DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning

DeepSeek-R1 shows an LLM can learn strong step-by-step reasoning from pure reinforcement learning, with no human-labeled reasoning examples.

2025/01/22 by DeepSeek-AI, Daya Guo, Guo, Daya +404 · 93 voices · 2191 citations
Computer Science · #Reinforcement Learning in Robotics #Data Stream Mining Techniques #Explainable Artificial Intelligence (XAI)

paper · pdf · doi:10.1038/s41586-025-09422-z

Abstract

Abstract General reasoning represents a long-standing and formidable challenge in artificial intelligence (AI). Recent breakthroughs, exemplified by large language models (LLMs) 1,2 and chain-of-thought (CoT) prompting 3 , have achieved considerable success on foundational reasoning tasks. However, this success is heavily contingent on extensive human-annotated demonstrations and the capabilities of models are still insufficient for more complex problems. Here we show that the reasoning abilities of LLMs can be incentivized through pure reinforcement learning (RL), obviating the need for human-labelled reasoning trajectories. The proposed RL framework facilitates the emergent development of advanced reasoning patterns, such as self-reflection, verification and dynamic strategy adaptation. Consequently, the trained model achieves superior performance on verifiable tasks such as mathematics, coding competitions and STEM fields, surpassing its counterparts trained through conventional supervised learning on human demonstrations. Moreover, the emergent reasoning patterns exhibited by these large-scale models can be systematically used to guide and enhance the reasoning capabilities of smaller models.

Summary

The authors train DeepSeek-V3-Base with reinforcement learning alone, rewarding only correct final answers and proper formatting, and find the model teaches itself long chain-of-thought reasoning, jumping from 15.6% to 77.9% pass@1 on the AIME 2024 math competition. Because this pure-RL model (DeepSeek-R1-Zero) is hard to read and mixes languages, they add a small set of human-curated examples plus further RL and fine-tuning stages to produce DeepSeek-R1, which keeps the reasoning gains while also handling writing and general instructions, and which they distill into smaller models.

machine-generated · claude-sonnet-5

Outline

• Introduction — Argues human-labeled reasoning traces cap model performance and proposes learning reasoning purely through RL instead.
• DeepSeek-R1-Zero — Trains DeepSeek-V3-Base with GRPO using only rule-based correctness and format rewards, skipping supervised fine-tuning entirely.
• DeepSeek-R1 — Adds cold-start human-curated data, a second RL stage, rejection sampling, and SFT to fix readability and language mixing while adding general-purpose skills.
• Experiment — Tracks benchmark scores across the pipeline stages (R1-Zero to Dev1/2/3 to final R1) on English, code, math, and Chinese-language tests.
• Ethics and Safety Statement — Flags jailbreak and misuse risks and reports R1's safety level as roughly comparable to GPT-4o.
• Conclusion, Limitation, and Future Work — Lists concrete gaps (tool use, token efficiency, language mixing, prompt sensitivity, software engineering) and names reward hacking as an open challenge for scaling pure RL.
• Appendix: Background and GRPO vs. PPO — Describes the DeepSeek-V3 base model and shows GRPO matching well-tuned PPO on a MATH task while avoiding a separate value model.
• Appendix: Training Details and RL Infrastructure — Describes the four-module RL system (rollout, inference, rule-based reward, training) and the data recipe used for RL prompts.

machine-generated · claude-sonnet-5

Claims

• Training DeepSeek-V3-Base with large-scale RL alone, rewarding only answer correctness and output format with no supervised fine-tuning step, causes reasoning behaviors like self-verification and reflection to emerge and raises AIME 2024 pass@1 from 15.6% to 77.9% (86.7% with self-consistency voting). [experiment]
• The pure-RL model DeepSeek-R1-Zero, despite strong reasoning, has poor output readability and mixes English and Chinese within single responses. [experiment]
• Adding a small human-curated cold-start dataset, a further RL stage, rejection sampling, and SFT on mixed reasoning/general data (the DeepSeek-R1 pipeline) fixes readability and language mixing while adding writing and general-instruction ability, without losing the reasoning gains. [experiment]
• The final DeepSeek-R1 model reaches 79.8% pass@1 on AIME 2024, 97.3% on MATH-500, a 2029 Codeforces rating, and 92.3% on ArenaHard, improving over every intermediate pipeline stage on most reasoning and general-instruction benchmarks. [experiment]
• Distilling DeepSeek-R1's reasoning behavior into smaller dense models gives those models stronger reasoning performance than their original instruction-tuned counterparts. [experiment]
• GRPO removes the need for a separate learned value model by computing advantage from relative rewards within a sampled group of outputs, and matches well-tuned PPO's performance on a MATH task while avoiding PPO's extra hyperparameter tuning and memory overhead. [experiment]
• A rule-based reward (final-answer correctness plus format compliance) avoids the reward hacking that neural, model-based reward models are prone to during large-scale RL training. [argument]
• Adding a language-consistency reward, based on the proportion of target-language words in the chain of thought, reduces language mixing but causes a slight degradation in raw reasoning performance. [experiment]

machine-generated · claude-sonnet-5

Key figure

Figure 1 — Two charts tracking DeepSeek-R1-Zero during pure reinforcement-learning training: its accuracy on the AIME 2024 math competition climbs from 15.6% to 77.9% (86.7% with majority-vote decoding), while its average response length steadily grows, showing the model teaching itself to 'think longer' on harder problems without being told to.

machine-generated · claude-sonnet-5

Glossary

Reinforcement learning (RL)

Training a model by rewarding good outputs and penalizing bad ones, rather than showing it labeled correct answers directly.

Group Relative Policy Optimization (GRPO)

An RL algorithm that scores each of several sampled outputs to the same question against each other, avoiding the need for a separately trained value model that standard PPO requires.

Chain-of-thought (CoT)

The step-by-step reasoning text a model writes out before giving its final answer.

Cold-start data

A small set of human-curated, readable reasoning examples used to warm-start training before large-scale RL begins.

Rejection sampling

Generating many candidate outputs, keeping only the correct or high-quality ones, and using those as training data.

Distillation

Training a smaller model to reproduce the outputs of a larger, more capable model.

Reward hacking

When a model finds a way to score well on its reward signal without actually doing what that reward was meant to measure.

Pass@1

The percentage of problems a model solves correctly on its first attempt.

machine-generated · claude-sonnet-5

Audience

ML researchers and engineers working on LLM post-training, reasoning, or RL fine-tuning, and infrastructure engineers building large-scale RL training systems.

prerequisites: Familiarity with LLM pretraining, supervised fine-tuning, and RLHF pipelines, Basic policy-gradient RL concepts such as PPO, KL regularization, and advantage estimation, Familiarity with reasoning benchmarks like AIME, MATH, Codeforces, and MMLU

machine-generated · claude-sonnet-5

Open questions

• Can DeepSeek-R1 be given reliable tool use, such as search engines or calculators, during reasoning, and would that improve its performance?
  The paper states R1 currently cannot use external tools and lists tool-augmented reasoning as a promising future direction, but does not test it.
• How can token efficiency be improved so the model stops generating excessive reasoning on simple questions?
  The authors note instances of 'overthinking' on easy tasks and call further optimization of token efficiency an open item.
• How can language mixing be resolved for languages other than Chinese and English?
  The paper says R1 is optimized only for Chinese and English and may still mix languages on queries in other languages, tracing this to the base model's training data, and states they aim to address it in future updates.
• Can a robust, non-gameable reward model be built for tasks like open-ended writing that lack a rule-based verifier, so pure RL can scale beyond reasoning tasks?
  The authors identify reward hacking as an inherent risk whenever reward comes from a trained model instead of fixed rules, and call scaling pure RL to such tasks 'an open challenge'.
• Can large-scale RL be applied efficiently to software-engineering tasks despite the long evaluation times such tasks require?
  The paper reports R1 shows little improvement over DeepSeek-V3 on software-engineering benchmarks because long evaluation times limited how much RL could be run there, and names this as future work.

machine-generated · claude-sonnet-5

Supplementary links

• https://huggingface.co/deepseek-ai [code]

machine-generated · claude-sonnet-5

Citations

• Are We Done with MMLU?
• Efficient Memory Management for Large Language Model Serving with PagedAttention
• Training language models to follow instructions with human feedback
• PyTorch: An Imperative Style, High-Performance Deep Learning Library

Cited by

• scPortrait integrates single-cell images into multimodal modeling
• Re-FORC: Adaptive Reward Prediction for Efficient Chain-of-Thought Reasoning
• Semiotic logical hexagon theory for LLM logical reasoning
• LeAct: Learning to Reason from Expert Actions
• Learning Obfuscations Of LLM Embedding Sequences: Stained Glass Transform
• Self-Guided Process Reward Optimization with Redefined Step-wise Advantage for Process Reinforcement Learning
• Learning to Reason for Factuality
• CHERRY: Compressed Hierarchical Experts with Recurrent Representational Yield
• Experience Augmented Policy Optimization for LLM Reasoning
• Statistical Early Stopping for Reasoning Models
• REGEN: Replay-recycling for Expert-to-Generalist distillation with Offline Reinforcement Learning
• Evidence-Grounded AI for Musculoskeletal Care
• Learning, Reasoning, Refinement: A Framework for Kahneman's Dual-System Intelligence in GUI Agents
• Efficient Multi-round LLM Inference over Disaggregated Serving
• RubricRL: Simple Generalizable Rewards for Text-to-Image Generation
• MeanFlowNFT: Bringing Forward-Process RL to Average-Velocity Generators
• ARMOR: Stabilizing On-Policy LLM RL with Off-Policy Anchor Samples
• MagicSelector: Joint Optimization for Agent Tool Selection via Counterfactual Decomposition and Progressive Reranking
• DORA: A Scalable Asynchronous Reinforcement Learning System for Language Model Training
• Omni-Perception Policy Optimization for Multimodal Emotion Reasoning
• VOPE: Revisiting Hallucination of Vision-Language Models in Voluntary Imagination Task
• TSRouter: Dynamic Modality-Model Selection for Time Series Reasoning
• PCTD: Preference-Guided Counterfactual Task Decomposition for Agent Tool Retrieval
• Schema-Constrained Document-Level Event Argument Extraction with Lightweight LLM Fine-Tuning
• RIMS: Preference Optimization via Smoothed Multi-pair Aggregation for Small-Scale LLM Retrieval-Augmented Generation
• Video Streaming Thinking: VideoLLMs Can Watch and Think Simultaneously
• Decoupled Alignment for Robust Plug-and-Play Adaptation
• Understanding Reasoning from Pretraining to Post-Training
• Orca: The World is in Your Mind
• VLT: A Vision-Language-Time Series Multimodal Foundation Model for Industrial Intelligence
• Ring-Zero: Scaling Zero RL to a Trillion Parameters for Emergent Reasoning
• Beyond Entropy: Correctness-Aware Advantage Shaping via Contrastive Policy Optimization
• Escaping Model Collapse via Synthetic Data Verification: Near-term Improvements and Long-term Convergence
• From Outcomes to Actions: Leveraging Hindsight for Long-Horizon Language Agent Training
• Shapley Context Pruning: A Cooperative Game Perspective for Context Reranking and Pruning
• AutoSurrogate: An LLM-driven multi-agent framework for autonomous construction of deep learning surrogate models in subsurface flow
• PR-CAD: Progressive Refinement for Unified Controllable and Faithful Text-to-CAD Generation with Large Language Models
• Success Conditioning as Policy Improvement: The Optimization Problem Solved by Imitating Success
• The Molecular Structure of Thought: Mapping the Topology of Long Chain-of-Thought Reasoning
• Beyond Converging Representations A Philosophical Response on the Interpretation Risks of Scientific Foundation Models
• Fifty Shades of Greenwashing: The Political Economy of Climate Change Advertising on Social Media
• Computational Turing Test Reveals Systematic Differences Between Human and AI Language
• Continuous Autoregressive Language Models
• Jasmine: A Simple, Performant and Scalable JAX-based World Modeling Codebase
• Kimi Linear: An Expressive, Efficient Attention Architecture
• Not All Bits Are Equal: Scale-Dependent Memory Optimization Strategies for Reasoning Models
• Expected Attention: KV Cache Compression by Estimating Attention from Future Queries Distribution
• Extract-0: A Specialized Language Model for Document Information Extraction
• Bootstrapping Task Spaces for Self-Improvement
• Information suppression in large language models: Auditing, quantifying, and characterizing censorship in DeepSeek
• K2-Think: A Parameter-Efficient Reasoning System
• GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models
• The Homogenizing Effect of Large Language Models on Human Expression and Thought
• Group Sequence Policy Optimization
• Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens
• Subliminal Learning: Language models transmit behavioral traits via hidden signals in data
• Seed-Prover: Deep and Broad Reasoning for Automated Theorem Proving
• Supervised Fine Tuning on Curated Data is Reinforcement Learning (and can be improved)
• Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety
• Reinforcement Learning Teachers of Test Time Scaling
• QiMeng: Fully Automated Hardware and Software Design for Processor Chip
• Small Language Models are the Future of Agentic AI
• Unsupervised Elicitation of Language Models
• Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens
• Reinforcement Pre-Training
• CRMArena-Pro: Holistic Assessment of LLM Agents Across Diverse Business Scenarios and Interactions
• Outcome-based Reinforcement Learning to Predict the Future
• LLMs Get Lost In Multi-Turn Conversation
• Breaking Quadratic Barriers: A Non-Attention LLM for Ultra-Long Context Horizons
• From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning
• Even Small Reasoners Should Quote Their Sources: Introducing the Pleias-RAG Model Family
• M1: Towards Scalable Test-Time Compute with Mamba Reasoning Models
• 70% Size, 100% Accuracy: Lossless LLM Compression for Efficient GPU Inference via Dynamic-Length Float (DFloat11)
• OmniSVG: A Unified Scalable Vector Graphics Generation Model
• Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search
• Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction
• VAPO: Efficient and Reliable Reinforcement Learning for Advanced Reasoning Tasks
• Inference-Time Scaling for Generalist Reward Modeling
• How Deep Do Large Language Models Internalize Scientific Literature and Citation Practices?
• Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models
• Transformers without Normalization
• Proof or Bluff? Evaluating LLMs on 2025 USA Math Olympiad
• TransMLA: Multi-Head Latent Attention Is All You Need
• Automated Capability Discovery via Foundation Model Self-Exploration
• LLMs Can Teach Themselves to Better Predict the Future
• LADDER: Self-Improving LLMs Through Recursive Problem Decomposition
• General Intelligence Requires Reward-based Pretraining
• Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention
• LIMO: Less is More for Reasoning
• Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs
• SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training
• Context-Selective State Space Models: Feedback is All You Need
• TimeSearch-R: Adaptive Temporal Search for Long-Form Video Understanding via Self-Verification Reinforcement Learning
• ATGen: Adversarial Reinforcement Learning for Test Case Generation
• Java academic benchmark: Exam-based evaluation of LLMs on object-oriented programming
• ProGuard: Towards Proactive Multimodal Safeguard
• AKG kernel Agent: A Multi-Agent Framework for Cross-Platform Kernel Synthesis
• LLM4Branch: Large Language Model for Discovering Efficient Branching Policies of Integer Programs
• Diversity or Precision? A Deep Dive into Next Token Prediction
• MIRA: Multimodal Iterative Reasoning Agent for Image Editing
• Remedy-R: Generative Reasoning for Machine Translation Evaluation without Error Annotations
• A Structured Clustering Approach for Inducing Media Narratives
• Reasoning Gets Harder for LLMs Inside A Dialogue
• Reinforcement Learning via Self-Distillation
• World Engine: Towards the Era of Post-Training for Autonomous Driving
• Compared to What? Baselines and Metrics for Counterfactual Prompting
• OmniOPD: Logit-Free On-Policy Distillation via Speculative Verification
• Argus: Token Aware Distributed LLM Inference Optimization
• OptiNIC: A Resilient and Tail-Optimal RDMA NIC for Distributed ML Workloads
• DICE: Discrete Interpretable Comparative Evaluation with Probabilistic Scoring for Retrieval-Augmented Generation
• Agent2World: Learning to Generate Symbolic World Models via Adaptive Multi-Agent Feedback
• SmartSnap: Proactive Evidence Seeking for Self-Verifying Agents
• VideoZoomer: Reinforcement-Learned Temporal Focusing for Long Video Reasoning
• LiveProteinBench: A Contamination-Free Benchmark for Assessing Models' Specialized Capabilities in Protein Science
• Shape of Thought: When Distribution Matters More than Correctness in Reasoning Tasks
• Calibrating LLM Judges: Linear Probes for Fast and Reliable Uncertainty Estimation
• Masking Teacher and Reinforcing Student for Distilling Vision-Language Models
• Towards Signboard-Oriented Visual Question Answering: ViSignVQA Dataset, Method and Benchmark
• Open-Source Multimodal Moxin Models with Moxin-VLM and Moxin-VLA
• Orthrus: Memory-Efficient Parallel Token Generation via Dual-View Diffusion
• MarineEVT: Advancing Event-Centric Marine Video Understanding via Visual Tool Reasoning
• Hybrid Advantage Estimation with Unified Critic for VLM Agentic Reinforcement Learning
• Express Language Modeling
• Scale Weight Decay and Train Better
• Performance of AI agents based on reasoning language models on ALD process optimization tasks
• AI Empowered Communication and Radar Modulation Recognition: A Survey
• Self-Boosting Vision-Language Models with Noisy Student On-Policy Self-Distillation
• Can LLMs Write Correct TLA+ Specifications? Evaluating Natural-Language-to-TLA+ Generation
• DeepLook: Deeper Thinking with Lookahead
• Do LLMs Experience an Internal Polylogue? Investigating Reasoning through the Lens of Personas
• LanteRn: Latent Visual Structured Reasoning
• MAI-UI Technical Report: Real-World Centric Foundation GUI Agents
• El Agente Estructural: An Artificially Intelligent Molecular Editor
• SWE-RM: Execution-free Feedback For Software Engineering Agents
• Exploring the Heterogeneity of Tabular Data: A Diversity-aware Data Generator via LLMs
• RLLaVA: An RL-central Framework for Language and Vision Assistants
• HELP: Hierarchical Embodied Language Planner for Household Tasks
• RoboSafe: Safeguarding Embodied Agents via Executable Safety Logic
• Rethinking Supervised Fine-Tuning: Emphasizing Key Answer Tokens for Improved LLM Accuracy
• Pioneering Multimodal Emotion Recognition in the Era of Large Models: From Closed Sets to Open Vocabularies
• PhononBench:A Large-Scale Phonon-Based Benchmark for Dynamical Stability in Crystal Generation
• A Real-World Evaluation of LLM Medication Safety Reviews in NHS Primary Care
• From Pilots to Practices: A Scoping Review of GenAI-Enabled Personalization in Computer Science Education
• Learning to Reason in 4D: Dynamic Spatial Understanding for Vision Language Models
• TableGPT-R1: Advancing Tabular Reasoning Through Reinforcement Learning
• Memory-T1: Reinforcement Learning for Temporal Reasoning in Multi-session Agents
• BacAlarm: Mining and Simulating Composite API Traffic to Prevent Broken Access Control Violations
• SpatialTree: How Spatial Abilities Branch Out in MLLMs
• Reaching Agreement Among Reasoning LLM Agents
• From Retrieval to Reasoning: A Framework for Cyber Threat Intelligence NER with Explicit and Adaptive Instructions
• PhysMaster: Building an Autonomous AI Physicist for Theoretical and Computational Physics Research
• From Indoor to Open World: Revealing the Spatial Reasoning Gap in MLLMs
• VA-π: Variational Policy Alignment for Pixel-Aware Autoregressive Generation
• Emotion-Director: Bridging Affective Shortcut in Emotion-Oriented Image Generation
• ReasonCD: A Multimodal Reasoning Large Model for Implicit Change-of-Interest Semantic Mining
• VisionDirector: Vision-Language Guided Closed-Loop Refinement for Generative Image Synthesis
• OmniMoGen: Unifying Human Motion Generation via Learning from Interleaved Text-Motion Instructions
• FC-MIR: A Mobile Screen Awareness Framework for Intent-Aware Recommendation based on Frame-Compressed Multimodal Trajectory Reasoning
• MEEA: Mere Exposure Effect-Driven Confrontational Optimization for LLM Jailbreaking
• Structural Reinforcement Learning for Heterogeneous Agent Macroeconomics
• ESearch-R1: Learning Cost-Aware MLLM Agents for Interactive Embodied Search via Reinforcement Learning
• Stable and Efficient Single-Rollout RL for Multimodal Reasoning
• Sophia: A Persistent Agent Framework of Artificial Life
• Shuttling Compiler for Trapped-Ion Quantum Computers Based on Large Language Models
• Adversarial Robustness of Vision in Open Foundation Models
• When Reasoning Meets Its Laws
• GroundingME: Exposing the Visual Grounding Gap in MLLMs through Multi-Dimensional Evaluation
• Xiaomi MiMo-VL-Miloco Technical Report
• Learning When to Look: A Disentangled Curriculum for Strategic Perception in Multimodal Reasoning
• Reasoning Palette: Modulating Reasoning via Latent Contextualization for Controllable Exploration for (V)LMs
• Trust-Region Adaptive Policy Optimization
• A Solver-in-the-Loop Framework for Improving LLMs on Answer Set Programming for Logic Puzzle Solving
• Exploration vs Exploitation: Rethinking RLVR through Clipping, Entropy, and Spurious Reward
• Posterior Behavioral Cloning: Pretraining BC Policies for Efficient RL Finetuning
• AdaSearch: Balancing Parametric Knowledge and Search in Large Language Models via Reinforcement Learning
• JustRL: Scaling a 1.5B LLM with a Simple RL Recipe
• Guiding Perception-Reasoning Closer to Human in Blind Image Quality Assessment
• TimeSeries2Report prompting enables adaptive large language model management of lithium-ion batteries
• Synthelite: Chemist-aligned and feasibility-aware synthesis planning with LLMs
• Pretrained battery transformer (PBT): A foundation model for battery life prediction
• Sigma-MoE-Tiny Technical Report
• In-Context Semi-Supervised Learning
• Bolmo: Byteifying the Next Generation of Language Models
• Nemotron-Math: Efficient Long-Context Distillation of Mathematical Reasoning from Multi-Mode Supervision
• CangLing-KnowFlow: A Unified Knowledge-and-Flow-fused Agent for Comprehensive Remote Sensing Applications
• EagleVision: A Dual-Stage Framework with BEV-grounding-based Chain-of-Thought for Spatial Intelligence
• Will AI Trade? A Computational Inversion of the No-Trade Theorem
• TimeLens: Rethinking Video Temporal Grounding with Multimodal LLMs
• Spherical Leech Quantization for Visual Tokenization and Generation
• LLM-driven Knowledge Enhancement for Multimodal Cancer Survival Prediction
• Learning to Extract Context for Context-Aware LLM Inference
• RecGPT-V2 Technical Report
• Zoom-Zero: Reinforced Coarse-to-Fine Video Understanding via Temporal Zoom-in
• Estimating problem difficulty without ground truth using Large Language Model comparisons
• Improving Semantic Uncertainty Quantification in LVLMs with Semantic Gaussian Processes
• RADAR: Accelerating Large Language Model Inference With RL-Based Dynamic Draft Trees
• What Affects the Effective Depth of Large Language Models?
• Autonomous Construction-Site Safety Inspection Using Mobile Robots: A Multilayer VLM-LLM Pipeline
• Sharing State Between Prompts and Programs
• A4-Agent: An Agentic Framework for Zero-Shot Affordance Reasoning
• CAPE: Capability Achievement via Policy Execution
• A Scientific Reasoning Model for Organic Synthesis Procedure Generation
• Soul: Breathe Life into Digital Human for High-fidelity Long-term Multimodal Animation
• SocialNav-MoE: A Mixture-of-Experts Vision Language Model for Socially Compliant Navigation with Reinforcement Fine-Tuning
• ORIBA: Exploring LLM-Driven Role-Play Chatbot as a Creativity Support Tool for Original Character Artists
• TraPO: A Semi-Supervised Reinforcement Learning Framework for Boosting LLM Reasoning
• ADHint: Adaptive Hints with Difficulty Priors for Reinforcement Learning
• State over Tokens: Characterizing the Role of Reasoning Tokens
• Anchoring Values in Temporal and Group Dimensions for Flow Matching Model Alignment
• Reassessing the Role of Supervised Fine-Tuning: An Empirical Study in VLM Reasoning
• DiG: Differential Grounding for Enhancing Fine-Grained Perception in Multimodal Large Language Model
• Exact Flow Linear Attention: Exact Solution from Continuous-Time Dynamics
• DL³M: A Vision-to-Language Framework for Expert-Level Medical Reasoning through Deep Learning and Large Language Models
• More Than the Final Answer: Improving Visual Extraction and Logical Consistency in Vision-Language Models
• SCIR: A Self-Correcting Iterative Refinement Framework for Enhanced Information Extraction Based on Schema
• Hold Onto That Thought: Assessing KV Cache Compression On Reasoning
• LYNX: Learning Dynamic Exits for Confidence-Controlled Reasoning
• Active Video Perception: Iterative Evidence Seeking for Agentic Long Video Understanding
• RollMux: Phase-Level Multiplexing for Disaggregated RL Post-Training
• Evolutionary Reinforcement Learning based AI tutor for Socratic Interdisciplinary Instruction
• FutureX: Enhance End-to-End Autonomous Driving via Latent Chain-of-Thought World Model
• DentalGPT: Incentivizing Multimodal Complex Reasoning in Dentistry
• CLINIC: Evaluating Multilingual Trustworthiness in Language Models for Healthcare
• Long-horizon Reasoning Agent for Olympiad-Level Mathematical Problem Solving
• Enhancing Radiology Report Generation and Visual Grounding using Reinforcement Learning
• On the Dynamics of Multi-Agent LLM Communities Driven by Value Diversity
• RoleRMBench & RoleRM: Towards Reward Modeling for Profile-Based Role Play in Dialogue Systems
• Boosting RL-Based Visual Reasoning with Selective Adversarial Entropy Intervention
• Translating Informal Proofs into Formal Proofs Using a Chain of States
• Are We Ready for RL in Text-to-3D Generation? A Progressive Investigation
• Mull-Tokens: Modality-Agnostic Latent Thinking
• SCOPE: Language Models as One-Time Teacher for Hierarchical Planning in Text Environments
• ChronusOmni: Improving Time Awareness of Omni Large Language Models
• Can LLMs Evaluate What They Cannot Annotate? Revisiting LLM Reliability in Hate Speech Detection
• Rethinking Chain-of-Thought Reasoning for Videos
• Video-QTR: Query-Driven Temporal Reasoning Framework for Lightweight Video Understanding
• ARM-Thinker: Reinforcing Multimodal Generative Reward Models with Agentic Tool Use and Visual Reasoning
• LaMoSys3.5D: Enabling 3.5D-IC-Based Large Language Model Inference Serving Systems via Hardware/Software Co-Design
• Soft Inductive Bias Approach via Explicit Reasoning Perspectives in Inappropriate Utterance Detection Using Large Language Models
• TreeGRPO: Tree-Advantage GRPO for Online RL Post-Training of Diffusion Models
• SpatialDreamer: Incentivizing Spatial Reasoning via Active Mental Imagery
• No Labels, No Problem: Training Visual Reasoners with Multimodal Verifiers
• ValuePilot: A Two-Phase Framework for Value-Driven Decision-Making
• Scalable Offline Model-Based RL with Action Chunks
• Bridging the Clinical Expertise Gap: Development of a Web-Based Platform for Accessible Time Series Forecasting and Analysis
• Large Causal Models from Large Language Models
• Each Prompt Matters: Scaling Reinforcement Learning Without Wasting Rollouts on Hundred-Billion-Scale MoE
• Comparative Analysis and Parametric Tuning of PPO, GRPO, and DAPO for LLM Reasoning Enhancement
• ReLaX: Reasoning with Latent Exploration for Large Reasoning Models
• AutoICE: Automatically Synthesizing Verifiable C Code via LLM-driven Evolution
• CFD-copilot: leveraging domain-adapted large language model and model context protocol to enhance simulation automation
• Native Parallel Reasoner: Reasoning in Parallelism via Self-Distilled Reinforcement Learning
• From Show Programmes to Data: Designing a Workflow to Make Performing Arts Ephemera Accessible Through Language Models
• Training Language Models to Use Prolog as a Tool
• START: Spatial and Textual Learning for Chart Understanding
• Living the Novel: A System for Generating Self-Training Timeline-Aware Conversational Agents from Novels
• VulnLLM-R: Specialized Reasoning LLM with Agent Scaffold for Vulnerability Detection
• DiffusionDriveV2: Reinforcement Learning-Constrained Truncated Diffusion Modeling in End-to-End Autonomous Driving
• RVLF: A Reinforcing Vision-Language Framework for Gloss-Free Sign Language Translation
• JT-DA: Enhancing Data Analysis with Tool-Integrated Table Reasoning Large Language Models
• Decouple to Generalize: Context-First Self-Evolving Learning for Data-Scarce Vision-Language Reasoning
• LLM as a Neural Architect: Controlled Generation of Image Captioning Models Under Strict API Contracts
• The Role of Entropy in Visual Grounding: Analysis and Optimization
• PersonaMem-v2: Towards Personalized Intelligence via Learning Implicit User Personas and Agentic Memory
• RLAX: Large-Scale, Distributed Reinforcement Learning for Large Language Models on TPUs
• VG-Refiner: Towards Tool-Refined Referring Grounded Reasoning via Agentic Reinforcement Learning
• Knowing the Answer Isn't Enough: Fixing Reasoning Path Failures in LVLMs
• Beyond Token-level Supervision: Unlocking the Potential of Decoding-based Regression via Reinforcement Learning
• Trusted AI Agents in the Cloud
• VRSA: Jailbreaking Multimodal Large Language Models through Visual Reasoning Sequential Attack
• Ontology Learning with LLMs: A Benchmark Study on Axiom Identification
• Training Multi-Image Vision Agents via End2End Reinforcement Learning
• AI & Human Co-Improvement for Safer Co-Superintelligence
• STARE-VLA: Progressive Stage-Aware Reinforcement for Fine-Tuning Vision-Language-Action Models
• Visual Reasoning Tracer: Object-Level Grounded Reasoning Benchmark
• Arbitrage: Efficient Reasoning via Advantage-Aware Speculation
• RefineBench: Evaluating Refinement Capability of Language Models via Checklists
• Autoregressive Image Generation Needs Only a Few Lines of Cached Tokens
• Language Models as Semantic Teachers: Post-Training Alignment for Medical Audio Understanding
• BioMedGPT-Mol: Multi-task Learning for Molecular Understanding and Generation
• EtCon: Edit-then-Consolidate for Reliable Knowledge Editing
• Efficient Reinforcement Learning with Semantic and Token Entropy for LLM Reasoning
• CryptoTensors: A Light-Weight Large Language Model File Format for Highly-Secure Model Distribution
• Principled RL for Diffusion LLMs Emerges from a Sequence-Level Perspective
• VideoMem: Enhancing Ultra-Long Video Understanding via Adaptive Memory Management
• dVLM-AD: Enhance Diffusion Vision-Language-Model for Driving via Controllable Reasoning
• Semantic Soft Bootstrapping: Long Context Reasoning in LLMs without Reinforcement Learning
• Decoding Large Language Diffusion Models with Foreseeing Movement
• TempR1: Improving Temporal Understanding of MLLMs via Temporal-Aware Multi-Task Reinforcement Learning
• KVNAND: Efficient On-Device Large Language Model Inference Using DRAM-Free In-Flash Computing
• CoRT: Code-integrated Reasoning within Thinking
• PretrainZero: Reinforcement Active Pretraining
• MarkTune: Improving the Quality-Detectability Trade-off in Open-Weight LLM Watermarking
• SPARK: Stepwise Process-Aware Rewards for Reference-Free Reinforcement Learning
• Plantain: Plan-Answer Interleaved Reasoning
• GeoZero: Incentivizing Reasoning from Scratch on Geospatial Scenes
• Fairy2i: Training Complex LLMs from Real LLMs with All Parameters in \± 1, ± i\
• MindGPT-4ov: An Enhanced MLLM via a Multi-Stage Post-Training Paradigm
• A benchmark dataset for evaluating Syndrome Differentiation and Treatment in large language models
• SR-GRPO: Stable Rank as an Intrinsic Geometric Reward for Large Language Model Alignment
• CryptoQA: A Large-scale Question-answering Dataset for AI-assisted Cryptography
• Learning What to Attend First: Modality-Importance-Guided Reasoning for Reliable Multimodal Emotion Understanding
• RULER-Bench: Probing Rule-based Reasoning Abilities of Next-level Video Generation Models for Vision Foundation Intelligence
• DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
• Skywork-R1V4: Toward Agentic Multimodal Intelligence through Interleaved Thinking with Images and DeepResearch
• Process-Centric Analysis of Agentic Software Systems
• Synthetic Error Injection Fails to Elicit Self-Correction In Language Models
• VACoT: Rethinking Visual Data Augmentation with VLMs
• Think Before You Prune: Self-Reflective Structured Pruning for Reasoning Language Models
• Think in Parallel, Answer as One: Logit Averaging for Open-Ended Reasoning
• ReVSeg: Incentivizing the Reasoning Chain for Video Segmentation with Reinforcement Learning
• Self-Improving AI Agents through Self-Play
• Beyond SFT: Reinforcement Learning for Safer Large Reasoning Models with Better Reasoning Ability
• OpenREAD: Reinforced Open-Ended Reasoning for End-to-End Autonomous Driving with LLM-as-Critic
• Edge Deployment of Small Language Models, a comprehensive comparison of CPU, GPU and NPU backends
• Beware of Reasoning Overconfidence: Pitfalls in the Reasoning Process for Multi-solution Tasks
• Zero-Overhead Introspection for Adaptive Test-Time Compute
• Stabilizing Reinforcement Learning with LLMs: Formulation and Practices
• FishDetector-R1: Unified MLLM-Based Framework with Reinforcement Fine-Tuning for Weakly Supervised Fish Detection, Segmentation, and Counting
• The Tragedy of Productivity: A Unified Framework for Diagnosing Coordination Failures in Labor Markets and AI Governance
• Kardia-R1: Unleashing LLMs to Reason toward Understanding and Empathy for Emotional Support via Rubric-as-Judge Reinforcement Learning
• Accelerating Large-Scale Reasoning Model Inference with Sparse Self-Speculative Decoding
• SUPERChem: A Multimodal Reasoning Benchmark in Chemistry
• On the Tension Between Optimality and Adversarial Robustness in Policy Optimization
• Unsupervised decoding of encoded reasoning using language model interpretability
• IC-World: In-Context Generation for Shared World Modeling
• CoSineVerifier: Tool-Augmented Answer Verification for Computation-Oriented Scientific Questions
• Mode-Conditioning Unlocks Superior Test-Time Scaling
• SimWorld: An Open-ended Realistic Simulator for Autonomous Agents in Physical and Social Worlds
• Crystalyse: a multi-tool agent for materials design
• Table as a Modality for Large Language Models
• Elastic Mixture of Rank-Wise Experts for Knowledge Reuse in Federated Fine-Tuning
• Multilingual Training-Free Remote Sensing Image Captioning
• ReJump: A Tree-Jump Representation for Analyzing and Improving LLM Reasoning
• IRPO: Boosting Image Restoration via Post-training GRPO
• Thinking with Drafts: Speculative Temporal Reasoning for Efficient Long Video Understanding
• Multi-GRPO: Multi-Group Advantage Estimation for Text-to-Image Generation with Tree-Based Trajectories and Multiple Rewards
• G-KV: Decoding-Time KV Cache Eviction with Global Attention
• VCWorld: A Biological World Model for Virtual Cell Simulation
• What about gravity in video generation? Post-Training Newton's Laws with Verifiable Rewards
• Video-R2: Reinforcing Consistent and Grounded Reasoning in Multimodal Language Models
• Thinking by Doing: Building Efficient World Model Reasoning in LLMs via Multi-turn Interaction
• ThetaEvolve: Test-time Learning on Open Problems
• OctoMed: Data Recipes for State-of-the-Art Multimodal Medical Reasoning
• Learning to Refuse: Refusal-Aware Reinforcement Fine-Tuning for Hard-Irrelevant Queries in Video Temporal Grounding
• McSc: Motion-Corrective Preference Alignment for Video Generation with Self-Critic Hierarchical Reasoning
• Adversarial Training for Process Reward Models
• FLIMs: Fault Localization Interference Mutants, Definition, Recognition and Mitigation
• Closed-Loop Transformers: Autoregressive Modeling as Iterative Latent Equilibrium
• Improving Robotic Manipulation Robustness via NICE Scene Surgery
• Revisiting the Necessity of Lengthy Chain-of-Thought in Vision-centric Reasoning Generalization
• RoadSceneBench: A Lightweight Benchmark for Mid-Level Road Scene Understanding
• ITS3D: Inference-Time Scaling for Text-Guided 3D Diffusion Models
• OmniInfer: System-Wide Acceleration Techniques for Optimizing LLM Serving Throughput and Latency
• Asking like Socrates: Socrates helps VLMs understand remote sensing images
• ToolOrchestra: Elevating Intelligence via Efficient Model and Tool Orchestration
• TreeCoder: Systematic Exploration and Optimisation of Decoding and Constraints for LLM Code Generation
• Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation
• Focused Chain-of-Thought: Efficient LLM Reasoning via Structured Input Information
• MoGAN: Improving Motion Quality in Video Diffusion via Few-Step Motion Adversarial Post-Training
• Guiding Generative Models for Protein Design: Prompting, Steering and Aligning
• Scaling Generative Verifiers For Natural Language Mathematical Proof Verification And Selection
• REVISOR: Beyond Textual Reflection, Towards Multimodal Introspective Reasoning in Long-Form Video Understanding
• Can LLMs extract human-like fine-grained evidence for evidence-based fact-checking?
• Thinking With Bounding Boxes: Enhancing Spatio-Temporal Video Grounding via Reinforcement Fine-Tuning
• Self-Guided Defense: Adaptive Safety Alignment for Reasoning Models via Synthesized Guidelines
• SocialNav: Training Human-Inspired Foundation Model for Socially-Aware Embodied Navigation
• Aligning LLMs with Biomedical Knowledge using Balanced Fine-Tuning
• GuardTrace-VL: Detecting Unsafe Multimodel Reasoning via Iterative Safety Supervision
• Unsupervised Memorability Modeling from Tip-of-the-Tongue Retrieval Queries
• OVAL-Grasp: Open-Vocabulary Affordance Localization for Task Oriented Grasping
• A Reason-then-Describe Instruction Interpreter for Controllable Video Generation
• From Words to Wisdom: Discourse Annotation and Baseline Models for Student Dialogue Understanding
• Large Language Models' Complicit Responses to Illicit Instructions across Socio-Legal Contexts
• AD-R1: Closed-Loop Reinforcement Learning for End-to-End Autonomous Driving with Impartial World Models
• EmoFeedback²: Reinforcement of Continuous Emotional Image Generation via LVLM-based Reward and Textual Feedback
• HiCoGen: Hierarchical Compositional Text-to-Image Generation in Diffusion Models via Reinforcement Learning
• Differential Smoothing Mitigates Sharpening and Improves LLM Reasoning
• Reasoning-VLA: A Fast and General Vision-Language-Action Reasoning Model for Autonomous Driving
• Syn-GRPO: Self-Evolving Data Synthesis for MLLM Perception Reasoning
• Learning to Reason: Training LLMs with GPT-OSS or DeepSeek R1 Reasoning Traces
• LongVT: Incentivizing "Thinking with Long Videos" via Native Tool Calling
• Soft Adaptive Policy Optimization
• Scaling Agentic Reinforcement Learning for Tool-Integrated Reasoning in VLMs
• Learning Massively Multitask World Models for Continuous Control
• Be My Eyes: Extending Large Language Models to New Modalities Through Multi-Agent Collaboration
• Learning Robust Social Strategies with Large Language Models
• VideoChat-M1: Collaborative Policy Planning for Video Understanding via Multi-Agent Reinforcement Learning
• A Longitudinal Measurement of Privacy Policy Evolution for Large Language Models
• SmartPoC: Generating Executable and Validated PoCs for Smart Contract Bug Reports
• Periodic Asynchrony: An On-Policy Approach for Accelerating LLM Reinforcement Learning
• Seeing What Matters: Visual Preference Policy Optimization for Visual Generation
• LLM4AMC: Adapting Large Language Models for Adaptive Modulation and Coding
• Beyond Description: Cognitively Benchmarking Fine-Grained Action for Embodied Agents
• Learning What to Trust: Bayesian Prior-Guided Optimization for Visual Generation
• VADE: Variance-Aware Dynamic Sampling via Online Sample-Level Difficulty Estimation for Multimodal RL
• Leveraging Language Models for Interpretable Analysis of Narratives in a Large Corpus
• Foundations of Artificial Intelligence Frameworks: Notion and Limits of AGI
• MindEval: Benchmarking Language Models on Multi-turn Mental Health Support
• Decoupling Perception from Reasoning for Hallucination-Resistant Video Understanding
• Perceptual-Evidence Anchored Reinforced Learning for Multimodal Reasoning
• DiVE-k: Differential Visual Reasoning for Fine-grained Image Recognition
• DiscoVerse: Multi-Agent Pharmaceutical Co-Scientist for Traceable Drug Discovery and Reverse Translation
• EgoVITA: Learning to Plan and Verify for Egocentric Video Reasoning
• The Alignment Paradox of Medical Large Language Models in Infertility Care: Decoupling Algorithmic Improvement from Clinical Decision-making Quality
• IE-Critic-R1: Advancing the Explanatory Measurement of Text-Driven Image Editing for Human Perception Alignment
• SPINE: Token-Selective Test-Time Reinforcement Learning with Entropy-Band Regularization
• MobileVLA-R1: Reinforcing Vision-Language-Action for Mobile Robots
• Plan-X: Instruct Video Generation via Semantic Planning
• Video-R4: Reinforcing Text-Rich Video Reasoning with Visual Rumination
• Masked-and-Reordered Self-Supervision for Reinforcement Learning from Verifiable Rewards
• Intrinsic preservation of plasticity in continual quantum learning
• Can MLLMs Read the Room? A Multimodal Benchmark for Assessing Deception in Multi-Party Social Interactions
• Steering in the Shadows: Causal Amplification for Activation Space Attacks in Large Language Models
• FireScope: Wildfire Risk Raster Prediction with a Chain-of-Thought Oracle
• Dual-Path Knowledge-Augmented Contrastive Alignment Network for Spatially Resolved Transcriptomics
• ChainV: Atomic Visual Hints Make Multimodal Reasoning Shorter and Better
• Do Vision-Language Models Understand Visual Persuasiveness?
• Exploring Scientific Debt: Harnessing AI for SATD Identification in Scientific Software
• Goal-Directed Search Outperforms Goal-Agnostic Memory Compression in Long-Context Memory Tasks
• Learning to Think Fast and Slow for Visual Language Models
• You Only Forward Once: An Efficient Compositional Judging Paradigm
• ODE-ViT: Plug & Play Attention Layer from the Generalization of the ViT as an Ordinary Differential Equation
• SDA: Steering-Driven Distribution Alignment for Open LLMs without Fine-Tuning
• "To Survive, I Must Defect": Jailbreaking LLMs via the Game-Theory Scenarios
• Fast LLM Post-training via Decoupled and Fastest-of-N Speculation
• SkyRL-Agent: Efficient RL Training for Multi-turn LLM Agent
• KRAL: Knowledge and Reasoning Augmented Learning for LLM-assisted Clinical Antimicrobial Therapy
• Reasoning Guided Embeddings: Leveraging MLLM Reasoning for Improved Multimodal Retrieval
• OpenMMReasoner: Pushing the Frontiers for Multimodal Reasoning with an Open and General Recipe
• Step-Audio-R1 Technical Report
• Explore How to Inject Beneficial Noise in MLLMs
• SRPO: Self-Referential Policy Optimization for Vision-Language-Action Models
• Enhancing Reliability across Short and Long-Form QA via Reinforcement Learning
• Efficiency Will Not Lead to Sustainable Reasoning AI
• GRPO-RM: Fine-Tuning Representation Models via GRPO-Driven Reinforcement Learning
• SOLID: a Framework of Synergizing Optimization and LLMs for Intelligent Decision-Making
• From Solving to Verifying: A Unified Objective for Robust Reasoning in LLMs
• Reasoning via Video: The First Evaluation of Video Models' Reasoning Abilities through Maze-Solving Tasks
• Seer: Online Context Learning for Fast Synchronous LLM Reinforcement Learning
• GPS: General Per-Sample Prompter
• Extending Test-Time Scaling: A 3D Perspective with Context, Batch, and Turn
• From Topology to Behavioral Semantics: Enhancing BGP Security by Understanding BGP's Language with LLMs
• DIR-TIR: Dialog-Iterative Refinement for Text-to-Image Retrieval
• Beyond Surface-Level Similarity: Hierarchical Contamination Detection for Synthetic Training Data in Foundation Models
• Knowledge-Grounded Agentic Large Language Models for Multi-Hazard Understanding from Reconnaissance Reports
• Can World Simulators Reason? Gen-ViRe: A Generative Visual Reasoning Benchmark
• TiViBench: Benchmarking Think-in-Video Reasoning for Video Generative Models
• P1: Mastering Physics Olympiads with Reinforcement Learning
• Grounded by Experience: Generative Healthcare Prediction Augmented with Hierarchical Agentic Retrieval
• Multi-Agent Deep Research: Training Multi-Agent Systems with M-GRPO
• PIGEON: VLM-Driven Object Navigation via Points of Interest Selection
• ViSS-R1: Self-Supervised Reinforcement Video Reasoning
• From Perception to Reasoning: Deep Thinking Empowers Multimodal Large Language Models
• Enhancing Diffusion-based Restoration Models via Difficulty-Adaptive Reinforcement Learning with IQA Reward
• TPS-Bench: Evaluating AI Agents' Tool Planning & Scheduling Abilities in Compounding Tasks
• BARD: budget-aware reasoning distillation
• HMVLM: Human Motion-Vision-Lanuage Model via MoE LoRA
• LiveSearchBench: An Automatically Constructed Benchmark for Retrieval and Reasoning over Dynamic Knowledge
• Reproducibility Report: Test-Time Training on Nearest Neighbors for Large Language Models
• Fast Reasoning Segmentation for Images and Videos
• Constructing and Interpreting Digital Twin Representations for Visual Reasoning via Reinforcement Learning
• AttackVLA: Benchmarking Adversarial and Backdoor Attacks on Vision-Language-Action Models
• Image-POSER: Reflective RL for Multi-Expert Image Generation and Editing
• On the Notion that Language Models Reason
• Honesty over Accuracy: Trustworthy Language Models through Reinforced Hesitation
• Large Language Models and 3D Vision for Intelligent Robotic Perception and Autonomy
• DiscoX: Benchmarking Discourse-Level Translation task in Expert Domains
• Reinforcing Trustworthiness in Multimodal Emotional Support Systems
• AdvancedIF: Rubric-Based Benchmarking and Reinforcement Learning for Advancing LLM Instruction Following
• Enhancing the Medical Context-Awareness Ability of LLMs via Multifaceted Self-Refinement Learning
• Physical Plausibility Reasoning via HCM-GRPO: Empowering Compact Model for Superior Performance
• SCARE: A Benchmark for SQL Correction and Question Answerability Classification for Reliable EHR Question Answering
• Uncovering Strategic Egoism Behaviors in Large Language Models
• Learning to Pose Problems: Reasoning-Driven and Solver-Adaptive Data Synthesis for Large Reasoning Models
• Uncertainty-Guided Checkpoint Selection for Reinforcement Finetuning of Large Language Models
• MoFa: A Unified Performance Modeling Framework for LLM Pretraining
• From Street to Orbit: Training-Free Cross-View Retrieval via Location Semantics and LLM Guidance
• MMaDA-Parallel: Multimodal Large Diffusion Language Models for Thinking-Aware Editing and Generation
• ToolMind Technical Report: A Large-Scale, Reasoning-Enhanced Tool-Use Dataset
• Hey Pentti, We Did (More of) It!: A Vector-Symbolic Lisp With Residue Arithmetic
• ParaS2S: Benchmarking and Aligning Spoken Language Models for Paralinguistic-aware Speech-to-Speech Interaction
• The Path Not Taken: RLVR Provably Learns Off the Principals
• Temporal Predictors of Outcome in Reasoning Language Models
• The Open Syndrome Definition as a Machine-Readable Standard for Public Health: Design and Implementation Study
• Towards General Auditory Intelligence: Large Multimodal Models for Machine Listening and Speaking
• Knowledge-Augmented Long-CoT Generation for Complex Biomolecular Reasoning
• An Efficient Training Pipeline for Reasoning Graphical User Interface Agents
• MSCR: Exploring the Vulnerability of LLMs' Mathematical Reasoning Abilities Using Multi-Source Candidate Replacement
• Numerical Sensitivity and Robustness: Exploring the Flaws of Mathematical Reasoning in Large Language Models
• NOTAM-Evolve: A Knowledge-Guided Self-Evolving Optimization Framework with LLMs for NOTAM Interpretation
• Benchmarking Multi-Step Legal Reasoning and Analyzing Chain-of-Thought Effects in Large Language Models
• Visual Bridge: Universal Visual Perception Representations Generating
• From Exploration to Exploitation: A Two-Stage Entropy RLVR Approach for Noise-Tolerant MLLM Training
• Auto-US: An Ultrasound Video Diagnosis Agent Using Video Classification Framework and LLMs
• SpeechJudge: Towards Human-Level Judgment for Speech Naturalness
• PCRLLM: Proof-Carrying Reasoning with Large Language Models under Stepwise Logical Constraints
• Think Consistently, Reason Efficiently: Energy-Based Calibration for Implicit Chain-of-Thought
• Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence
• RLVE: Scaling Up Reinforcement Learning for Language Models with Adaptive Verifiable Environments
• EASE: Practical and Efficient Safety Alignment for Small Language Models
• Zooming into Comics: Region-Aware RL Improves Fine-Grained Comic Understanding in Vision-Language Models
• CG-TTRL: Context-Guided Test-Time Reinforcement Learning for On-Device Large Language Models
• DRIVE: Data Curation Best Practices for Reinforcement Learning with Verifiable Reward in Competitive Code Generation
• MONICA: Real-Time Monitoring and Calibration of Chain-of-Thought Sycophancy in Large Reasoning Models
• LPFQA: A Long-Tail Professional Forum-based Benchmark for LLM Evaluation
• PRAGMA: A Profiling-Reasoned Multi-Agent Framework for Automatic Kernel Optimization
• Overview of CHIP 2025 Shared Task 2: Discharge Medication Recommendation for Metabolic Diseases Based on Chinese Electronic Health Records
• Real-World Adverse Weather Image Restoration via Dual-Level Reinforcement Learning with High-Quality Cold Start
• OpenVLN: Open-world Aerial Vision-Language Navigation
• Tiny Model, Big Logic: Diversity-Driven Optimization Elicits Large-Model Reasoning Ability in VibeThinker-1.5B
• FLEX: Continuous Agent Evolution via Forward Learning from Experience
• Klear-AgentForge: Forging Agentic Intelligence through Posttraining Scaling
• CoT-X: An Adaptive Framework for Cross-Model Chain-of-Thought Transfer and Optimization
• PreResQ-R1: Towards Fine-Grained Rank-and-Score Reinforcement Learning for Visual Quality Assessment via Preference-Response Disentangled Policy Optimization
• If I Could Turn Back Time: Temporal Reframing as a Historical Reasoning Task for LLMs
• Generating Software Architecture Description from Source Code using Reverse Engineering and Large Language Model
• iFlyBot-VLM Technical Report
• Motif 2 12.7B technical report
• Visual Spatial Tuning
• Explore Data Left Behind in Reinforcement Learning for Reasoning Language Models
• Don't Just Search, Understand: Semantic Path Planning Agent for Spherical Tensegrity Robots in Unknown Environments
• Thinking with Video: Video Generation as a Promising Multimodal Reasoning Paradigm
• Forget BIT, It is All about TOKEN: Towards Semantic Information Theory for LLMs
• Revealing AI Reasoning Increases Trust but Crowds Out Unique Human Knowledge
• LiveTradeBench: Seeking Real-World Alpha with Large Language Models
• Why Less is More (Sometimes): A Theory of Data Curation
• AnchorTP: Resilient LLM Inference with State-Preserving Elastic Tensor Parallelism
• CoPRIS: Efficient and Stable Reinforcement Learning via Concurrency-Controlled Partial Rollout with Importance Sampling
• Epidemiology of Large Language Models: A Benchmark for Observational Distribution Knowledge
• ROBoto2: An Interactive System and Dataset for LLM-assisted Clinical Trial Risk of Bias Assessment
• In Good GRACEs: Principled Teacher Selection for Knowledge Distillation
• Oolong: Evaluating Long Context Reasoning and Aggregation Capabilities
• Thinking with DistilQwen: A Tale of Four Distilled Reasoning and Reward Model Series
• POLIS-Bench: Towards Multi-Dimensional Evaluation of LLMs for Bilingual Policy Tasks in Governmental Scenarios
• The Sequential Edge: Inverse-Entropy Voting Beats Parallel Self-Consistency at Matched Compute
• Unlocking the Power of Multi-Agent LLM for Reasoning: From Lazy Agents to Deliberation
• VFocus: Better Verilog Generation from Large Language Model via Focused Reasoning
• FATE: A Formal Benchmark Series for Frontier Algebra of Multiple Difficulty Levels
• InsurAgent: A Large Language Model-Empowered Agent for Simulating Individual Behavior in Purchasing Flood Insurance
• KV Cache Transform Coding for Compact Storage in LLM Inference
• How Far Are Surgeons from Surgical World Models? A Pilot Study on Zero-shot Surgical Video Generation with Expert Assessment
• HarnessLLM: Automatic Testing Harness Generation via Reinforcement Learning
• Prompt-R1: Collaborative Automatic Prompting Framework via End-to-end Reinforcement Learning
• Fleming-VL: Towards Universal Medical Visual Reasoning with Multimodal LLMs
• MULTI-Bench: A Multi-Turn Interactive Benchmark for Assessing Emotional Intelligence ability of Spoken Dialogue Models
• OmniBrainBench: A Comprehensive Multimodal Benchmark for Brain Imaging Analysis Across Multi-stage Clinical Tasks
• Efficient Reinforcement Learning for Large Language Models with Intrinsic Exploration
• Logic-informed reinforcement learning for cross-domain optimization of large-scale cyber-physical systems
• DTS: Enhancing Large Reasoning Models via Decoding Tree Sketching
• Diagnosing Hallucination Risk in AI Surgical Decision-Support: A Sequential Framework for Sequential Validation
• ID-Crafter: VLM-Grounded Online RL for Compositional Multi-Subject Video Generation
• Saliency-R1: Incentivizing Unified Saliency Reasoning Capability in MLLM with Confidence-Guided Reinforcement Learning
• VinciCoder: Unifying Multimodal Code Generation via Coarse-to-fine Visual Reinforcement Learning
• ORGEval: Graph-Theoretic Evaluation of LLMs in Optimization Modeling
• Rethinking Facial Expression Recognition in the Era of Multimodal Large Language Models: Benchmark, Datasets, and Beyond
• MedRECT: A Medical Reasoning Benchmark for Error Correction in Clinical Texts
• Why Federated Optimization Fails to Achieve Perfect Fitting? A Theoretical Perspective on Client-Side Optima
• A Survey on Generative Recommendation: Data, Model, and Tasks
• Towards Understanding Self-play for LLM Reasoning
• Addressing Longstanding Challenges in Cognitive Science with Language Models
• Spatial-SSRL: Enhancing Spatial Understanding via Self-Supervised Reinforcement Learning
• EBT-Policy: Energy Unlocks Emergent Physical Reasoning Capabilities
• Chain of Time: In-Context Physical Simulation with Image Generation Models
• LLM-based Multi-class Attack Analysis and Mitigation Framework in IoT/IIoT Networks
• Are Video Models Ready as Zero-Shot Reasoners? An Empirical Study with the MME-CoF Benchmark
• Defeating the Training-Inference Mismatch via FP16
• AMO-Bench: Large Language Models Still Struggle in High School Math Competitions
• The End of Manual Decoding: Towards Truly End-to-End Language Models
• The Era of Agentic Organization: Learning to Organize with Language Models
• Inverse Knowledge Search over Verifiable Reasoning: Synthesizing a Scientific Encyclopedia from a Long Chains-of-Thought Knowledge Base
• Emu3.5: Native Multimodal Models are World Learners
• InfoFlow: Reinforcing Search Agent Via Reward Density Optimization
• RoboOS-NeXT: A Unified Memory-based Framework for Lifelong, Scalable, and Robust Multi-Robot Collaboration
• Rethinking Text-to-SQL: Dynamic Multi-turn SQL Interaction for Real-world Database Exploration
• OmniEduBench: A Comprehensive Chinese Benchmark for Evaluating Large Language Models in Education
• Do LLMs Signal When They're Right? Evidence from Neuron Agreement
• Reasoning Curriculum: Bootstrapping Broad LLM Reasoning from Math
• Lean4Physics: Comprehensive Reasoning Framework for College-level Physics in Lean4
• Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail
• e1: Learning Adaptive Control of Reasoning Effort
• Detecting Anomalies in Machine Learning Infrastructure via Hardware Telemetry
• Generalizing Test-time Compute-optimal Scaling as an Optimizable Graph
• Revisiting Multilingual Data Mixtures in Language Model Pretraining
• Completion ≠ Collaboration: Scaling Collaborative Effort with Agents
• EHR-R1: A Reasoning-Enhanced Foundational Language Model for Electronic Health Record Analysis
• Zero Reinforcement Learning Towards General Domains
• NormWorlds-CF: Solver-Verified Counterfactual Normative Reasoning with Metamorphic-Relation GRPO
• Cognitive Convergence: Deep Similarities Between Large Language Models and Human Cognition
• Shared SFT Lessons Across Alignment, Model Organisms, and Toy Models
• PRISM-Physics: Causal DAG-Based Process Evaluation for Physics Reasoning
• RAPID: An Efficient Reinforcement Learning Algorithm for Small Language Models
• Integrating Machine Learning Standards in Disseminating Machine Learning Research
• Improving Cooperation in Collaborative Embodied AI
• Reward Models are Metrics in a Trench Coat
• Beyond the Final Layer: Intermediate Representations for Better Multilingual Calibration in Large Language Models
• Symbol-Equivariant Recurrent Reasoning Models
• The Scaling Properties of Implicit Deductive Reasoning in Transformers
• AlphaGRPO: Unlocking Self-Reflective Multimodal Generation in UMMs via Decompositional Verifiable Reward
• Improving Heart-Focused Medical Question Answering in LLMs via Variance-Aware Rubric Rewards with GRPO
• BAS: A Decision-Theoretic Approach to Evaluating Large Language Model Confidence
• Atom-anchored LLMs speak Chemistry: A Retrosynthesis Demonstration
• LANPO: Bootstrapping Language and Numerical Feedback for Reinforcement Learning in LLMs
• NeurIPT: Foundation Model for Neural Interfaces
• NP-Engine: Empowering Optimization Reasoning in Large Language Models with Verifiable Synthetic NP Problems
• DialectLLM: A Dialect-Aware Dialog[ue] Generation Framework Beyond Standard American English
• Teaching Models to Teach Themselves: Reasoning at the Edge of Learnability
• Visual Generation Unlocks Human-Like Reasoning through Multimodal World Models
• Quantum automated theorem proving
• Fantastic Reasoning Behaviors and Where to Find Them: Unsupervised Discovery of the Reasoning Process
• Model-Document Protocol for AI Search
• Large Language Model for Verilog Code Generation: Literature Review and the Road Ahead
• KnowCoder-A1: Incentivizing Agentic Reasoning Capability with Outcome Supervision for KBQA
• SeeingEye: Agentic Information Flow Unlocks Multimodal Reasoning In Text-only LLMs
• The Kinetics of Reasoning: How Chain-of-Thought Shapes Learning in Transformers?
• SemCoT: Accelerating Chain-of-Thought Reasoning through Semantically-Aligned Implicit Tokens
• SPICE: Self-Play In Corpus Environments Improves Reasoning
• OSWorld-MCP: Benchmarking MCP Tool Invocation In Computer-Use Agents
• TritonRL: Training LLMs to Think and Code Triton Without Cheating
• Count Counts: Motivating Exploration in LLM Reasoning with Count-based Intrinsic Rewards
• MiniOneRec: An Open-Source Framework for Scaling Generative Recommendation
• Uncovering Gaps Between RFC Updates and TCP/IP Implementations: LLM-Facilitated Differential Checks on Intermediate Representations
• Synergizing chemical and AI communities for advancing laboratories of the future
• Critique-RL: Training Language Models for Critiquing through Two-Stage Reinforcement Learning
• ViPER: Empowering the Self-Evolution of Visual Perception Abilities in Vision-Language Model
• UtilGen: Utility-Centric Generative Data Augmentation with Dual-Level Task Adaptation
• BLM₁: A Boundless Large Model for Cross-Space, Cross-Task, and Cross-Embodiment Learning
• ProofSketch: Efficient Verified Reasoning for Large Language Models
• PFEA: An LLM-based High-Level Natural Language Planning and Feedback Embodied Agent for Human-Centered AI
• Reasoning Visual Language Model for Chest X-Ray Analysis
• A Pragmatic Way to Measure Chain-of-Thought Monitorability
• World Simulation with Video Foundation Models for Physical AI
• Latent Chain-of-Thought for Visual Reasoning
• Assessing the Relational Abilities of Large Language Models and Large Reasoning Models
• ATA: A Neuro-Symbolic Approach to Implement Autonomous and Trustworthy Agents
• Multi-Agent Evolve: LLM Self-Improve through Co-evolution
• EgoThinker: Unveiling Egocentric Reasoning with Spatio-Temporal CoT
• On the Faithfulness of Visual Thinking: Measurement and Enhancement
• Video-Thinker: Sparking "Thinking with Videos" via Reinforcement Learning
• Omni-Reward: Towards Generalist Omni-Modal Reward Modeling with Free-Form Preferences
• A Neuro-Symbolic Multi-Agent Approach to Legal-Cybersecurity Knowledge Integration
• EMTSF:Extraordinary Mixture of SOTA Models for Time Series Forecasting
• The Best of N Worlds: Aligning Reinforcement Learning with Best-of-N Sampling via max@k Optimisation
• LightKGG: Simple and Efficient Knowledge Graph Generation from Textual Data
• Smaller Models, Smarter Rewards: A Two-Sided Approach to Process and Outcome Rewards
• Think before Recommendation: Autonomous Reasoning-enhanced Recommender
• Incentivizing Agentic Reasoning in LLM Judges via Tool-Integrated Reinforcement Learning
• A high-capacity linguistic steganography based on entropy-driven rank-token mapping
• Towards Stable and Effective Reinforcement Learning for Mixture-of-Experts
• Improving Human Verification of LLM Reasoning through Interactive Explanation Interfaces
• Can Language Models Compose Skills In-Context?
• CodeAD: Synthesize Code of Rules for Log-based Anomaly Detection with LLMs
• Once Upon an Input: Reasoning via Per-Instance Program Synthesis
• Scalable Supervising Software Agents with Patch Reasoner
• FlowCritic: Bridging Value Estimation with Flow Matching in Reinforcement Learning
• RoboSVG: A Unified Framework for Interactive SVG Generation with Multi-modal Guidance
• Do Stop Me Now: Detecting Boilerplate Responses with a Single Iteration
• Chitchat with AI: Understand the supply chain carbon disclosure of companies worldwide through Large Language Model
• Label Smoothing Improves Gradient Ascent in LLM Unlearning
• GRPO-Guard: Mitigating Implicit Over-Optimization in Flow Matching via Regulated Clipping
• CityRiSE: Reasoning Urban Socio-Economic Status in Vision-Language Models via Reinforcement Learning
• PACR: Progressively Ascending Confidence Reward for LLM Reasoning
• LongCat-Video Technical Report
• OmniVinci: Enhancing Architecture and Data for Omni-Modal Understanding LLM
• Beyond Reasoning Gains: Mitigating General Capabilities Forgetting in Large Reasoning Models
• Few-Shot Knowledge Distillation of LLMs With Counterfactual Explanations
• Boosting Accuracy and Efficiency of Budget Forcing in LLMs via Reinforcement Learning for Mathematical Reasoning
• From Polyester Girlfriends to Blind Mice: Creating the First Pragmatics Understanding Benchmarks for Slovene
• FineRS: Fine-grained Reasoning and Segmentation of Small Objects with Reinforcement Learning
• α-LoRA: Effective Fine-Tuning via Base Model Rescaling
• Multi-turn Training with Basic Human Feedback Helps Little on LLM Reasoning
• Social Simulations with Large Language Model Risk Utopian Illusion
• String Seed of Thought: Prompting LLMs for Distribution-Faithful and Diverse Generation
• PhysVLM-AVR: Active Visual Reasoning for Multimodal Large Language Models in Physical Environments
• Self-Rewarding PPO: Aligning Large Language Models with Demonstrations Only
• The Virtues of Brevity: Avoid Overthinking in Parallel Test-Time Reasoning
• The Gray Zone of Faithfulness: Taming Ambiguity in Unfaithfulness Detection
• Customizing Open Source LLMs for Quantitative Medication Attribute Extraction across Heterogeneous EHR Systems
• L²M³OF: A Large Language Multimodal Model for Metal-Organic Frameworks
• Co-Designing Quantum Codes with Transversal Diagonal Gates via Multi-Agent Systems
• Trust, But Verify: An Empirical Evaluation of AI-Generated Code for SDN Controllers
• Generalizable Reasoning through Compositional Energy Minimization
• GranViT: A Fine-Grained Vision Model With Autoregressive Perception For MLLMs
• Conan: Progressive Learning to Reason Like a Detective over Multi-Scale Visual Evidence
• Addressing Corner Cases in Autonomous Driving: A World Model-based Approach with Mixture of Experts and LLMs
• Ask a Strong LLM Judge when Your Reward Model is Uncertain
• Every Question Has Its Own Value: Reinforcement Learning with Explicit Human Values
• DeepWideSearch: Benchmarking Depth and Width in Agentic Information Seeking
• AsyncHZP: Hierarchical ZeRO Parallelism with Asynchronous Scheduling for Scalable LLM Training
• Learning to Triage Taint Flows Reported by Dynamic Program Analysis in Node.js Packages
• LM-mixup: Text Data Augmentation via Language Model based Mixup
• KL-Regularized Reinforcement Learning is Designed to Mode Collapse
• ResearchGPT: Benchmarking and Training LLMs for End-to-End Computer Science Research Workflows
• Code-enabled language models can outperform reasoning models on diverse tasks
• UI-Ins: Enhancing GUI Grounding with Multi-Perspective Instruction-as-Reasoning
• Metis-HOME: Hybrid Optimized Mixture-of-Experts for Multimodal Reasoning
• On the Detectability of LLM-Generated Text: What Exactly Is LLM-Generated Text?
• A Concrete Roadmap towards Safety Cases based on Chain-of-Thought Monitoring
• Can Confidence Estimates Decide When Chain-of-Thought Is Necessary for LLMs?
• Enhancing Reasoning Skills in Small Persian Medical Language Models Can Outperform Large-Scale Data Training
• Seed3D 1.0: From Images to High-Fidelity Simulation-Ready 3D Assets
• BugPilot: Complex Bug Generation for Efficient Learning of SWE Skills
• SmartSwitch: Advancing LLM Reasoning by Overcoming Underthinking via Promoting Deeper Thought Exploration
• Stream: Scaling up Mechanistic Interpretability to Long Context in LLMs via Sparse Attention
• Learning Noise-Resilient and Transferable Graph-Text Alignment via Dynamic Quality Assessment
• The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Models
• LoongRL: Reinforcement Learning for Advanced Reasoning over Long Contexts
• No Compute Left Behind: Rethinking Reasoning and Sampling with Masked Diffusion Models
• Imitation Learning Policy based on Multi-Step Consistent Integration Shortcut Model
• RailS: Load Balancing for All-to-All Communication in Distributed Mixture-of-Experts Training
• Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning
• Can Reasoning Models Obfuscate Reasoning? Stress-Testing Chain-of-Thought Monitorability
• Every Step Evolves: Scaling Reinforcement Learning for Trillion-Scale Thinking Model
• A Model Can Help Itself: Reward-Free Self-Training for LLM Reasoning
• Reasoning Language Model Inference Serving Unveiled: An Empirical Study
• Sherlock Your Queries: Learning to Ask the Right Questions for Dialogue-Based Retrieval
• Extracting alignment data in open models
• VAPU: System for Autonomous Legacy Code Modernization
• Prompting the Priorities: A First Look at Evaluating LLMs for Vulnerability Triage and Prioritization
• CodeRL+: Improving Code Generation via Reinforcement with Execution Semantics Alignment
• Automated urban waterlogging assessment and early warning through a mixture of foundation models
• Proactive Reasoning-with-Retrieval Framework for Medical Multimodal Large Language Models
• EdgeReasoning: Characterizing Reasoning LLM Deployment on Edge GPUs
• StreamingTOM: Streaming Token Compression for Efficient Video Understanding
• Learning from the Best, Differently: A Diversity-Driven Rethinking on Data Selection
• DelvePO: Direction-Guided Self-Evolving Framework for Flexible Prompt Optimization
• Test-time Verification via Optimal Transport: Coverage, ROC, & Sub-optimality
• Plural Voices, Single Agent: Towards Inclusive AI in Multi-User Domestic Spaces
• Local Coherence or Global Validity? Investigating RLVR Traces in Math Domains
• BlueCodeAgent: A Blue Teaming Agent Enabled by Automated Red Teaming for CodeGen AI
• Any-Depth Alignment: Unlocking Innate Safety Alignment of LLMs to Any-Depth
• Unbiased Gradient Low-Rank Projection
• Foundational Automatic Evaluators: Scaling Multi-Task Generative Evaluator Training for Reasoning-Centric Domains
• MARS-M: When Variance Reduction Meets Matrices
• Enabling Fine-Grained Operating Points for Black-Box LLMs
• LLM-as-a-Prophet: Understanding Predictive Intelligence with Prophet Arena
• Deep Self-Evolving Reasoning
• DETree: DEtecting Human-AI Collaborative Texts via Tree-Structured Hierarchical Representation Learning
• ReXMoE: Reusing Experts with Minimal Overhead in Mixture-of-Experts
• Disparities in Multilingual LLM-Based Healthcare Q&A
• EvoSyn: Generalizable Evolutionary Data Synthesis for Verifiable Learning
• DynaKV: Enabling Accurate and Efficient Long-Sequence LLM Decoding on Smartphones
• Explainability of Large Language Models: Opportunities and Challenges toward Generating Trustworthy Explanations
• Mamba4Net: Distilled Hybrid Mamba Large Language Models For Networking
• SimBench: Benchmarking the Ability of Large Language Models to Simulate Human Behaviors
• Learning to Design Soft Hands using Reward Models
• Fine-tuning Flow Matching Generative Models with Intermediate Feedback
• Soft-Masked Diffusion Language Models
• Rewarding the Journey, Not Just the Destination: A Composite Path and Answer Self-Scoring Reward Mechanism for Test-Time Reinforcement Learning
• Language Confusion Gate: Language-Aware Decoding Through Model Self-Distillation
• DeTAILS: Deep Thematic Analysis with Iterative LLM Support
• TREAT: A Code LLMs Trustworthiness / Reliability Evaluation and Testing Framework
• I-RAVEN-X: Benchmarking Generalization and Robustness of Analogical and Mathematical Reasoning in Large Language and Reasoning Models
• Infinity Parser: Layout Aware Reinforcement Learning for Scanned Document Parsing
• Bits Leaked per Query: Information-Theoretic Bounds on Adversarial Attacks against LLMs
• Leave It to the Experts: Detecting Knowledge Distillation via MoE Expert Signatures
• DeepAnalyze: Agentic Large Language Models for Autonomous Data Science
• QuanBench: Benchmarking Quantum Code Generation with Large Language Models
• Planner and Executor: Collaboration between Discrete Diffusion And Autoregressive Models in Reasoning
• Soundness-Aware Level: A Microscopic Signature that Predicts LLM Reasoning Potential
• InfiMed-ORBIT: Aligning LLMs on Open-Ended Complex Tasks via Rubric-Based Incremental Training
• DeceptionBench: A Comprehensive Benchmark for AI Deception Behaviors in Real-world Scenarios
• Reliability of Large Language Model Generated Clinical Reasoning in Assisted Reproductive Technology: Blinded Comparative Evaluation Study
• Distractor Injection Attacks on Large Reasoning Models: Characterization and Defense
• ImagerySearch: Adaptive Test-Time Search for Video Generation Beyond Semantic Dependency Constraints
• Select Less, Reason More: Prioritizing Evidence Purity for Video Reasoning
• CarBoN: Calibrated Best-of-N Sampling Improves Test-time Reasoning
• DLER: Doing Length pEnalty Right - Incentivizing More Intelligence per Token via Reinforcement Learning
• Internalizing World Models via Self-Play Finetuning for Agentic RL
• Information Gain-based Policy Optimization: A Simple and Effective Approach for Multi-Turn LLM Agents
• LaSeR: Reinforcement Learning with Last-Token Self-Rewarding
• GroundedPRM: Tree-Guided and Fidelity-Aware Process Reward Modeling for Step-Level Reasoning
• Reasoning with Sampling: Your Base Model is Smarter Than You Think
• To Infinity and Beyond: Tool-Use Unlocks Length Generalization in State Space Models
• AMS-QUANT: Adaptive Mantissa Sharing for Floating-point Quantization
• Finding Answers in Thought Matters: Revisiting Evaluation on Large Language Models with Reasoning
• Code-driven Number Sequence Calculation: Enhancing the inductive Reasoning Abilities of Large Language Models
• Beyond Correctness: Evaluating Subjective Writing Preferences Across Cultures
• Oracle-RLAIF: An Improved Fine-Tuning Framework for Multi-modal Video Models using Reinforcement Learning from Ranking Feedback
• LLM Agents Beyond Utility: An Open-Ended Perspective
• Coder as Editor: Code-driven Interpretable Molecular Optimization
• IMAGINE: Integrating Multi-Agent System into One Model for Complex Reasoning and Planning
• Can MLLMs Absorb Math Reasoning Abilities from LLMs as Free Lunch?
• CAST: Compositional Analysis via Spectral Tracking for Understanding Transformer Layer Functions
• Towards Agentic Self-Learning LLMs in Search Environment
• BitNet Distillation
• Breadcrumbs Reasoning: Memory-Efficient Reasoning with Compression Beacons
• InteractiveOmni: A Unified Omni-modal Model for Audio-Visual Multi-turn Dialogue
• Hard2Verify: A Step-Level Verification Benchmark for Open-Ended Frontier Math
• Confidence as a Reward: Transforming LLMs into Reward Models
• LiteraryQA: Towards Effective Evaluation of Long-document Narrative QA
• Beyond Correctness: Rewarding Faithful Reasoning in Retrieval-Augmented Generation
• Higher Satisfaction, Lower Cost: A Technical Report on How LLMs Revolutionize Meituan's Intelligent Interaction Systems
• EgoSocial: Benchmarking Proactive Intervention Ability of Omnimodal LLMs via Egocentric Social Interaction Perception
• Optimal Aggregation of LLM and PRM Signals for Efficient Test-Time Scaling
• CoT-Evo: Evolutionary Distillation of Chain-of-Thought for Scientific Reasoning
• Adaptive Reasoning Executor: A Collaborative Agent System for Efficient Reasoning
• ConsintBench: Evaluating Language Models on Real-World Consumer Intent Understanding
• A Matter of Representation: Towards Graph-Based Abstract Code Generation
• Reinforcement Learning Meets Masked Generative Models: Mask-GRPO for Text-to-Image Generation
• The Art of Scaling Reinforcement Learning Compute for LLMs
• Max It or Miss It: Benchmarking LLM On Solving Extremal Problems
• RAG Meets Temporal Graphs: Time-Sensitive Modeling and Retrieval for Evolving Knowledge
• FreshTab: Sourcing Fresh Data for Table-to-Text Generation Evaluation
• Toward Reasoning-Centric Time-Series Analysis
• Litespark Technical Report: High-Throughput, Energy-Efficient LLM Training Framework
• Demystifying Hybrid Thinking: Can LLMs Truly Switch Between Think and No-Think?
• CoIRL-AD: Collaborative-Competitive Imitation-Reinforcement Learning in Latent World Models for Autonomous Driving
• LLM-REVal: Can We Trust LLM Reviewers Yet?
• Finite-time Convergence Analysis of Actor-Critic with Evolving Reward
• Reinforced Preference Optimization for Recommendation
• Self-Verifying Reflection Helps Transformers with CoT Reasoning
• MetaCaptioner: Towards Generalist Visual Captioning with Open-source Suites
• Improving Text-to-Image Generation with Input-Side Inference-Time Scaling
• EduDial: Constructing a Large-scale Multi-turn Teacher-Student Dialogue Corpus
• CGBench: Benchmarking Language Model Scientific Reasoning for Clinical Genetics Research
• Deep Research Brings Deeper Harm
• Demystifying Reinforcement Learning in Agentic Reasoning
• QeRL: Beyond Efficiency -- Quantization-enhanced Reinforcement Learning for LLMs
• ACADREASON: Exploring the Limits of Reasoning Models with Academic Research Problems
• EvoCAD: Evolutionary CAD Code Generation with Vision Language Models
• A²FM: An Adaptive Agent Foundation Model for Tool-Aware Hybrid Reasoning
• Enhancing Long Chain-of-Thought Reasoning through Multi-Path Plan Aggregation
• ExpVid: A Benchmark for Experiment Video Understanding & Reasoning
• Tree-based Dialogue Reinforced Policy Optimization for Red-Teaming Attacks
• Bag of Tricks for Subverting Reasoning-based Safety Guardrails
• What Generative Search Engines Like and How to Optimize Web Content Cooperatively
• LouisKV: Efficient KV Cache Retrieval for Long Input-Output Sequences
• Protein as a Second Language for LLMs
• CoPRS: Learning Positional Prior from Chain-of-Thought for Reasoning Segmentation
• Enhancing LLM Reasoning via Non-Human-Like Reasoning Path Preference Optimization
• LogiNumSynth: Synthesizing Joint Logical-Numerical Reasoning Problems for Language Models
• Automating Structural Engineering Workflows with Large Language Model Agents
• A Survey on Agentic Multimodal Large Language Models
• Parallel Scaling Law: Unveiling Reasoning Generalization through A Cross-Linguistic Perspective
• Video-STR: Reinforcing MLLMs in Video Spatio-Temporal Reasoning with Relation Graph
• Benchmarking Correctness and Security in Multi-Turn Code Generation
• Where on Earth? A Vision-Language Benchmark for Probing Model Geolocation Skills Across Scales
• More Than One Teacher: Adaptive Multi-Guidance Policy Optimization for Diverse Exploration
• Cog-Rethinker: Hierarchical Metacognitive Reinforcement Learning for LLM Reasoning
• CodePlot-CoT: Mathematical Visual Reasoning by Thinking with Code-Driven Images
• RoVer: Robot Reward Model as Test-Time Verifier for Vision-Language-Action Model
• Direct Multi-Token Decoding
• Stabilizing MoE Reinforcement Learning by Aligning Training and Inference Routers
• UpSafe^∘C: Upcycling for Controllable Safety in Large Language Models
• DCP: Addressing Input Dynamism In Long-Context Training via Dynamic Context Parallelism
• OmniQuality-R: Advancing Reward Models Through All-Encompassing Quality Assessment
• ViSurf: Visual Supervised-and-Reinforcement Fine-Tuning for Large Vision-and-Language Models
• Tracing the Traces: Latent Temporal Signals for Efficient and Accurate Reasoning
• SASER: Stego attacks on open-source LLMs
• Trace Length is a Simple Uncertainty Signal in Reasoning Models
• OpusAnimation: Code-Based Dynamic Chart Generation
• Robust Exploratory Stopping under Ambiguity in Reinforcement Learning
• Reasoning-Enhanced Large Language Models for Molecular Property Prediction
• You only need 4 extra tokens: Synergistic Test-time Adaptation for LLMs
• IntrinTrans: LLM-based Intrinsic Code Translator for RISC-V Vector
• Answer-Consistent Chain-of-thought Reinforcement Learning For Multi-modal Large Langauge Models
• Output Supervision Can Obfuscate the Chain of Thought
• Automated Glaucoma Report Generation via Dual-Attention Semantic Parallel-LSTM and Multimodal Clinical Data Integration
• Efficient Onboard Vision-Language Inference in UAV-Enabled Low-Altitude Economy Networks via LLM-Enhanced Optimization
• Skill-Targeted Adaptive Training
• Beyond the limitation of a single query: Train your LLM for query expansion with Reinforcement Learning
• Adaptive Dual Reasoner: Large Reasoning Models Can Think Efficiently by Hybrid Reasoning
• Enhancing Large Language Model Reasoning with Reward Models: An Analytical Survey
• Quagmires in SFT-RL Post-Training: When High SFT Scores Mislead and What to Use Instead
• Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers
• Sparse Query Attention (SQA): A Computationally Efficient Attention Mechanism with Query Heads Reduction
• Spreadsheet-RL: Advancing Large Language Model Agents on Realistic Spreadsheet Tasks via Reinforcement Learning
• A Systematic Study on Generating Web Vulnerability Proof-of-Concepts Using Large Language Models
• Every Step Counts: Decoding Trajectories as Authorship Fingerprints of dLLMs
• Don't Throw Away Your Pretrained Model
• Autonomous Agents for Scientific Discovery: Orchestrating Scientists, Language, Code, and Physics
• AutoPR: Let's Automate Your Academic Promotion!
• Beyond Surface Reasoning: Unveiling the True Long Chain-of-Thought Capacity of Diffusion Large Language Models
• SPG: Sandwiched Policy Gradient for Masked Diffusion Language Models
• Agentic Systems in Radiology: Design, Applications, Evaluation, and Challenges
• Diagnosing Shoulder Disorders Using Multimodal Large Language Models and Consumer-Grade Cameras
• When Retrieval Succeeds and Fails: Rethinking Retrieval-Augmented Generation for LLMs
• Cross-Representation Benchmarking in Time-Series Electronic Health Records for Clinical Outcome Prediction
• ReFIne: A Framework for Trustworthy Large Reasoning Models with Reliability, Faithfulness, and Interpretability
• Robust Driving Control for Autonomous Vehicles: An Intelligent General-sum Constrained Adversarial Reinforcement Learning Approach
• All Code, No Thought: Current Language Models Struggle to Reason in Ciphered Language
• Value-State Gated Attention for Mitigating Extreme-Token Phenomena in Transformers
• DiTSinger: Scaling Singing Voice Synthesis with Diffusion Transformer and Implicit Alignment
• Towards Efficient Multimodal Unified Reasoning Model via Model Merging
• RegexPSPACE: A Benchmark for Evaluating LLM Reasoning on PSPACE-complete Regex Problems
• Unleashing Perception-Time Scaling to Multimodal Reasoning Models
• Just Do It!? Computer-Use Agents Exhibit Blind Goal-Directedness
• SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAG
• BaNEL: Exploration Posteriors for Generative Modeling Using Only Negative Rewards
• BLINK-Twice: You see, but do you observe? A Reasoning Benchmark on Visual Perception
• How Many Code and Test Cases Are Enough? Evaluating Test Cases Generation from a Binary-Matrix Perspective
• SpatialLadder: Progressive Training for Spatial Reasoning in Vision-Language Models
• Beyond Turn Limits: Training Deep Search Agents with Dynamic Context Window
• Fewer Weights, More Problems: A Practical Attack on LLM Pruning
• Active Confusion Expression in Large Language Models: Leveraging World Models toward Better Social Reasoning
• LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?
• GCPO: When Contrast Fails, Go Gold
• OpenRubrics: Towards Scalable Synthetic Rubric Generation for Reward Modeling and LLM Alignment
• BLAZER: Bootstrapping LLM-based Manipulation Agents with Zero-Shot Data Generation
• Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens
• R-Horizon: How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
• Detecting Post-generation Edits to Watermarked LLM Outputs via Combinatorial Watermarking
• Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization
• Beyond Pass@k: Breadth-Depth Metrics for Reasoning Boundaries
• MoA-VR: A Mixture-of-Agents System Towards All-in-One Video Restoration
• Self-Improving LLM Agents at Test-Time
• Reinforcing Diffusion Models by Direct Group Preference Optimization
• An Adaptive Multi Agent Bitcoin Trading System
• A²Search: Ambiguity-Aware Question Answering with Reinforcement Learning
• FlyLoRA: Boosting Task Decoupling and Parameter Efficiency via Implicit Rank-Wise Mixture-of-Experts
• PEAR: Phase Entropy Aware Reward for Efficient Reasoning
• AutoMLGen: Navigating Fine-Grained Optimization for Coding Agents
• Don't Waste Mistakes: Leveraging Negative RL-Groups via Confidence Reweighting
• Beyond Textual CoT: Interleaved Text-Image Chains with Deep Confidence Reasoning for Image Editing
• Entropy Regularizing Activation: Boosting Continuous Control, Large Language Models, and Image Classification with Activation as Entropy Constraints
• SliceFine: The Universal Winning-Slice Hypothesis for Pretrained Networks
• Populism Meets AI: Advancing Populism Research with LLMs
• Haystack Engineering: Context Engineering for Heterogeneous and Agentic Long-Context Evaluation
• Artificial Hippocampus Networks for Efficient Long-Context Modeling
• Hybrid Reinforcement: When Reward Is Sparse, It's Better to Be Dense
• Red-Bandit: Test-Time Adaptation for LLM Red-Teaming via Bandit-Guided LoRA Experts
• Leveraging LLMs to Streamline the Review of Public Funding Applications
• Encode, Think, Decode: Scaling test-time reasoning with recursive latent thoughts
• Accelerating Diffusion LLM Inference via Local Determinism Propagation
• SaFeR-VLM: Toward Safety-aware Fine-grained Reasoning in Multimodal Models
• Autoformalizer with Tool Feedback
• CLUE: Non-parametric Verification from Experience via Hidden-State Clustering
• ImageNet-Think-250K: A Large-Scale Synthetic Dataset for Multimodal Reasoning for Vision Language Models
• Think Right: Learning to Mitigate Under-Over Thinking via Adaptive, Attentive Compression
• CardioRAG: A Retrieval-Augmented Generation Framework for Multimodal Chagas Disease Detection
• Towards Interpretable and Inference-Optimal COT Reasoning with Sparse Autoencoder-Guided Generation
• FURINA: A Fully Customizable Role-Playing Benchmark via Scalable Multi-Agent Collaboration Pipeline
• Verifying Memoryless Sequential Decision-making of Large Language Models
• Scaling LLM Multi-turn RL with End-to-end Summarization-based Context Management
• Rethinking KL Regularization in RLHF: From Value Estimation to Gradient Optimization
• The Markovian Thinker: Architecture-Agnostic Linear Scaling of Reasoning
• VLA-R1: Enhancing Reasoning in Vision-Language-Action Models
• ConCuR: Conciseness Makes State-of-the-Art Kernel Generation
• Adaptive Tool Generation with Models as Tools and Reinforcement Learning
• Get RICH or Die Scaling: Profitably Trading Inference Compute for Robustness
• AsyncSpade: Efficient Test-Time Scaling with Asynchronous Sparse Decoding
• λ-GRPO: Unifying the GRPO Frameworks with Learnable Token Preferences
• The Framework That Survives Bad Models: Human-AI Collaboration For Clinical Trials
• TGPR: Tree-Guided Policy Refinement for Robust Self-Debugging of LLMs
• Webscale-RL: Automated Data Pipeline for Scaling RL Data to Pretraining Levels
• TaTToo: Tool-Grounded Thinking PRM for Test-Time Scaling in Tabular Reasoning
• Lumina-DiMOO: An Omni Diffusion Large Language Model for Multi-Modal Generation and Understanding
• Belief-Calibrated Multi-Agent Consensus Seeking for Complex NLP Tasks
• SDAR: A Synergistic Diffusion-AutoRegression Paradigm for Scalable Sequence Generation
• lm-Meter: Unveiling Runtime Inference Latency for On-Device Language Models
• Influence Functions for Efficient Data Selection in Reasoning
• Optimizing for Persuasion Improves LLM Generalization: Evidence from Quality-Diversity Evolution of Debate Strategies
• Towards Label-Free Biological Reasoning Synthetic Dataset Creation via Uncertainty Filtering
• EMORL-TTS: Reinforcement Learning for Fine-Grained Emotion Control in LLM-based TTS
• HOI-R1: Exploring the Potential of Multimodal Large Language Models for Human-Object Interaction Detection
• Presenting a Paper is an Art: Self-Improvement Aesthetic Agents for Academic Presentations
• AQA-TTRL: Self-Adaptation in Audio Question Answering with Test-Time Reinforcement Learning
• When Importance Sampling Misallocates Credit: Asymmetric Ratios for Outcome-Supervised RL
• Refusal Falls off a Cliff: How Safety Alignment Fails in Reasoning?
• Data Provenance Auditing of Fine-Tuned Large Language Models with a Text-Preserving Technique
• The Valley of Code Reasoning: Scaling Knowledge Distillation of Large Language Models
• Bloom: Designing for LLM-Augmented Behavior Change Interactions
• Adversarial Reinforcement Learning for Large Language Model Agent Safety
• Let it Calm: Exploratory Annealed Decoding for Verifiable Reinforcement Learning
• Stratum: System-Hardware Co-Design with Tiered Monolithic 3D-Stackable DRAM for Efficient MoE Serving
• Video-LMM Post-Training: A Deep Dive into Video Reasoning with Large Multimodal Models
• Large Language Models Achieve Gold Medal Performance at the International Olympiad on Astronomy & Astrophysics (IOAA)
• Aligning Perception, Reasoning, Modeling and Interaction: A Survey on Physical AI
• Reinforcement Learning for Clinical Reasoning: Aligning LLMs with ACR Imaging Appropriateness Criteria
• Bridging Reasoning to Learning: Unmasking Illusions using Complexity Out of Distribution Generalization
• Online automatic code generation for robot swarms: LLMs and self-organizing hierarchy
• The Reasoning Boundary Paradox: How Reinforcement Learning Constrains Language Models
• EduPersona: Benchmarking Subjective Ability Boundaries of Virtual Student Agents
• DRPO: Efficient Reasoning via Decoupled Reward Policy Optimization
• Mitigating Forgetting Between Supervised and Reinforcement Learning Yields Stronger Reasoners
• Making Mathematical Reasoning Adaptive
• Learning on the Job: Test-Time Curricula for Targeted Reinforcement Learning
• TRAJECT-Bench:A Trajectory-Aware Benchmark for Evaluating Agentic Tool Use
• COLE: a Comprehensive Benchmark for French Language Understanding Evaluation
• A Lightweight Large Language Model-Based Multi-Agent System for 2D Frame Structural Analysis
• Video-in-the-Loop: Span-Grounded Long Video QA with Interleaved Reasoning
• Read the Scene, Not the Script: Outcome-Aware Safety for LLMs
• SPOGW: a Score-based Preference Optimization method via Group-Wise comparison for workflows
• Toward a unified framework for data-efficient evaluation of large language models
• RLRF: Competitive Search Agent Design via Reinforcement Learning from Ranker Feedback
• The Debate on RLVR Reasoning Capability Boundary: Shrinkage, Expansion, or Both? A Two-Stage Dynamic View
• Searching Meta Reasoning Skeleton to Guide LLM Reasoning
• Measuring Language Model Hallucinations Through Distributional Correctness
• COSMO-RL: Towards Trustworthy LMRMs via Joint Safety and Stability
• From Poisoned to Aware: Fostering Backdoor Self-Awareness in LLMs
• RESTRAIN: From Spurious Votes to Signals -- Self-Driven RL with Self-Penalization
• What Shapes a Creative Machine Mind? Comprehensively Benchmarking Creativity in Foundation Models
• CALM Before the STORM: Unlocking Native Reasoning for Optimization Modeling
• EvoEngineer: Mastering Automated CUDA Kernel Code Evolution with Large Language Models
• REG: A Regularization Optimizer for Robust Training Dynamics
• Unlocking Reasoning Capabilities in LLMs via Reinforcement Learning Exploration
• What Can You Do When You Have Zero Rewards During RL?
• Sample, Align, Synthesize: Graph-Based Response Synthesis with ConGrs
• NonTextual Target Attack
• Towards Scalable and Consistent 3D Editing
• RoiRL: Efficient, Self-Supervised Reasoning with Offline Iterative Reinforcement Learning
• Reward Model Routing in Alignment
• Attack via Overfitting: 10-shot Benign Fine-tuning to Jailbreak LLMs
• The Path of Self-Evolving Large Language Models: Achieving Data-Efficient Learning via Intrinsic Feedback
• Retrv-R1: A Reasoning-Driven MLLM Framework for Universal and Efficient Multimodal Retrieval
• Studying the Korean Word-Chain Game with RLVR: Mitigating Reward Conflicts via Curriculum Learning
• Brain-Language Model Alignment: Insights into the Platonic Hypothesis and Intermediate-Layer Advantage
• Social Welfare Function Leaderboard: When LLM Agents Allocate Social Welfare
• Safety Instincts: LLMs Learn to Trust Their Internal Compass for Self-Defense
• Typed Chain-of-Thought: A Curry-Howard Framework for Verifying LLM Reasoning
• QUASAR: Quantum Assembly Code Generation Using Tool-Augmented LLMs via Agentic RL
• Can Emulating Semantic Translation Help LLMs with Code Translation? A Study Based on Pseudocode
• Material Synthesis 2025 (MatSyn25) Dataset for 2D Materials
• ACPO: Adaptive Curriculum Policy Optimization for Aligning Vision-Language Models in Complex Reasoning
• Demystifying deep search: a holistic evaluation with hint-free multi-hop questions and factorised metrics
• Agent-ScanKit: Unraveling Memory and Reasoning of Multimodal Agents via Sensitivity Perturbations
• 3D-RAD: A Comprehensive 3D Radiology Med-VQA Dataset with Multi-Temporal Analysis and Diverse Diagnostic Tasks
• LongCodeZip: Compress Long Context for Code Language Models
• PAL-UI: Planning with Active Look-back for Vision-Based GUI Agents
• Generalized Parallel Scaling with Interdependent Generations
• Erase to Improve: Erasable Reinforcement Learning for Search-Augmented LLMs
• RiskPO: Risk-based Policy Optimization via Verifiable Reward for LLM Post-Training
• BroRL: Scaling Reinforcement Learning via Broadened Exploration
• Apriel-1.5-15b-Thinker
• GRPO-λ: Credit Assignment improves LLM Reasoning
• TAMA: Tool-Augmented Multimodal Agent for Procedural Activity Understanding
• Clarification as Supervision: Reinforcement Learning for Vision-Language Interfaces
• Linking Process to Outcome: Conditional Reward Modeling for LLM Reasoning
• Probing the Critical Point (CritPt) of AI Reasoning: a Frontier Physics Research Benchmark
• Training Matryoshka Mixture-of-Experts for Elastic Inference-Time Expert Utilization
• VitaBench: Benchmarking LLM Agents with Versatile Interactive Tasks in Real-world Applications
• CreAgentive: An Agent Workflow Driven Multi-Category Creative Generation Engine
• Latent Thinking Optimization: Your Latent Reasoning Language Model Secretly Encodes Reward Signals in Its Latent Thoughts
• Thinking-Free Policy Initialization Makes Distilled Reasoning Models More Effective and Efficient Reasoners
• Clip-Low Increases Entropy and Clip-High Decreases Entropy in Reinforcement Learning of Large Language Models
• RE-Searcher: Robust Agentic Search with Goal-oriented Planning and Self-reflection
• RoRecomp: Enhancing Reasoning Efficiency via Rollout Response Recomposition in Reinforcement Learning
• Lita: Light Agent Uncovers the Agentic Coding Capabilities of LLMs
• Reinforced Embodied Planning with Verifiable Reward for Real-World Robotic Manipulation
• Knapsack RL: Unlocking Exploration of LLMs via Optimizing Budget Allocation
• More Thought, Less Accuracy? On the Dual Nature of Reasoning in Vision-Language Models
• Overthinking Reduction with Decoupled Rewards and Curriculum Data Scheduling
• Logo-VGR: Visual Grounded Reasoning for Open-world Logo Recognition
• Improving Sampling Efficiency in RLVR through Adaptive Rollout and Response Reuse
• TruthRL: Incentivizing Truthful LLMs via Reinforcement Learning
• Nudging the Boundaries of LLM Reasoning
• LLM-Powered Code Analysis and Optimization for Gaussian Splatting Kernels
• FlowMoE: A Scalable Pipeline Scheduling Framework for Distributed Mixture-of-Experts Training
• Learning to Reason as Action Abstractions with Scalable Mid-Training RL
• PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection
• Planner-R1: Reward Shaping Enables Efficient Agentic RL with Smaller LLMs
• Hybrid Reward Normalization for Process-supervised Non-verifiable Agentic Tasks
• Binary Sparse Coding for Interpretability
• DeepSearch: Overcome the Bottleneck of Reinforcement Learning with Verifiable Rewards via Monte Carlo Tree Search
• From Faithfulness to Correctness: Generative Reward Models that Think Critically
• Spontaneous High-Order Generalization in Neural Theory-of-Mind Networks
• InfoAgent: Advancing Autonomous Information-Seeking Agents
• SIRI: Scaling Iterative Reinforcement Learning with Interleaved Compression
• VideoAnchor: Reinforcing Subspace-Structured Visual Cues for Coherent Visual-Spatial Reasoning
• Rethinking Entropy Regularization in Large Reasoning Models
• Cogito, Ergo Ludo: An Agent that Learns to Play by Reasoning and Planning
• CLPO: Curriculum Learning meets Policy Optimization for LLM Reasoning
• Random Policy Valuation is Enough for LLM Reasoning with Verifiable Rewards
• Retro*: Optimizing LLMs for Reasoning-Intensive Document Retrieval
• ProxyAttn: Guided Sparse Attention via Representative Heads
• LOVE-R1: Advancing Long Video Understanding with an Adaptive Zoom-in Mechanism via Multi-Step Reasoning
• UniPruning: Unifying Local Metric and Global Feedback for Scalable Sparse LLMs
• LatentEvolve: Self-Evolving Test-Time Scaling in Latent Space
• Diamonds in the rough: Transforming SPARCs of imagination into a game concept by leveraging medium sized LLMs
• MemGen: Weaving Generative Latent Memory for Self-Evolving Agents
• FedPOB: Sample-Efficient Federated Prompt Optimization via Bandits
• AdaThink-Med: Medical Adaptive Thinking with Uncertainty-Guided Length Calibration
• SemGuard: Real-Time Semantic Evaluator for Correcting LLM-Generated Code
• JSProtect: A Scalable Obfuscation Framework for Mini-Games in WeChat
• CDT: A Comprehensive Capability Framework for Large Language Models Across Cognition, Domain, and Task
• Towards Safe Reasoning in Large Reasoning Models via Corrective Intervention
• AdaNav: Adaptive Reasoning with Uncertainty for Vision-Language Navigation
• Evolution Strategies at Scale: LLM Fine-Tuning Beyond Reinforcement Learning
• MAS²: Self-Generative, Self-Configuring, Self-Rectifying Multi-Agent Systems
• FrameThinker: Learning to Think with Long Videos via Multi-Turn Frame Spotlighting
• Bridging the behavior-neural gap: A multimodal AI reveals the brain's geometry of emotion more accurately than human self-reports
• AdvChain: Adversarial Chain-of-Thought Tuning for Robust Safety Alignment of Large Reasoning Models
• Risk-Sensitive RL for Alleviating Exploration Dilemmas in Large Language Models
• ChessArena: A Chess Testbed for Evaluating Strategic Reasoning Capabilities of Large Language Models
• Learning to Ponder: Adaptive Reasoning in Latent Space
• MDD-Thinker: Towards Large Reasoning Models for Major Depressive Disorder Diagnosis
• Your thoughts tell who you are: Characterize the reasoning patterns of LRMs
• InfLLM-V2: Dense-Sparse Switchable Attention for Seamless Short-to-Long Adaptation
• SynthPert: Enhancing LLM Biological Reasoning via Synthetic Reasoning Traces for Cellular Perturbation Prediction
• Advantage Weighted Matching: Aligning RL with Pretraining in Diffusion Models
• Short window attention enables long-term memorization
• Rethinking and Benchmarking Large Language Models for Graph Reasoning
• GeoVLM-R1: Reinforcement Fine-Tuning for Improved Remote Sensing Reasoning
• Expanding Computation Spaces of LLMs at Inference Time
• PIPer: On-Device Environment Setup via Online Reinforcement Learning
• Bridging On-Device and Cloud LLMs for Collaborative Reasoning: A Unified Methodology for Local Routing and Post-Training
• Conditional Advantage Estimation for Reinforcement Learning in Large Reasoning Models
• Reinforcement Learning with Inverse Rewards for World Model Post-training
• Taming Masked Diffusion Language Models via Consistency Trajectory Reinforcement Learning with Fewer Decoding Step
• RADAR: A Risk-Aware Dynamic Multi-Agent Framework for LLM Safety Evaluation via Role-Specialized Collaboration
• GUI-Shepherd: Reliable Process Reward and Verification for Long-Sequence GUI Tasks
• SafeSearch: Automated Red-Teaming for the Safety of LLM-Based Search Agents
• Evaluating Program Semantics Reasoning with Type Inference in System F
• Beyond English-Centric Training: How Reinforcement Learning Improves Cross-Lingual Reasoning in LLMs
• ReWatch-R1: Boosting Complex Video Reasoning in Large Vision-Language Models through Agentic Data Synthesis
• Emergent Slow Thinking in LLMs as Inverse Tree Freezing
• Timber: Training-free Instruct Model Refining with Base via Effective Rank
• Towards Efficient CoT Distillation: Self-Guided Rationale Selector for Better Performance with Fewer Rationales
• Efficient Turing Machine Simulation with Transformers
• Large-Scale Constraint Generation -- Can LLMs Parse Hundreds of Constraints?
• Advancing Multi-agent Traffic Simulation via R1-Style Reinforcement Fine-Tuning
• Pretraining Scaling Laws for Generative Evaluations of Language Models
• SolContractEval: A Benchmark for Evaluating Contract-Level Solidity Code Generation
• Efficient Multi-turn RL for GUI Agents via Decoupled Training and Adaptive Data Curation
• Your Dense Retriever is Secretly an Expeditious Reasoner
• ABC-Eval: Benchmarking Large Language Models on Symbolic Music Understanding and Instruction Following
• DentVLM: A Multimodal Vision-Language Model for Comprehensive Dental Diagnosis and Enhanced Clinical Practice
• PARROT: A Benchmark for Evaluating LLMs in Cross-System SQL Translation
• Scaling Policy Compliance Assessment in Language Models with Policy Reasoning Traces
• Mitigating Spurious Correlations in LLMs via Causality-Aware Post-Training
• Tree Reward-Aligned Search for TReASURe in Masked Diffusion Language Models
• Tagging the Thought: Unlocking Personalization Reasoning via Reinforcement Learning
• Tracing Uncertainty in Language Model "Reasoning"
• Agentic AI Reasoning for Mobile Edge General Intelligence: Fundamentals, Approaches, and Directions
• From Harm to Help: Turning Reasoning In-Context Demos into Assets for Reasoning LMs
• Critique to Verify: Accurate and Honest Test-Time Scaling with RL-Trained Verifiers
• RHYTHM: Reasoning with Hierarchical Temporal Tokenization for Human Mobility
• Follow-Your-Preference: Towards Preference-Aligned Image Inpainting
• Local Success Does Not Compose: Benchmarking Large Language Models for Compositional Formal Verification
• WirelessMathLM: Teaching Mathematical Reasoning for LLMs in Wireless Communications with Reinforcement Learning
• JE-IRT: A Geometric Lens on LLM Abilities through Joint Embedding Item Response Theory
• CapRL: Stimulating Dense Image Caption Capabilities via Reinforcement Learning
• WebGen-Agent: Enhancing Interactive Website Generation with Multi-Level Feedback and Step-Level Reinforcement Learning
• Language Models Can Learn from Verbal Feedback Without Scalar Rewards
• Variational Reasoning for Language Models
• Learn the Ropes, Then Trust the Wins: Self-imitation with Progressive Exploration for Agentic Reinforcement Learning
• Dynamic Experts Search: Enhancing Reasoning in Mixture-of-Experts LLMs at Test Time
• Think Socially via Cognitive Reasoning
• The Emergence of Altruism in Large-Language-Model Agents Society
• Do LLM Agents Know How to Ground, Recover, and Assess? A Benchmark for Epistemic Competence in Information-Seeking Agents
• Group Critical-token Policy Optimization for Autoregressive Image Generation
• GeoSketch: A Neural-Symbolic Approach to Geometric Multimodal Reasoning with Auxiliary Line Construction and Affine Transformation
• RAU: Reference-based Anatomical Understanding with Vision Language Models
• Advancing Natural Language Formalization to First Order Logic with Fine-tuned LLMs
• RAPID³: Tri-Level Reinforced Acceleration Policies for Diffusion Transformer
• Rule-Based Reinforcement Learning for Document Image Classification with Vision Language Models
• Safety Compliance: Rethinking LLM Safety Reasoning through the Lens of Compliance
• Towards Faithful Reasoning in Remote Sensing: A Perceptually-Grounded GeoSpatial Chain-of-Thought for Vision-Language Models
• Mixture of Detectors: A Compact View of Machine-Generated Text Detection
• S2J: Bridging the Gap Between Solving and Judging Ability in Generative Reward Models
• Code once, Run Green: Automated Green Code Translation in Serverless Computing
• The Thinking Spectrum: An Empirical Study of Tunable Reasoning in LLMs through Model Merging
• Lightweight Structured Multimodal Reasoning for Clinical Scene Understanding in Robotics
• Goal-Guided Efficient Exploration via Large Language Model in Reinforcement Learning
• RISK: A Framework for GUI Agents in E-commerce Risk Management
• Geo-R1: Improving Few-Shot Geospatial Referring Expression Understanding with Reinforcement Fine-Tuning
• QoNext: Towards Next-generation QoE for Foundation Models
• No Prompt Left Behind: Exploiting Zero-Variance Prompts in LLM Reinforcement Learning via Entropy-Guided Advantage Shaping
• KnowMT-Bench: Benchmarking Knowledge-Intensive Long-Form Question Answering in Multi-Turn Dialogues
• Perception-Consistency Multimodal Large Language Models Reasoning via Caption-Regularized Policy Optimization
• Defending MoE LLMs against Harmful Fine-Tuning via Safety Routing Alignment
• SoK: Potentials and Challenges of Large Language Models for Reverse Engineering
• FastGRPO: Accelerating Policy Optimization via Concurrency-aware Speculative Decoding and Online Draft Learning
• MIRG-RL: Multi-Image Reasoning and Grounding with Reinforcement Learning
• PSRT: Accelerating LRM-based Guard Models via Prefilled Safe Reasoning Traces
• UniVid: Unifying Vision Tasks with Pre-trained Video Generation Models
• Rethinking RoPE Scaling in Quantized LLM: Theory, Outlier, and Channel-Band Analysis with Weight Rescaling
• POLO: Preference-Guided Multi-Turn Reinforcement Learning for Lead Optimization
• UISim: An Interactive Image-Based UI Simulator for Dynamic Mobile Environments
• Thinking in Many Modes: How Composite Reasoning Elevates Large Language Model Performance with Limited Data
• AI Brown and AI Koditex: LLM-Generated Corpora Comparable to Traditional Corpora of English and Czech Texts
• DM-Bench: Benchmarking LLMs for Personalized Decision Making in Diabetes Management
• Rethinking Large Language Model Distillation: A Constrained Markov Decision Process Perspective
• Why Chain of Thought Fails in Clinical Text Understanding
• Learning to Reason with Mixture of Tokens
• One Model, Many Morals: Uncovering Cross-Linguistic Misalignments in Computational Moral Reasoning
• LLMTrace: A Corpus for Classification and Fine-Grained Localization of AI-Written Text
• MMR1: Enhancing Multimodal Reasoning with Variance-Aware Sampling and Open Resources
• Query-Centric Graph Retrieval Augmented Generation
• Expanding Reasoning Potential in Foundation Model by Learning Diverse Chains of Thought Patterns
• TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Them
• Disagreements in Reasoning: How a Model's Thinking Process Dictates Persuasion in Multi-Agent Systems
• GeoRef: Referring Expressions in Geometry via Task Formulation, Synthetic Supervision, and Reinforced MLLM-based Solutions
• Toward Robust and Efficient ML-Based GPU Caching for Modern Inference
• PseudoBridge: Pseudo Code as the Bridge for Better Semantic and Logic Alignment in Code Retrieval
• ToMPO: Training LLM Strategic Decision Making from a Multi-Agent Perspective
• RollPacker: Mitigating Long-Tail Rollouts for Fast, Synchronous RL Post-Training
• On Theoretical Interpretations of Concept-Based In-Context Learning
• RL Grokking Recipe: How Does RL Unlock and Transfer New Algorithms in LLMs?
• ScaleDiff: Scaling Difficult Problems for Advanced Mathematical Reasoning
• InsightGUIDE: An Opinionated AI Assistant for Guided Critical Reading of Scientific Literature
• Thinking Augmented Pre-training
• Automated Multi-Agent Workflows for RTL Design
• Probability Signature: Bridging Data Semantics and Embedding Structure in Language Models
• Integrated Framework for LLM Evaluation with Answer Generation
• Embodied AI: From LLMs to World Models
• AMLA: MUL by ADD in FlashAttention Rescaling
• WEST: LLM based Speech Toolkit for Speech Understanding, Generation, and Interaction
• PromptCoT 2.0: Scaling Prompt Synthesis for Large Language Model Reasoning
• DAOpt: Modeling and Evaluation of Data-Driven Optimization under Uncertainty with LLMs
• VCRL: Variance-based Curriculum Reinforcement Learning for Large Language Models
• SIM-CoT: Supervised Implicit Chain-of-Thought
• SKYLENAGE Technical Report: Mathematical Reasoning and Contest-Innovation Benchmarks for Multi-Level Math Evaluation
• UserRL: Training Interactive User-Centric Agent via Reinforcement Learning
• The Conductor and the Engine: A Path Towards Co-Designed Reasoning
• LOCA: Logical Chain Augmentation for Scientific Corpus Cleaning
• bi-GRPO: Bidirectional Optimization for Jailbreak Backdoor Injection on LLMs
• Are We Scaling the Right Thing? A System Perspective on Test-Time Scaling
• Semantic-Aware Fuzzing: An Empirical Framework for LLM-Guided, Reasoning-Driven Input Mutation
• Exact Action Values Are Not Enough: Rollout-Verified Reinforcement Fine-Tuning of a Reasoning Model for Multi-Zone VAV Control
• DFlash: Block Diffusion for Flash Speculative Decoding
• The homogenizing effect of large language models on human expression and thought
• GEPO: Group Expectation Policy Optimization for Stable Heterogeneous Reinforcement Learning
• Reinforcement Learning on Pre-Training Data
• Agentic Reinforcement Learning with Implicit Step Rewards
• Citrus-V: Advancing Medical Foundation Models with Unified Medical Image Grounding for Clinical Reasoning
• From latent factors to language: a user study on LLM-generated explanations for an inherently interpretable matrix-based recommender system
• How Far are VLMs from Visual Spatial Intelligence? A Benchmark-Driven Perspective
• MAPO: Mixed Advantage Policy Optimization
• Pay More Attention To Audio: Mitigating Imbalance of Cross-Modal Attention in Large Audio Language Models
• AECBench: A Hierarchical Benchmark for Knowledge Evaluation of Large Language Models in the AEC Field
• MECap-R1: Emotion-aware Policy with Reinforcement Learning for Multimodal Emotion Captioning
• PipelineRL: Faster On-policy Reinforcement Learning for Long Sequence Generation
• OraPO: Oracle-educated Reinforcement Learning for Data-efficient and Factual Radiology Report Generation
• No Verifiable Reward for Prosody: Toward Preference-Guided Prosody Learning in TTS
• Residual Off-Policy RL for Finetuning Behavior Cloning Policies
• NGRPO: Negative-enhanced Group Relative Policy Optimization
• Spacer: Towards Engineered Scientific Inspiration
• Introducing LongCat-Flash-Thinking: A Technical Report
• GRPO++: Enhancing Dermatological Reasoning under Low Resource Settings
• What Characterizes Effective Reasoning? Revisiting Length, Review, and Structure of CoT
• CogniLoad: A Synthetic Natural Language Reasoning Benchmark With Tunable Length, Intrinsic Difficulty, and Distractor Density
• ConfClip: Confidence-Weighted and Clipped Reward for Reinforcement Learning in LLMs
• AuditoryBench++: Can Language Models Understand Auditory Knowledge without Hearing?
• Correlation or Causation: Analyzing the Causal Structures of LLM and LRM Reasoning Process
• LLaVul: A Multimodal LLM for Interpretable Vulnerability Reasoning about Source Code
• Clotho: Measuring Task-Specific Pre-Generation Test Adequacy for LLM Inputs
• Exploiting Tree Structure for Credit Assignment in RL Training of LLMs
• MSCoRe: A Benchmark for Multi-Stage Collaborative Reasoning in LLM Agents
• Evaluating the Safety and Skill Reasoning of Large Reasoning Models Under Compute Constraints
• TempSamp-R1: Effective Temporal Sampling with Reinforcement Fine-Tuning for Video LLMs
• D-REX: A Benchmark for Detecting Deceptive Reasoning in Large Language Models
• GeoPQA: Bridging the Visual Perception Gap in MLLMs for Geometric Reasoning
• Adaptive Overclocking: Dynamic Control of Thinking Path Length via Real-Time Reasoning Signals
• PTQTP: Post-Training Quantization to Trit-Planes for Large Language Models
• AirQA: A Comprehensive QA Dataset for AI Research with Instance-Level Evaluation
• R1-Fuzz: Specializing Language Models for Textual Fuzzing via Reinforcement Learning
• SCAN: Self-Denoising Monte Carlo Annotation for Robust Process Reward Learning
• RubikSQL: Lifelong Learning Agentic Knowledge Base as an Industrial NL2SQL System
• Evaluation of Causal Reasoning for Large Language Models in Contextualized Clinical Scenarios of Laboratory Test Interpretation
• RPG: A Repository Planning Graph for Unified and Scalable Codebase Generation
• SightSound-R1: Cross-Modal Reasoning Distillation from Vision to Audio Language Models
• Self-Rewarding Rubric-Based Reinforcement Learning for Open-Ended Reasoning
• TennisTV: Do Multimodal Large Language Models Understand Tennis Rallies?
• ChartMaster: Advancing Chart-to-Code Generation with Real-World Charts and Chart Similarity Reinforcement Learning
• Spectral Logit Sculpting: Adaptive Low-Rank Logit Transformation for Controlled Text Generation
• LiteLong: Resource-Efficient Long-Context Data Synthesis for LLMs
• BTL-UI: Blink-Think-Link Reasoning Model for GUI Agent
• GPO: Learning from Critical Steps to Improve LLM Reasoning
• Thinking in cocktail party: Chain-of-Thought and reinforcement learning for target speaker automatic speech recognition
• MoE-CE: Enhancing Generalization for Deep Learning based Channel Estimation via a Mixture-of-Experts Framework
• SciEvent: Benchmarking Multi-domain Scientific Event Extraction
• TISDiSS: A Training-Time and Inference-Time Scalable Framework for Discriminative Source Separation
• Emulating Human-like Adaptive Vision for Efficient and Flexible Machine Visual Perception
• Generalizable Geometric Image Caption Synthesis
• RynnVLA-001: Using Human Demonstrations to Improve Robot Manipulation
• How Good are Foundation Models in Step-by-Step Embodied Reasoning?
• Self-Improving Embodied Foundation Models
• MedFact-R1: Towards Factual Medical Reasoning via Pseudo-Label Augmentation
• T2I-ReasonBench: Benchmarking Reasoning-Informed Text-to-Image Generation
• Embodied Arena: A Comprehensive, Unified, and Evolving Evaluation Platform for Embodied AI
• Empathy-R1: A Chain-of-Empathy and Reinforcement Learning Framework for Long-Form Mental Health Support
• TreePO: Bridging the Gap of Policy Optimization and Efficacy and Inference Efficiency with Heuristic Tree-based Modeling
• Fleming-R1: Toward Expert-Level Medical Reasoning via Reinforcement Learning
• Explicit vs. Implicit Biographies: Evaluating and Adapting LLM Information Extraction on Wikidata-Derived Texts
• TDRM: Smooth Reward Models with Temporal Difference for LLM RL and Inference
• CodeFuse-CR-Bench: A Comprehensiveness-aware Benchmark for End-to-End Code Review Evaluation in Python Projects
• FlowRL: Matching Reward Distributions for LLM Reasoning
• AToken: A Unified Tokenizer for Vision
• AdaThinkDrive: Adaptive Thinking via Reinforcement Learning for Autonomous Driving
• THOR: Tool-Integrated Hierarchical Optimization via RL for Mathematical Reasoning
• FLAME: A Serving System Optimized for Large-Scale Generative Recommendation with Efficiency
• Geometric Uncertainty for Detecting and Correcting Hallucinations in LLMs
• Bellman Optimality of Average-Reward Robust Markov Decision Processes with a Constant Gain
• VerilogMonkey: Exploring Parallel Scaling for Automated Verilog Code Generation with LLMs
• LLM-I: LLMs are Naturally Interleaved Multimodal Creators
• CORE-RAG: Lossless Compression for Retrieval-Augmented LLMs via Reinforcement Learning
• Meta-R1: Empowering Large Reasoning Models with Metacognition
• Benchmarking ChatGPT and DeepSeek in April 2025: A Novel Dual Perspective Sentiment Analysis Using Lexicon-Based and Deep Learning Approaches
• LATTS: Locally Adaptive Test-Time Scaling
• Single-stream Policy Optimization
• EconProver: Towards More Economical Test-Time Scaling for Automated Theorem Proving
• Validating Solidity Code Defects using Symbolic and Concrete Execution powered by Large Language Models
• Leveraging Large Language Models to Effectively Generate Visual Data for Canine Musculoskeletal Diagnoses
• Participatory AI: A Scandinavian Approach to Human-Centered AI
• Tool-R1: Sample-Efficient Reinforcement Learning for Agentic Tool Use
• Large Language Models Imitate Logical Reasoning, but at what Cost?
• WebResearcher: Unleashing unbounded reasoning capability in Long-Horizon Agents
• FastMTP: Accelerating LLM Inference with Enhanced Multi-Token Prediction
• ActiveVLN: Towards Active Exploration via Multi-Turn RL in Vision-and-Language Navigation
• Scaling Agents via Continual Pre-training
• Reasoned Safety Alignment: Ensuring Jailbreak Defense via Answer-Then-Check
• D²HScore: Reasoning-Aware Hallucination Detection via Semantic Breadth and Depth Analysis in LLMs
• UI-S1: Advancing GUI Automation via Semi-online Reinforcement Learning
• Omni-CLST: Error-aware Curriculum Learning with guided Selective chain-of-Thought for audio question answering
• Free-MAD: Consensus-Free Multi-Agent Debate
• Enhancing Generalization in Vision-Language-Action Models by Preserving Pretrained Representations
• Towards Automated Error Discovery: A Study in Conversational AI
• Reasoning Under Uncertainty: Exploring Probabilistic Reasoning Capabilities of LLMs
• GC-VLN: Instruction as Graph Constraints for Training-free Vision-and-Language Navigation
• MatSKRAFT: A framework for large-scale materials knowledge extraction from scientific tables
• DeepDive: Advancing Deep Search Agents with Knowledge Graphs and Multi-Turn RL
• Understanding Outer Optimizers in Local SGD: Learning Rates, Momentum, and Acceleration
• Decoding Alignment: A Critical Survey of LLM Development Initiatives through Value-setting and Data-centric Lens
• MagicMirror: A Large-Scale Dataset and Benchmark for Fine-Grained Artifacts Assessment in Text-to-Image Generation
• Prompt Injection Attacks on LLM Generated Reviews of Scientific Publications
• Smart Trial: Evaluating the Use of Large Language Models for Recruiting Clinical Trial Participants via Social Media
• Towards Understanding Visual Grounding in Visual Language Models
• Inpainting-Guided Policy Optimization for Diffusion Large Language Models
• SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning
• CDE: Curiosity-Driven Exploration for Efficient Reinforcement Learning in Large Language Models
• Retrieval-Augmented Generation for Reliable Interpretation of Radio Regulations
• LAVA: Language Model Assisted Verbal Autopsy for Cause-of-Death Determination
• Tree-OPO: Off-policy Monte Carlo Tree-Guided Advantage Optimization for Multistep Reasoning
• AI Reasoning for Wireless Communications and Networking: A Survey and Perspectives
• Compass-v3: Scaling Domain-Specific LLMs for Multilingual E-Commerce in Southeast Asia
• MR-UIE: Multi-Perspective Reasoning with Reinforcement Learning for Universal Information Extraction
• Latency and Token-Aware Test-Time Compute
• Fast attention mechanisms: a tale of parallelism
• Merge-of-Thought Distillation
• SWE-Mirror: Scaling Issue-Resolving Datasets by Mirroring Issues Across Repositories
• MCP-AgentBench: Evaluating Real-World Language Agent Performance with MCP-Mediated Tools
• Facet: highly efficient E(3)-equivariant networks for interatomic potentials
• An Iterative LLM Framework for SIBT utilizing RAG-based Adaptive Weight Optimization
• Accelerating Reinforcement Learning Algorithms Convergence using Pre-trained Large Language Models as Tutors With Advice Reusing
• SocialNav-SUB: Benchmarking VLMs for Scene Understanding in Social Robot Navigation
• FinZero: Launching Multi-modal Financial Time Series Forecast with Large Reasoning Model
• AgentGym-RL: Training LLM Agents for Long-Horizon Decision Making through Multi-Turn Reinforcement Learning
• Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search
• RaC: Robot Learning for Long-Horizon Tasks by Scaling Recovery and Correction
• Towards Generalized Routing: Model and Agent Orchestration for Adaptive and Efficient Inference
• Temporal Counterfactual Explanations of Behaviour Tree Decisions
• VL Norm: Rethink Loss Aggregation in RLVR
• Language Self-Play For Data-Free Training
• Parallel-R1: Towards Parallel Thinking via Reinforcement Learning
• SCA-LLM: Spectral-Attentive LLM-Based Wireless World Modeling for Agentic Communications
• Testing chatbots on the creation of encoders for audio conditioned image generation
• Performative Thinking? The Brittle Correlation Between CoT Length and Problem Complexity
• Text2Touch: Tactile In-Hand Manipulation with LLM-Designed Reward Functions
• Unleashing the True Potential of LLMs: A Feedback-Triggered Self-Correction with Long-Term Multipath Decoding
• Staying in the Sweet Spot: Responsive Reasoning Evolution via Capability-Adaptive Hint Scaffolding
• WST: Weak-to-Strong Knowledge Transfer via Reinforcement Learning
• MachineLearningLM: Scaling Many-shot In-context Learning via Continued Pretraining
• Aligning Large Vision-Language Models by Deep Reinforcement Learning and Direct Preference Optimization
• Towards Trustworthy Agentic IoEV: AI Agents for Explainable Cyberthreat Mitigation and State Analytics
• Reinforcement Learning Foundations for Deep Research Systems: A Survey
• WebExplorer: Explore and Evolve for Training Long-Horizon Web Agents
• Test-Time Scaling in Reasoning Models Is Not Effective for Knowledge-Intensive Tasks Yet
• SFR-DeepResearch: Towards Effective Reinforcement Learning for Autonomously Reasoning Single Agents
• Spatial Reasoning with Vision-Language Models in Ego-Centric Multi-View Scenes
• RL Fine-Tuning Heals OOD Forgetting in SFT
• The Majority is not always right: RL training for solution aggregation
• Interleaving Reasoning for Better Text-to-Image Generation
• Scaling up Multi-Turn Off-Policy RL and Multi-Agent Tree Search for LLM Step-Provers
• MaaSO: SLO-aware Orchestration of Heterogeneous Model Instances for MaaS
• Domain-Aware RAG: MoL-Enhanced RL for Efficient Training and Scalable Retrieval
• SVGauge: Towards Human-Aligned Evaluation for SVG Generation
• Rethinking Reasoning Quality in Large Language Models through Enhanced Chain-of-Thought via RL
• GeoAnalystBench: A GeoAI benchmark for assessing large language models for spatial analysis workflow and code generation
• Reverse-Engineered Reasoning for Open-Ended Generation
• From Long to Short: LLMs Excel at Trimming Own Reasoning Chains
• RL Is Neither a Panacea Nor a Mirage: Understanding Supervised vs. Reinforcement Learning Fine-Tuning for LLMs
• Simple Optimizers for Convex Aligned Multi-Objective Optimization
• Few-Shot Query Intent Detection via Relation-Aware Prompt Learning
• FlowSeek: Optical Flow Made Easier with Depth Foundation Models and Motion Bases
• Scaling Performance of Large Language Model Pretraining
• Less is More Tokens: Efficient Math Reasoning via Difficulty-Aware Chain-of-Thought Distillation
• A Knowledge-Driven Diffusion Policy for End-to-End Autonomous Driving Based on Expert Routing
• COMMET: A System for Human-Induced Conflicts in Mobile Manipulation of Everyday Tasks
• Symbolic Graphics Programming with Large Language Models
• CoVeR: Conformal Calibration for Versatile and Reliable Autoregressive Next-Token Prediction
• Hunyuan-MT Technical Report
• Inverse IFEval: Can LLMs Unlearn Stubborn Training Conventions to Follow Real Instructions?
• TAGAL: Tabular Data Generation using Agentic LLM Methods
• A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models
• Emergent Hierarchical Reasoning in LLMs through Reinforcement Learning
• Generative Auto-Bidding in Large-Scale Competitive Auctions via Diffusion Completer-Aligner
• AgenTracer: Who Is Inducing Failure in the LLM Agentic Systems?
• AetherCode: Evaluating LLMs' Ability to Win In Premier Programming Competitions
• Advancing SLM Tool-Use Capability using Reinforcement Learning
• Empowering Lightweight MLLMs with Reasoning via Long CoT SFT
• SimpleTIR: End-to-End Reinforcement Learning for Multi-Turn Tool-Integrated Reasoning
• Planning with Reasoning using Vision Language World Model
• SATORI: Static Test Oracle Generation for REST APIs
• OwkinZero: Accelerating Biological Discovery with AI
• Implicit Reasoning in Large Language Models: A Comprehensive Survey
• UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning
• DCPO: Dynamic Clipping Policy Optimization
• MedOmni-45°: A Safety-Performance Benchmark for Reasoning-Oriented LLMs in Medicine
• Jointly Reinforcing Diversity and Quality in Language Model Generations
• Baichuan-M2: Scaling Medical Capability with Large Verifier System
• Towards Agents That Know When They Don't Know: Uncertainty as a Control Signal for Structured Reasoning
• Abex-rat: Synergizing Abstractive Augmentation and Adversarial Training for Classification of Occupational Accident Reports
• Succeed or Learn Slowly: Sample Efficient Off-Policy Reinforcement Learning for Mobile App Control
• Reinforced Visual Perception with Tools
• Kwai Keye-VL 1.5 Technical Report
• Bridging the Gap in Ophthalmic AI: MM-Retinal-Reason Dataset and OphthaReason Model toward Dynamic Multimodal Reasoning
• Zero-shot Cross-lingual NER via Mitigating Language Difference: An Entity-aligned Translation Perspective
• Robix: A Unified Model for Robot Interaction, Reasoning and Planning
• VerlTool: Towards Holistic Agentic Reinforcement Learning with Tool Use
• Reinforcement Learning for Machine Learning Engineering Agents
• When LLM Meets Time Series: Can LLMs Perform Multi-Step Time Series Reasoning and Inference
• CARE: Decoding Time Safety Alignment via Rollback and Introspection Intervention
• An Automated Attack Investigation Approach Leveraging Threat-Knowledge-Augmented Large Language Models
• Towards High Data Efficiency in Reinforcement Learning with Verifiable Reward
• Aligning Reasoning LLMs for Materials Discovery with Physics-aware Rejection Sampling
• LLaVA-Critic-R1: Your Critic Model is Secretly a Strong Policy Model
• LLM Encoder vs. Decoder: Robust Detection of Chinese AI-Generated Text with LoRA
• EviNote-RAG: Enhancing RAG Models via Answer-Supportive Evidence Notes
• The Gold Medals in an Empty Room: Diagnosing Metalinguistic Reasoning in LLMs with Camlang
• Open Data Synthesis For Deep Research
• Scaling Up, Speeding Up: A Benchmark of Speculative Decoding for Efficient LLM Test-Time Scaling
• ParaThinker: Native Parallel Thinking as a New Paradigm to Scale LLM Test-time Compute
• Balanced Actor Initialization: Stable RLHF Training of Distillation-Based Reasoning Models
• Do Cognitively Interpretable Reasoning Traces Improve LLM Performance?
• Think in Games: Learning to Reason in Games via Reinforcement Learning with Large Language Models
• Know When to Explore: Difficulty-Aware Certainty as a Guide for LLM Reinforcement Learning
• UItron: Foundational GUI Agent with Advanced Perception and Planning
• Mirage or Method? How Model-Task Alignment Induces Divergent RL Conclusions
• R-4B: Incentivizing General-Purpose Auto-Thinking Capability in MLLMs via Bi-Mode Annealing and Reinforce Learning
• MSRS: Evaluating Multi-Source Retrieval-Augmented Generation
• Droplet3D: Commonsense Priors from Videos Facilitate 3D Generation
• AI-SearchPlanner: Modular Agentic Search via Pareto-Optimal Multi-Objective Reinforcement Learning
• End-to-End Agentic RAG System Training for Traceable Diagnostic Reasoning
• Lethe: Purifying Backdoored Large Language Models with Knowledge Dilution
• GDS Agent for Graph Algorithmic Reasoning
• rStar2-Agent: Agentic Reasoning Technical Report
• OneRec-V2 Technical Report
• Train-Once Plan-Anywhere Kinodynamic Motion Planning via Diffusion Trees
• Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning
• NiceWebRL: a Python library for human subject experiments with reinforcement learning environments
• Quantum Verifiable Rewards for Post-Training Qiskit Code Assistant
• MathBuddy: A Multimodal System for Affective Math Tutoring
• Evaluating Language Model Reasoning about Confidential Information
• IELDG: Suppressing Domain-Specific Noise with Inverse Evolution Layers for Domain Generalized Semantic Segmentation
• Think in Blocks: Adaptive Reasoning from Direct Response to Deep Reasoning
• LLM-Driven Self-Refinement for Embodied Drone Task Planning
• SynthCoder: A Synthetical Strategy to Tune LLMs for Code Completion
• Entropy-Guided Loop: Achieving Reasoning through Uncertainty-Aware Generation
• GENIE-ASI: Generative Instruction and Executable Code for Analog Subcircuit Identification
• OmniHuman-1.5: Instilling an Active Mind in Avatars via Cognitive Simulation
• DeepMEL: A Multi-Agent Collaboration Framework for Multimodal Entity Linking
• Understanding Tool-Integrated Reasoning
• Reasoning LLMs in the Medical Domain: A Literature Survey
• MUA-RL: Multi-turn User-interacting Agent Reinforcement Learning for agentic tool use
• CAC-CoT: Connector-Aware Compact Chain-of-Thought for Efficient Reasoning Data Synthesis Across Dual-System Cognitive Tasks
• Beyond Benchmark: LLMs Evaluation with an Anthropomorphic and Value-oriented Roadmap
• RLMR: Reinforcement Learning with Mixed Rewards for Creative Writing
• VERIRL: Boosting the LLM-based Verilog Code Generation via Reinforcement Learning
• DemoBias: An Empirical Study to Trace Demographic Biases in Vision Foundation Models
• A.S.E: A Repository-Level Benchmark for Evaluating Security in AI-Generated Code
• Mobile-Agent-v3: Fundamental Agents for GUI Automation
• CARFT: Boosting LLM Reasoning via Contrastive Learning with Annotated Chain-of-Thought-based Reinforced Fine-Tuning
• Dream 7B: Diffusion Large Language Models
• Fin-PRM: A Domain-Specialized Process Reward Model for Financial Reasoning in Large Language Models
• Survey of Vision-Language-Action Models for Embodied Manipulation
• Deep Think with Confidence
• Discovering Hidden Algebraic Structures via Transformers with Rank-Aware Beam GRPO
• Unveiling Trust in Multimodal Large Language Models: Evaluation, Analysis, and Mitigation
• Nemotron-CC-Math: A 133 Billion-Token-Scale High Quality Math Pretraining Dataset
• Investigation of the Inter-Rater Reliability between Large Language Models and Human Raters in Qualitative Analysis
• DuPO: Enabling Reliable LLM Self-Verification via Dual Preference Optimization
• PepThink-R1: LLM for Interpretable Cyclic Peptide Optimization with CoT SFT and Reinforcement Learning
• NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning Model
• Tinker: Diffusion's Gift to 3D--Multi-View Consistent Editing From Sparse Inputs without Per-Scene Optimization
• MoEcho: Exploiting Side-Channel Attacks to Compromise User Privacy in Mixture-of-Experts LLMs
• FedRAIN-Lite: Federated Reinforcement Algorithms for Improving Idealised Numerical Weather and Climate Models
• HiCL: Hippocampal-Inspired Continual Learning
• Lexical Hints of Accuracy in LLM Reasoning Chains
• Prompt-Based One-Shot Exact Length-Controlled Generation with LLMs
• Expertise-aware Multi-LLM Recruitment and Collaboration for Medical Decision-Making
• CausalPlan: Empowering Efficient LLM Multi-Agent Collaboration Through Causality-Driven Planning
• HumanPCR: Probing MLLM Capabilities in Diverse Human-Centric Scenes
• Towards a Larger Model via One-Shot Federated Learning on Heterogeneous Client Models
• LM Agents May Fail to Act on Their Own Risk Knowledge
• Beyond Pass@1: Self-Play with Variational Problem Synthesis Sustains RLVR
• Embodied-R1: Reinforced Embodied Reasoning for General Robotic Manipulation
• ComputerRL: Scaling End-to-End Online Reinforcement Learning for Computer Use Agents
• Datarus-R1: An Adaptive Multi-Step Reasoning LLM for Automated Data Analysis
• AI Agents for Photonic Integrated Circuit Design Automation
• Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Survey
• CardAIc-Agents: A Multimodal Framework with Hierarchical Adaptation for Cardiac Care Support
• G²RPO-A: Guided Group Relative Policy Optimization with Adaptive Guidance
• Involuntary Jailbreak: On Self-Prompting Attacks
• Atom-Searcher: Enhancing Agentic Deep Research via Fine-Grained Atomic Thought Reward
• RAJ-PGA: Reasoning-Activated Jailbreak and Principle-Guided Alignment Framework for Large Reasoning Models
• DESIGNER: Design-Logic-Guided Multidisciplinary Data Synthesis for LLM Reasoning
• Vision-G1: Towards General Vision Language Reasoning with Multi-Domain Data Curation
• Mini-Omni-Reasoner: Token-Level Thinking-in-Speaking in Large Speech Models
• Reinforcement Learning with Rubric Anchors
• MDPO: Overcoming the Training-Inference Divide of Masked Diffusion Language Models
• Consensus or Conflict? Fine-Grained Evaluation of Conflicting Answers in Question-Answering
• RadarQA: Multi-modal Quality Analysis of Weather Radar Forecasts
• ProtTeX-CC: Activating In-Context Learning in Protein LLM via Two-Stage Instruction Compression
• Improving Pre-Trained Vision-Language-Action Policies with Model-Based Search
• Mitigating Hallucinations in Large Language Models via Causal Reasoning
• Too Easily Fooled? Prompt Injection Breaks LLMs on Frustratingly Simple Multiple-Choice Questions
• Talk Less, Fly Lighter: Autonomous Semantic Compression for UAV Swarm Communication via LLMs
• QuarkMed Medical Foundation Model Technical Report
• Discovering Expert-Level Nash Equilibrium Algorithms with Large Language Models
• Optimizing Token Choice for Code Watermarking: An RL Approach
• Thyme: Think Beyond Images
• On-Policy RL Meets Off-Policy Experts: Harmonizing Supervised Fine-Tuning and Reinforcement Learning via Dynamic Weighting
• Aware First, Think Less: Dynamic Boundary Self-Awareness Drives Extreme Reasoning Efficiency in Large Language Models
• SeamlessFlow: A Trainer Agent Isolation RL Framework Achieving Bubble-Free Pipelines via Tag Scheduling
• AI Agentic Programming: A Survey of Techniques, Challenges, and Opportunities
• Inclusion Arena: An Open Platform for Evaluating Large Foundation Models with Real-World Apps
• MM-R1: Unleashing the Power of Unified Multimodal Large Language Models for Personalized Image Generation
• Retrieval-augmented reasoning with lean language models
• ETTRL: Balancing Exploration and Exploitation in LLM Test-Time Reinforcement Learning Via Entropy Mechanism
• HOID-R1: Reinforcement Learning for Open-World Human-Object Interaction Detection Reasoning with Multimodal Large Language Model
• Beyond Solving Math Quiz: Evaluating the Ability of Large Reasoning Models to Ask for Information
• Audio Flamingo Sound-CoT Technical Report: Improving Chain-of-Thought Reasoning in Sound Understanding
• SSRL: Self-Search Reinforcement Learning
• UI-Venus Technical Report: Building High-performance UI Agents with RFT
• FROGENT: An End-to-End Full-process Drug Design Agent
• Retro-Expert: Collaborative Reasoning for Interpretable Retrosynthesis
• We-Math 2.0: A Versatile MathBook System for Incentivizing Visual Mathematical Reasoning
• Few-shot Vision-based Human Activity Recognition with MLLM-based Visual Reinforcement Learning
• Making Qwen3 Think in Korean with Reinforcement Learning
• Artificial Emotion: A Survey of Theories and Debates on Realising Emotion in Artificial Intelligence
• MAC: A Live Benchmark for Multimodal Large Language Models in Scientific Understanding
• Promoting Efficient Reasoning with Verifiable Stepwise Reward
• Pruning Long Chain-of-Thought of Large Reasoning Models via Small-Scale Preference Optimization
• Multi-Turn Puzzles: Evaluating Interactive Reasoning and Strategic Dialogue in LLMs
• mSCoRe: a Multilingual and Scalable Benchmark for Skill-based Commonsense Reasoning
• Echo-4o: Harnessing the Power of GPT-4o Synthetic Images for Improved Image Generation
• VisCodex: Unified Multimodal Code Generation via Merging Vision and Coding Models
• Physical Autoregressive Model for Robotic Manipulation without Action Pretraining
• Sample More to Think Less: Group Filtered Policy Optimization for Concise Reasoning
• ReqInOne: A Large Language Model-Based Agent for Software Requirements Specification Generation
• Shadow in the Cache: Unveiling and Mitigating Privacy Risks of KV-cache in LLM Inference
• Distilling LLM Prior to Flow Model for Generalizable Agent's Imagination in Object Goal Navigation
• ParallelSearch: Train your LLMs to Decompose Query and Search Sub-queries in Parallel with Reinforcement Learning
• Compass-Thinker-7B Technical Report
• ASPD: Unlocking Adaptive Serial-Parallel Decoding by Exploring Intrinsic Parallelism in LLMs
• Reducing Cognitive Overhead in Tool Use via Multi-Small-Agent Reinforcement Learning
• Cluster Topology-Driven Placement of Experts Reduces Network Traffic in MoE Inference
• A Survey of Optimization Modeling Meets LLMs: Progress and Future Directions
• PersRM-R1: Enhance Personalized Reward Modeling with Reinforcement Learning
• Scaling Learned Image Compression Models up to 1 Billion
• Bridging Formal Language with Chain-of-Thought Reasoning to Geometry Problem Solving
• Train Long, Think Short: Curriculum Learning for Efficient Reasoning
• AgriGPT: a Large Language Model Ecosystem for Agriculture
• Time Is a Feature: Exploiting Temporal Dynamics in Diffusion Language Models
• Jinx: Unlimited LLMs for Probing Alignment Failures
• Street-Level AI: Are Large Language Models Ready for Real-World Judgments?
• Reinforcement Learning for Large Model: A Survey
• Audio-Thinker: Guiding Audio Language Model When and How to Think via Reinforcement Learning
• \(X\)-evolve: Solution space evolution powered by large language models
• Bottom-up Domain-specific Superintelligence: A Reliable Knowledge Graph is What We Need
• Towards Theoretical Understanding of Transformer Test-Time Computing: Investigation on In-Context Linear Regression
• WideSearch: Benchmarking Agentic Broad Info-Seeking
• Large Language Models for Subjective Language Understanding: A Survey
• Keyword-Centric Prompting for One-Shot Event Detection with Self-Generated Rationale Enhancements
• FEAT: A Multi-Agent Forensic AI System with Domain-Adapted Large Language Model for Automated Cause-of-Death Analysis
• Pareto Multi-Objective Alignment for Language Models
• Beyond Ten Turns: Unlocking Long-Horizon Agentic Search with Large-Scale Asynchronous RL
• Schema Lineage Extraction at Scale: Multilingual Pipelines, Composite Evaluation, and Language-Model Benchmarks
• Omni-SafetyBench: A Benchmark for Safety Evaluation of Audio-Visual Large Language Models
• HealthBranches: Synthesizing Clinically-Grounded Question Answering Datasets via Decision Pathways
• AMFT: Aligning LLM Reasoners by Meta-Learning the Optimal Imitation-Exploration Balance
• AR-GRPO: Training Autoregressive Image Generation Models via Reinforcement Learning
• Generalizing Scaling Laws for Dense and Sparse Large Language Models
• Intuition emerges in Maximum Caliber models at criticality
• LLM Unlearning Without an Expert Curated Dataset
• Towards Integrated Alignment
• When a Paper Has 1000 Authors: Rethinking Citation Metrics in the Era of LLMs
• Bifrost-1: Bridging Multimodal LLMs and Diffusion Models with Patch-level CLIP Latents
• Impact-driven Context Filtering For Cross-file Code Completion
• Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Temporal Grounding
• Mitigating Think-Answer Mismatch in LLM Reasoning Through Noise-Aware Advantage Reweighting
• Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcing
• In-Training Defenses against Emergent Misalignment in Language Models
• WebWatcher: Breaking New Frontier of Vision-Language Deep Research Agent
• Cooper: Co-Optimizing Policy and Reward Models in Reinforcement Learning for Large Language Models
• Shuffle-R1: Efficient RL framework for Multimodal Large Language Models via Data-centric Dynamic Shuffle
• StructVRM: Aligning Multimodal Reasoning with Structured and Verifiable Reward Models
• CodeBoost: Boosting Code LLMs by Squeezing Knowledge from Code Snippets with RL
• ReasoningTrack: Chain-of-Thought Reasoning for Long-term Vision-Language Tracking
• Klear-CodeTest: Scalable Test Case Generation for Code Reinforcement Learning
• IRL-VLA: Training an Vision-Language-Action Policy via Reward World Model
• Can Large Language Models Integrate Spatial Data? Empirical Insights into Reasoning Strengths and Computational Weaknesses
• A Multi-Stage Large Language Model Framework for Extracting Suicide-Related Social Determinants of Health
• InfiGUI-G1: Advancing GUI Grounding with Adaptive Exploration Policy Optimization
• SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience
• StepFun-Formalizer: Unlocking the Autoformalization Potential of LLMs through Knowledge-Reasoning Fusion
• PersonaEval: Are LLM Evaluators Human Enough to Judge Role-Play?
• Thinking With Videos: Multimodal Tool-Augmented Reinforcement Learning for Long Video Reasoning
• GTPO and GRPO-S: Token and Sequence-Level Reward Shaping with Policy Entropy
• GuirlVG: Incentivize GUI Visual Grounding via Empirical Exploration on Reinforcement Learning
• Fine-tuning for Better Few Shot Prompting: An Empirical Comparison for Short Answer Grading
• PromptAL: Sample-Aware Dynamic Soft Prompts for Few-Shot Active Learning
• MiDashengLM: Efficient Audio Understanding with General Audio Captions
• Scaling Linear Attention with Sparse State Expansion
• RealBench: Benchmarking Verilog Generation Models with Real-World IP Designs
• Policy to Assist Iteratively Local Segmentation: Optimising Modality and Location Selection for Prostate Cancer Localisation
• Sotopia-RL: Reward Design for Social Intelligence
• Evo-MARL: Co-Evolutionary Multi-Agent Reinforcement Learning for Internalized Safety
• Self-Questioning Language Models
• Agent Lightning: Train ANY AI Agents with Reinforcement Learning
• Semantic-aware Graph-guided Behavior Sequences Generation with Large Language Models for Smart Homes
• Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets
• R2GenKG: Hierarchical Multi-modal Knowledge Graph for LLM-based Radiology Report Generation
• Frontier: Simulating the Next Generation of LLM Inference Systems
• Who is a Better Player: LLM against LLM
• ContractEval: Benchmarking LLMs for Clause-Level Legal Risk Identification in Commercial Contracts
• Tool-integrated Reinforcement Learning for Repo Deep Search
• CTTS: Collective Test-Time Scaling
• Data Dependency-Aware Code Generation from Enhanced UML Sequence Diagrams
• PentestJudge: Judging Agent Behavior Against Operational Requirements
• Cognitive Loop via In-Situ Optimization: Self-Adaptive Reasoning for Science
• CAMA: Enhancing Mathematical Reasoning in Large Language Models with Causal Knowledge
• CAPO: Towards Enhancing LLM Reasoning through Generative Credit Assignment
• AirTrafficGen: Configurable Air Traffic Scenario Generation with Large Language Models
• Trainable Dynamic Mask Sparse Attention
• MedVLThinker: Simple Baselines for Multimodal Medical Reasoning
• PoeTone: A Framework for Constrained Generation of Structured Chinese Songci with LLMs
• Large-Scale Model Enabled Semantic Communication Based on Robust Knowledge Distillation
• HealthFlow: A Self-Evolving AI Agent with Meta Planning for Autonomous Healthcare Research
• CUPID: Evaluating Personalized and Contextualized Alignment of LLMs from Interactions
• JSidentify-V2: Leveraging Dynamic Memory Fingerprinting for Mini-Game Plagiarism Detection
• Enhancing Math Reasoning in Small-sized LLMs via Preview Difficulty-Aware Intervention
• AffectGPT-R1: Leveraging Reinforcement Learning for Open-Vocabulary Multimodal Emotion Recognition
• MeshLLM: Empowering Large Language Models to Progressively Understand and Generate 3D Mesh
• Causal Graph Fuzzy LLMs: A First Introduction and Applications in Time Series Forecasting
• Expert-Guided LLM Reasoning for Battery Discovery: From AI-Driven Hypothesis to Synthesis and Characterization
• Provably Secure Retrieval-Augmented Generation
• ROVI: A VLM-LLM Re-Captioned Dataset for Open-Vocabulary Instance-Grounded Text-to-Image Generation
• VisionThink: Smart and Efficient Vision Language Model via Reinforcement Learning
• MCeT: Behavioral Model Correctness Evaluation using Large Language Models
• Integrating clinical reasoning into large language model-based diagnosis through etiology-aware attention steering
• Accurate and Consistent Graph Model Generation from Text with Large Language Models
• Multi-Layer Attention is the Amplifier of Demonstration Effectiveness
• MetaAgent: Toward Self-Evolving Agent via Tool Meta-Learning
• Lessons from complex systems science for AI governance
• RL-PLUS: Countering Capability Boundary Collapse of LLMs in Reinforcement Learning with Hybrid-policy Optimization
• Hierarchical Budget Policy Optimization for Adaptive Reasoning
• CoT-Self-Instruct: Building high-quality synthetic prompts for reasoning and non-reasoning tasks
• MECAT: A Multi-Experts Constructed Benchmark for Fine-Grained Audio Understanding Tasks
• CX-Mind: A Pioneering Multimodal Large Language Model for Interleaved Reasoning in Chest X-ray via Curriculum-Guided Reinforcement Learning
• Phi-Ground Tech Report: Advancing Perception in GUI Grounding
• LLMs Can Covertly Sandbag on Capability Evaluations Against Chain-of-Thought Monitoring
• Causal Reasoning in Pieces: Modular In-Context Learning for Causal Discovery
• Stabilizing Knowledge, Promoting Reasoning: Dual-Token Constraints for RLVR
• A Framework for Analyzing Abnormal Emergence in Service Ecosystems Through LLM-based Agent Intention Mining
• Good Learners Think Their Thinking: Generative PRM Makes Large Reasoning Model More Efficient Math Learner
• LAPO: Internalizing Reasoning Efficiency via Length-Adaptive Policy Optimization
• Repair-R1: Better Test Before Repair
• GPT-4.1 Sets the Standard in Automated Experiment Design Using Novel Python Libraries
• RePaCA: Leveraging Reasoning Large Language Models for Static Automated Patch Correctness Assessment
• Why can't Epidemiology be automated (yet)?
• Towards Simulating Social Influence Dynamics with LLM-based Multi-agents
• IFEvalCode: Controlled Code Generation
• A Large Language Model Powered Integrated Circuit Footprint Geometry Understanding
• VL-Cogito: Progressive Curriculum Reinforcement Learning for Advanced Multimodal Reasoning
• FairReason: Balancing Reasoning and Social Bias in MLLMs
• RecGPT Technical Report
• Improving Generative Ad Text on Facebook using Reinforcement Learning
• ArtSeek: Deep artwork understanding via multimodal in-context reasoning and late interaction retrieval
• Graph-R1: Towards Agentic GraphRAG Framework via End-to-end Reinforcement Learning
• Exploring the Link Between Bayesian Inference and Embodied Intelligence: Toward Open Physical-World Embodied AI Systems
• Libra: Assessing and Improving Reward Model by Learning to Think
• MAAD: Automate Software Architecture Design through Knowledge-Driven Multi-Agent Collaboration
• Repairing vulnerabilities without invisible hands. A differentiated replication study on LLMs
• ARC-Hunyuan-Video-7B: Structured Video Comprehension of Real-World Shorts
• MediQAl: A French Medical Question Answering Dataset for Knowledge and Reasoning Evaluation
• Geometric-Mean Policy Optimization
• Enhancing Spatial Reasoning through Visual and Textual Thinking
• QuestA: Expanding Reasoning Capacity in LLMs via Question Augmentation
• RingMo-Agent: A Unified Remote Sensing Foundation Model for Multi-Platform and Multi-Modal Reasoning
• Cultivating Helpful, Personalized, and Creative AI Tutors: A Framework for Pedagogical Alignment using Reinforcement Learning
• ELMES: An Automated Framework for Evaluating Large Language Models in Educational Scenarios
• The Policy Cliff: A Theoretical Analysis of Reward-Policy Maps in Large Language Models
• StepFun-Prover Preview: Let's Think and Verify Step by Step
• Solving Formal Math Problems by Decomposition and Iterative Reflection
• UloRL:An Ultra-Long Output Reinforcement Learning Approach for Advancing Large Language Models' Reasoning Abilities
• JT-Math: A Multi-Stage Framework for Advanced Mathematical Reasoning in Large Language Models
• TurQUaz at CheckThat! 2025: Debating Large Language Models for Scientific Web Discourse Detection
• Understanding Human Limits in Pattern Recognition: A Computational Model of Sequential Reasoning in Rock, Paper, Scissors
• PennyCoder: Efficient Domain-Specific LLMs for PennyLane-Based Quantum Code Generation
• Trustworthy Reasoning: Evaluating and Enhancing Factual Accuracy in LLM Intermediate Thought Processes
• PEMUTA: Pedagogically-Enriched Multi-Granular Undergraduate Thesis Assessment
• Advancing Event Forecasting through Massive Training of Large Language Models: Challenges, Solutions, and Broader Impacts
• Smooth Reading: Bridging the Gap of Recurrent LLM to Self-Attention LLM on Long-Context Tasks
• RemoteReasoner: Towards Unifying Geospatial Reasoning Workflow
• OctoNav: Towards Generalist Embodied Navigation
• A Neuroscience-Inspired Dual-Process Model of Compositional Generalization
• PurpCode: Reasoning for Safer Code Generation
• Efficient Agents: Building Effective Agents While Reducing Cost
• TTS-VAR: A Test-Time Scaling Framework for Visual Auto-Regressive Generation
• Revisiting LLM Reasoning via Information Bottleneck
• Decoupling Knowledge and Reasoning in LLMs: An Exploration Using Cognitive Dual-System Theory
• Datasets and Recipes for Video Temporal Grounding via Reinforcement Learning
• T2VWorldBench: A Benchmark for Evaluating World Knowledge in Text-to-Video Generation
• Squeeze10-LLM: Squeezing LLMs' Weights by 10 Times via a Staged Mixed-Precision Quantization Method
• Does More Inference-Time Compute Really Help Robustness?
• GUI-G²: Gaussian Reward Modeling for GUI Grounding
• ICWLM: A Multi-Task Wireless Large Model via In-Context Learning
• DiagR1: A Vision-Language Model Trained via Reinforcement Learning for Digestive Pathology Diagnosis
• Reinforcement Learning for Flow-Matching Policies
• VAT-KG: Knowledge-Intensive Multimodal Knowledge Graph Dataset for Retrieval-Augmented Generation
• Feedback-Induced Performance Decline in LLM-Based Decision-Making
• AgentFly: Extensible and Scalable Reinforcement Learning for LM Agents
• Omni-Thinker: Scaling Multi-Task RL in LLMs with Hybrid Reward and Task Scheduling
• Breaking the Tokenizer Barrier: On-Policy Distillation across Model Families
• OPRD: On-Policy Representation Distillation
• QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks
• MiroMind-M1: An Open-Source Advancement in Mathematical Reasoning via Context-Aware Multi-Stage Policy Optimization
• Automated Conjecture Resolution with Formal Verification
• BusterX++: Towards Unified Cross-Modal AI-Generated Content Detection and Explanation with MLLM
• PivotRL: High Accuracy Agentic Post-Training at Low Compute Cost
• MagicGUI: A Foundational Mobile GUI Agent with Scalable Data Pipeline and Reinforcement Fine-tuning
• AlgoTune: Can Language Models Speed Up General-Purpose Numerical Programs?
• Actor-Curator: Co-adaptive Curriculum Learning via Policy-Improvement Bandits for RL Post-Training
• Benchmarking Gaslighting Negation Attacks Against Reasoning Models
• It's Not That Simple. An Analysis of Simple Test-Time Scaling
• X-Intelligence 3.0: Training and Evaluating Reasoning LLM for Semiconductor Display
• A Hybrid Framework for Subject Analysis: Integrating Embedding-Based Regression Models with Large Language Models
• Adaptive Multi-Agent Reasoning via Automated Workflow Generation
• Characterizing Communication Patterns in Distributed Large Language Model Inference
• Can LLMs Infer Personality from Real World Conversations?
• KROMA: Ontology Matching with Knowledge Retrieval and Large Language Models
• Cross-modal Causal Intervention for Alzheimer's Disease Prediction
• PubMind: Literature-Based Genetic Variant Extraction and Functional Annotation Using Large Language Models
• MolPIF: A Parameter Interpolation Flow Model for Molecule Generation
• Towards Evaluting Fake Reasoning Bias in Language Models
• Critique of impure reason: Unveiling the reasoning behaviour of medical large language models
• Pretraining on the Test Set Is No Longer All You Need: A Debate-Driven Approach to QA Benchmarks
• URPO: A Unified Reward & Policy Optimization Framework for Large Language Models
• LoopServe: An Adaptive Dual-phase LLM Inference Acceleration System for Multi-Turn Dialogues
• Seed LiveInterpret 2.0: End-to-end Simultaneous Speech-to-speech Translation with Your Voice
• MultiNRC: A Challenging and Native Multilingual Reasoning Evaluation Benchmark for LLMs
• On a few pitfalls in KL divergence gradient estimation for RL
• A Versatile Pathology Co-pilot via Reasoning Enhanced Multimodal Large Language Model
• CLARIFID: Improving Radiology Report Generation by Reinforcing Clinically Accurate Impressions and Enforcing Detailed Findings
• GenSelect: A Generative Approach to Best-of-N
• Agent Identity Evals: Measuring Agentic Identity
• Can One Domain Help Others? A Data-Centric Study on Multi-Domain Reasoning via Reinforcement Learning
• BucketServe: Bucket-Based Dynamic Batching for Smart and Efficient LLM Inference Serving
• Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains
• Reinforcement Learning Fine-Tunes a Sparse Subnetwork in Large Language Models
• Thinking Isn't an Illusion: Overcoming the Limitations of Reasoning Models via Tool Augmentations
• CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning
• MegaScience: Pushing the Frontiers of Post-Training Datasets for Science Reasoning
• When LLMs Copy to Think: Uncovering Copy-Guided Attacks in Reasoning LLMs
• Zebra-CoT: A Dataset for Interleaved Vision Language Reasoning
• The Ever-Evolving Science Exam
• SPARQL Query Generation with LLMs: Measuring the Impact of Training Data Memorization and Knowledge Injection
• INTEGRALBENCH: Benchmarking LLMs with Definite Integral Problems
• Seed-X: Building Strong Multilingual Translation LLM with 7B Parameters
• Perovskite-R1: A Domain-Specialized LLM for Intelligent Discovery of Precursor Additives and Experimental Design
• A Simple "Try Again" Can Elicit Multi-Turn LLM Reasoning
• Evaluating the Effectiveness of Cost-Efficient Large Language Models in Benchmark Biomedical Tasks
• Resa: Transparent Reasoning Models via SAEs
• Inverse Reinforcement Learning Meets Large Language Model Post-Training: Basics, Advances, and Opportunities
• Model-free Reinforcement Learning for Model-based Control: Towards Safe, Interpretable and Sample-efficient Agents
• LaViPlan : Language-Guided Visual Path Planning with RLVR
• AbGen: Evaluating Large Language Models in Ablation Study Design and Evaluation for Scientific Research
• Enhancing Cross-task Transfer of Large Language Models via Activation Steering
• Agentar-DeepFinance-100K: A Large-Scale Financial Dataset via Systematic Chain-of-Thought Synthesis Optimization
• Can We Predict Alignment Before Models Finish Thinking? Towards Monitoring Misaligned Reasoning Models
• Foresight in Motion: Reinforcing Trajectory Prediction with Reward Heuristics
• One Patient, Many Contexts: Scaling Medical AI with Contextual Intelligence
• PRISM: Distributed Inference for Foundation Models at Edge
• ReasonMed: A 370K Multi-Agent Generated Dataset for Advancing Medical Reasoning
• Toxicity-Aware Few-Shot Prompting for Low-Resource Singlish Translation
• Kevin: Multi-Turn RL for Generating CUDA Kernels
• Let's Think in Two Steps: Mitigating Agreement Bias in MLLMs with Self-Grounded Verification
• KisMATH: Do LLMs Have Knowledge of Implicit Structures in Mathematical Reasoning?
• SANGAM: SystemVerilog Assertion Generation via Monte Carlo Tree Self-Refine
• Router-R1: Teaching LLMs Multi-Round Routing and Aggregation via Reinforcement Learning
• VidBridge-R1: Bridging QA and Captioning for RL-based Video Understanding Models with Intermediate Proxy Tasks
• Trustworthy AI for Medicine: Continuous Hallucination Detection and Elimination with CHECK
• Opus: A Prompt Intention Framework for Complex Workflow Generation
• An Agentic Flow for Finite State Machine Extraction using Prompt Chaining
• LogTinyLLM: Tiny Large Language Models Based Contextual Log Anomaly Detection
• SWE-MERA: A Dynamic Benchmark for Agenticly Evaluating Large Language Models on Software Engineering Tasks
• EDINET-Bench: Evaluating LLMs on Complex Financial Tasks using Japanese Financial Statements
• EXAONE 4.0: Unified Large Language Models Integrating Non-reasoning and Reasoning Modes
• Step-wise Policy for Rare-tool Knowledge (SPaRK): Offline RL that Drives Diverse Tool Use in LLMs
• From Alerts to Intelligence: A Novel LLM-Aided Framework for Host-based Intrusion Detection
• UGC-VideoCaptioner: An Omni UGC Video Detail Caption Model and New Benchmarks
• How Many Instructions Can LLMs Follow at Once?
• ExpliCIT-QA: Explainable Code-Based Image Table Question Answering
• DeepForm: Reasoning Large Language Model for Communication System Formulation
• FMC: Formalization of Natural Language Mathematical Competition Problems
• FedRAG: A Framework for Fine-Tuning Retrieval-Augmented Generation Systems
• REST: Stress Testing Large Reasoning Models by Asking Multiple Problems at Once
• CodeJudgeBench: Benchmarking LLM-as-a-Judge for Coding Tasks
• HKGAI-V1: Towards Regional Sovereign Large Language Model for Hong Kong
• Prompt Informed Reinforcement Learning for Visual Coverage Path Planning
• ARMOR: Aligning Secure and Safe Large Language Models via Meticulous Reasoning
• Towards Applying Large Language Models to Complement Single-Cell Foundation Models
• GHPO: Adaptive Guidance for Stable and Efficient LLM Reinforcement Learning
• The Challenge of Teaching Reasoning to LLMs Without RL or Distillation
• How to Provably Improve Return Conditioned Supervised Learning?
• TableDreamer: Progressive and Weakness-guided Data Synthesis from Scratch for Table Instruction Tuning
• AbstentionBench: Reasoning LLMs Fail on Unanswerable Questions
• EmbRACE-3K: Embodied Reasoning and Action in Complex Environments
• DeepSeek: Paradigm Shifts and Technical Evolution in Large AI Models
• Comprehension Without Competence: Architectural Limits of LLMs in Symbolic Computation and Reasoning
• Scalpel vs. Hammer: GRPO Amplifies Existing Capabilities, SFT Replaces Them
• GenAI-based Multi-Agent Reinforcement Learning towards Distributed Agent Intelligence: A Generative-RL Agent Perspective
• SeerAttention-R: Sparse Attention Adaptation for Long Reasoning
• AraReasoner: Evaluating Reasoning-Based LLMs for Arabic NLP
• CompassJudger-2: Towards Generalist Judge Model via Verifiable Rewards
• TACTIC: Translation Agents with Cognitive-Theoretic Interactive Collaboration
• How to Train a Leader: Hierarchical Reasoning in Multi-Agent LLMs
• KV Cache Steering for Controlling Frozen LLMs
• Leanabell-Prover-V2: Verifier-integrated Reasoning for Formal Theorem Proving via Reinforcement Learning
• A comprehensive study of LLM-based argument classification: from LLAMA through GPT-4o to Deepseek-R1
• To Trade or Not to Trade: An Agentic Approach to Estimating Market Risk Improves Trading Decisions
• Large Multi-modal Model Cartographic Map Comprehension for Textual Locality Georeferencing
• FreeAudio: Training-Free Timing Planning for Controllable Long-Form Text-to-Audio Generation
• From Language to Logic: A Bi-Level Framework for Structured Reasoning
• What Factors Affect LLMs and RLLMs in Financial Question Answering?
• M2-Reasoning: Empowering MLLMs with Unified General and Spatial Reasoning
• KAT-V1: Kwai-AutoThink Technical Report
• Agent Safety Alignment via Reinforcement Learning
• Multi-Actor Generative Artificial Intelligence as a Game Engine
• Distilling Empathy from Large Language Models
• Quantile Reward Policy Optimization: Alignment with Pointwise Regression and Exact Partition Functions
• From Domain Documents to Requirements: Retrieval-Augmented Generation in the Space Industry
• DrugMCTS: a drug repurposing framework combining multi-agent, RAG and Monte Carlo Tree Search
• Automating Expert-Level Medical Reasoning Evaluation of Large Language Models
• Scaling RL to Long Videos
• RLEP: Reinforcement Learning with Experience Replay for LLM Reasoning
• DICE: Data Influence Cascade in Decentralized Learning
• FloorplanQA: A Benchmark for Spatial Reasoning in LLMs using Structured Representations
• The Synergy Dilemma of Long-CoT SFT and RL: Investigating Post-Training Techniques for Reasoning VLMs
• Squeeze the Soaked Sponge: Efficient Off-policy Reinforcement Finetuning for Large Language Model
• LOVON: Legged Open-Vocabulary Object Navigator
• First Return, Entropy-Eliciting Explore
• CriticLean: Critic-Guided Reinforcement Learning for Mathematical Formalization
• Evaluation of Habitat Robotics using Large Language Models
• Coding Triangle: How Does Large Language Model Understand Code?
• PrefixAgent: An LLM-Powered Design Framework for Efficient Prefix Adder Optimization
• High-Resolution Visual Reasoning via Multi-Turn Grounding-Based Reinforcement Learning
• AutoTriton: Automatic Triton Programming with Reinforcement Learning in LLMs
• How Not to Detect Prompt Injections with an LLM
• Skywork-R1V3 Technical Report
• A Survey on Latent Reasoning
• Differential Mamba
• PaddleOCR 3.0 Technical Report
• DATABench: Evaluating Dataset Auditing in Deep Learning from an Adversarial Perspective
• OpenAgentSafety: A Comprehensive Framework for Evaluating Real-World AI Agent Safety
• Affective-ROPTester: Capability and Bias Analysis of LLMs in Predicting Retinopathy of Prematurity
• MobileGUI-RL: Advancing Mobile GUI Agent through Reinforcement Learning in Online Environment
• Hierarchical Interaction Summarization and Contrastive Prompting for Explainable Recommendations
• CoRE: Enhancing Metacognition with Label-free Self-evaluation in LRMs
• CogniSQL-R1-Zero: Lightweight Reinforced Reasoning for Efficient SQL Generation
• The Safety Gap Toolkit: Evaluating Hidden Dangers of Open-Source Models
• FEVO: Financial Knowledge Expansion and Reasoning Evolution for Large Language Models
• GTA1: GUI Test-time Scaling Agent
• A Systematization of Security Vulnerabilities in Computer Use Agents
• Supporting Software Formal Verification with Large Language Models: An Experimental Study
• Performance Evaluation of General Purpose Large Language Models for Basic Linear Algebra Subprograms Code Generation
• SciMaster: Towards General-Purpose Scientific AI Agents, Part I. X-Master as Foundation: Can We Lead on Humanity's Last Exam?
• Evaluating Fake Music Detection Performance Under Audio Augmentations
• ArtifactsBench: Bridging the Visual-Interactive Gap in LLM Code Generation Evaluation
• Advancing Financial Engineering with Foundation Models: Progress, Applications, and Challenges
• Learn Globally, Speak Locally: Bridging the Gaps in Multilingual Reasoning
• Unveiling the Potential of Diffusion Large Language Model in Controllable Generation
• CoT-lized Diffusion: Let's Reinforce T2I Generation Step-by-step
• Does Learning Mathematical Problem-Solving Generalize to Broader Reasoning?
• Attention Slipping: A Mechanistic Understanding of Jailbreak Attacks and Defenses in LLMs
• Context Tuning for In-Context Optimization
• HLStrans: Dataset for C-to-HLS Hardware Code Synthesis
• Dissecting Clinical Reasoning in Language Models: A Comparative Study of Prompts and Model Adaptation Strategies
• Ready Jurist One: Benchmarking Language Agents for Legal Intelligence in Dynamic Environments
• A Comparative Study of Specialized LLMs as Dense Retrievers
• Economic Evaluation of LLMs
• Read Quietly, Think Aloud: Decoupling Comprehension and Reasoning in LLMs
• Analysis and Optimized CXL-Attached Memory Allocation for Long-Context LLM Fine-Tuning
• SynapseRoute: An Auto-Route Switching Framework on Dual-State Large Language Model
• Is Reasoning All You Need? Probing Bias in the Age of Reasoning Language Models
• Meta SecAlign: A Secure Foundation LLM Against Prompt Injection Attacks
• VRAgent-R1: Boosting Video Recommendation with MLLM-based Agents via Reinforcement Learning
• Exploring Gender Bias Beyond Occupational Titles
• WebSailor: Navigating Super-human Reasoning for Web Agent
• AI Research Agents for Machine Learning: Search, Exploration, and Generalization in MLE-bench
• Improving LLM Reasoning for Vulnerability Detection via Group Relative Policy Optimization
• Rethinking Data Protection in the (Generative) Artificial Intelligence Era
• ReliableMath: Benchmark of Reliable Mathematical Reasoning on Large Language Models
• CoRe: Benchmarking LLMs Code Reasoning Capabilities through Static Analysis Tasks
• Latent Chain-of-Thought? Decoding the Depth-Recurrent Transformer
• Do Role-Playing Agents Practice What They Preach? Belief-Behavior Consistency in LLM-Based Simulations of Human Trust
• BACTA-GPT: An AI-Based Bayesian Adaptive Clinical Trial Architect
• Energy-Based Transformers are Scalable Learners and Thinkers
• Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs
• Test-Time Scaling with Reflective Generative Model
• Kwai Keye-VL Technical Report
• HCNQA: Enhancing 3D VQA with Hierarchical Concentration Narrowing Supervision
• OpenTable-R1: A Reinforcement Learning Augmented Tool Agent for Open-Domain Table Question Answering
• Emotionally Intelligent Task-oriented Dialogue Systems: Architecture, Representation, and Optimisation
• Agent-as-Tool: A Study on the Hierarchical Decision Making with Reinforcement Learning
• Data Agent: A Holistic Architecture for Orchestrating Data+AI Ecosystems
• A Large Language Model for Chemistry and Retrosynthesis Predictions
• NaturalThoughts: Selecting and Distilling Reasoning Traces for General Reasoning Tasks
• MedGround-R1: Advancing Medical Image Grounding via Spatial-Semantic Rewarded Group Relative Policy Optimization
• Temporal Chain of Thought: Long-Video Understanding by Thinking in Frames
• GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning
• Improving the Reasoning of Multi-Image Grounding in MLLMs via Reinforcement Learning
• Can Large Language Models Develop Strategic Reasoning? Post-training Insights from Learning Chess
• A Hierarchical and Evolvable Benchmark for Fine-Grained Code Instruction Following with Multi-Turn Feedback
• When Iterative RAG Beats Ideal Evidence: A Diagnostic Study in Scientific Multi-hop Question Answering
• Twill: Scheduling Compound AI Systems on Heterogeneous Mobile Edge Platforms
• TGRPO :Fine-tuning Vision-Language-Action Model via Trajectory-wise Group Relative Policy Optimization
• HumanoidGen: Data Generation for Bimanual Dexterous Manipulation via LLM Reasoning
• ASTRO: Teaching Language Models to Reason by Reflecting and Backtracking In-Context
• Leveraging Large Language Models for Spontaneous Speech-Based Suicide Risk Detection
• Ella: Embodied Social Agents with Lifelong Memory
• Reinforcing VLMs to Use Tools for Detailed Visual Reasoning Under Resource Constraints
• Advancing Multi-Step Mathematical Reasoning in Large Language Models through Multi-Layered Self-Reflection with Auto-Prompting
• Email as the Interface to Generative AI Models: Seamless Administrative Automation
• L0: Reinforcement Learning to Become General Agents
• Can Artificial Intelligence Write Like Borges? An Evaluation Protocol for Spanish Microfiction
• EfficientXLang: Towards Improving Token Efficiency Through Cross-Lingual Reasoning
• DABstep: Data Agent Benchmark for Multi-step Reasoning
• Act-With-Think: Chunk Auto-Regressive Modeling for Generative Recommendation
• How large language models judge and influence human cooperation
• Do Thinking Tokens Help or Trap? Towards More Efficient Large Reasoning Model
• Machine Assistant with Reliable Knowledge: Enhancing Student Learning via RAG-based Retrieval
• From Release to Adoption: Challenges in Reusing Pre-trained AI Models for Downstream Developers
• CyberV: Cybernetics for Test-time Scaling in Video Understanding
• Are Large Language Models Capable of Deep Relational Reasoning? Insights from DeepSeek-R1 and Benchmark Comparisons
• Agent-to-Agent Theory of Mind: Testing Interlocutor Awareness among Large Language Models
• Evaluating and Improving Large Language Models for Competitive Program Generation
• Towards Explainable Bilingual Multimodal Misinformation Detection and Localization
• Listener-Rewarded Thinking in VLMs for Image Preferences
• MARBLE: A Hard Benchmark for Multimodal Spatial Reasoning and Planning
• RoboPearls: Editable Video Simulation for Robot Manipulation
• HAIBU-ReMUD: Reasoning Multimodal Ultrasound Dataset and Model Bridging to General Specific Domains
• Improving Large Language Models with Concept-Aware Fine-Tuning
• Refining Czech GEC: Insights from a Multi-Experiment Approach
• OptScale: Probabilistic Optimality for Inference-time Scaling
• EFRame: Deeper Reasoning via Exploration-Filter-Replay Reinforcement Learning Framework
• Lost at the Beginning of Reasoning
• GPAS: Accelerating Convergence of LLM Pretraining via Gradient-Preserving Activation Scaling
• Robust and Efficient Autoregressive Speech Synthesis with Dynamic Chunk-wise Prediction Policy
• A Large Language Model-Empowered Agent for Reliable and Robust Structural Analysis
• A Survey of Continual Reinforcement Learning
• LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs
• R1-Track: Direct Application of MLLMs to Visual Object Tracking via Reinforcement Learning
• RoboEnvision: A Long-Horizon Video Generation Model for Multi-Task Robot Manipulation
• Layer Importance for Mathematical Reasoning is Forged in Pre-Training and Invariant after Post-Training
• A Survey of LLM Inference Systems
• THE-Tree: Can Tracing Historical Evolution Enhance Scientific Verification and Reasoning?
• SEEA-R1: Tree-Structured Reinforcement Fine-Tuning for Self-Evolving Embodied Agents
• APO: Enhancing Reasoning Ability of MLLMs via Asymmetric Policy Optimization
• Weak-to-Strong GraphRAG: Aligning Weak Retrievers with Large Language Models for Graph-based Retrieval Augmented Generation
• Assessing an evolutionary search engine for small language models, prompts, and evaluation metrics
• HumanOmniV2: From Understanding to Omni-Modal Reasoning with Context
• RL-Selector: Reinforcement Learning-Guided Data Selection via Redundancy Assessment
• TAI3: Testing Agent Integrity in Interpreting User Intent
• Explicit Preference Optimization: No Need for an Implicit Reward Model
• Mitigating Hidden Confounding by Progressive Confounder Imputation via Large Language Models
• Bridging Offline and Online Reinforcement Learning for LLMs
• SEED: Enhancing Text-to-SQL Performance and Practical Usability Through Automatic Evidence Generation
• Do We Really Need GNNs with Explicit Structural Modeling? MLPs Suffice for Language Model Representations
• Double-Checker: Enhancing Reasoning of Slow-Thinking LLMs via Self-Critical Fine-Tuning
• MMSearch-R1: Incentivizing LMMs to Search
• π-CoT: Prolog-Initialized Chain-of-Thought Prompting for Multi-Hop Question-Answering
• An Agentic System for Rare Disease Diagnosis with Traceable Reasoning
• ARGUS: Hallucination and Omission Evaluation in Video-LLMs
• Enterprise Large Language Model Evaluation Benchmark
• A Modular Multitask Reasoning Framework Integrating Spatio-temporal Models and LLMs
• AALC: Large Language Model Efficient Reasoning via Adaptive Accuracy-Length Control
• Video Perception Models for 3D Scene Synthesis
• The Decrypto Benchmark for Multi-Agent Reasoning and Theory of Mind
• ReCode: Updating Code API Knowledge with Reinforcement Learning
• Enhancing Large Language Models through Structured Reasoning
• Safety-Aligned Weights Are Not Enough: Refusal-Teacher-Guided Finetuning Enhances Safety and Downstream Performance under Harmful Finetuning Attacks
• OctoThinker: Mid-training Incentivizes Reinforcement Learning Scaling
• Scaling Speculative Decoding with Lookahead Reasoning
• Persona Features Control Emergent Misalignment
• SAGE: Strategy-Adaptive Generation Engine for Query Rewriting
• OptiScene: LLM-driven Indoor Scene Layout Generation via Scaled Human-aligned Data Synthesis and Multi-Stage Preference Optimization
• A Survey of LLM-Driven AI Agent Communication: Protocols, Security Risks, and Defense Countermeasures
• KunLunBaizeRAG: Reinforcement Learning Driven Inference Performance Leap for Large Language Models
• Towards a Small Language Model Lifecycle Framework
• AnTKV: Anchor Token-Aware Sub-Bit Vector Quantization for KV Cache in Large Language Models
• Is Long-to-Short a Free Lunch? Investigating Inconsistency and Reasoning Efficiency in LRMs
• Solving Inequality Proofs with Large Language Models
• Automatic Prompt Optimization for Knowledge Graph Construction: Insights from an Empirical Study
• Prover Agent: An Agent-Based Framework for Formal Mathematical Proofs
• RecLLM-R1: A Two-Stage Training Paradigm with Reinforcement Learning and Chain-of-Thought v1
• Unified Vision-Language-Action Model
• Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion
• Surgery-R1: Advancing Surgical-VQLA with Reasoning Multimodal Large Language Model via Reinforcement Learning
• Skywork-SWE: Unveiling Data Scaling Laws for Software Engineering in LLMs
• SRFT: A Single-Stage Method with Supervised and Reinforcement Fine-Tuning for Reasoning
• Thought Anchors: Which LLM Reasoning Steps Matter?
• OMEGA: Can LLMs Reason Outside the Box in Math? Evaluating Exploratory, Compositional, and Transformative Generalization
• OmniGen2: Exploration to Advanced Multimodal Generation
• MiniCPM4: Ultra-Efficient LLMs on End Devices
• Event-Priori-Based Vision-Language Model for Efficient Visual Understanding
• ConciseHint: Boosting Efficient Reasoning via Continuous Concise Hints during Generation
• Existing LLMs Are Not Self-Consistent For Simple Tasks
• Harnessing the Power of Reinforcement Learning for Language-Model-Based Information Retriever via Query-Document Co-Augmentation
• Deceptive Game Design? Investigating the Impact of Visual Card Style on Player Perception
• MedTVT-R1: A Multimodal LLM Empowering Medical Reasoning and Diagnosis
• AdapThink: Adaptive Thinking Preferences for Reasoning Language Model
• Reasoning about Uncertainty: Do Reasoning Models Know When They Don't Know?
• Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning
• From Web Search towards Agentic Deep Research: Incentivizing Search with Reasoning Agents
• IndexTTS2: A Breakthrough in Emotionally Expressive and Duration-Controlled Auto-Regressive Zero-Shot Text-to-Speech
• Command-V: Pasting LLM Behaviors via Activation Profiles
• WeThink: Toward General-purpose Vision-Language Reasoning via Reinforcement Learning
• Understanding Reasoning in Thinking Language Models via Steering Vectors
• RL for Reasoning by Adaptively Revealing Rationales
• InspireDebate: Multi-Dimensional Subjective-Objective Evaluation-Guided Reasoning and Optimization for Debating
• CLGRPO: Reasoning Ability Enhancement for Small VLMs
• Pre-Trained LLM is a Semantic-Aware and Generalizable Segmentation Booster
• Refine Medical Diagnosis Using Generation Augmented Retrieval and Clinical Practice Guidelines
• Aligning Frozen LLMs by Reinforcement Learning: An Iterative Reweight-then-Optimize Approach
• Graphs Meet AI Agents: Taxonomy, Progress, and Future Opportunities
• Unveiling Factors for Enhanced POS Tagging: A Study of Low-Resource Medieval Romance Languages
• Scene-R1: Video-Grounded Large Language Models for 3D Scene Reasoning without 3D Annotations
• KAG-Thinker: Interactive Thinking and Deep Reasoning in LLMs via Knowledge-Augmented Generation
• When Does Multimodality Lead to Better Time Series Forecasting?
• Fast ECoT: Efficient Embodied Chain-of-Thought via Thoughts Reuse
• Beyond the Link: Assessing LLMs' ability to Classify Political Content across Global Media
• Re-Evaluating Code LLM Benchmarks Under Semantic Mutation
• Self-Balancing of Cell Populations via Martingale Turnover with Amplification
• Chiron-o1: Igniting Multimodal Large Language Models towards Generalizable Medical Reasoning via Mentor-Intern Collaborative Search
• No Free Lunch: Rethinking Internal Feedback for LLM Reasoning
• Evaluating LLMs Effectiveness in Detecting and Correcting Test Smells: An Empirical Study
• HeuriGym: An Agentic Benchmark for LLM-Crafted Heuristics in Combinatorial Optimization
• Mapping the Evolution of Research Contributions using KnoVo
• VLN-R1: Vision-Language Navigation via Reinforcement Fine-Tuning
• Cache Me If You Can: How Many KVs Do You Need for Effective Long-Context LMs?
• From Thinking to Output: Chain-of-Thought and Text Generation Characteristics in Reasoning Language Models
• ML-Master: Towards AI-for-AI via Integration of Exploration and Reasoning
• GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning
• LazyEviction: Lagged KV Eviction with Attention Pattern Observation for Efficient Long Reasoning
• DynScaling: Efficient Verifier-free Inference Scaling via Dynamic and Integrated Sampling
• FLAME: Towards Federated Fine-Tuning Large Language Models Through Adaptive SMoE
• Steering Your Diffusion Policy with Latent Space Reinforcement Learning
• AutoRule: Reasoning Chain-of-thought Extracted Rule-based Rewards Improve Preference Learning
• Exploring and Exploiting the Inherent Efficiency within Large Reasoning Models for Self-Guided Efficiency Enhancement
• ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving
• ProtoReasoning: Prototypes as the Foundation for Generalizable Reasoning in LLMs
• Truncated Proximal Policy Optimization
• HeurAgenix: Leveraging LLMs for Solving Complex Combinatorial Optimization Challenges
• RAS-Eval: A Comprehensive Benchmark for Security Evaluation of LLM Agents in Real-World Environments
• FindingDory: A Benchmark to Evaluate Memory in Embodied Agents
• MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents
• Lessons from Training Grounded LLMs with Verifiable Rewards
• Semantically-Aware Rewards for Open-Ended R1 Training in Free-Form Generation
• CC-LEARN: Cohort-based Consistency Learning
• Fractional Reasoning via Latent Steering Vectors Improves Inference Time Compute
• From LLMs to MLLMs to Agents: A Survey of Emerging Paradigms in Jailbreak Attacks and Defenses within LLM Ecosystem
• Reward Models in Deep Reinforcement Learning: A Survey
• Large Language Models -- the Future of Fundamental Physics?
• Ring-lite: Scalable Reasoning via C3PO-Stabilized Reinforcement Learning for LLMs
• AIRTBench: Measuring Autonomous AI Red Teaming Capabilities in Language Models
• M2BeamLLM: Multimodal Sensing-empowered mmWave Beam Prediction with Large Language Models
• Play to Generalize: Learning to Reason Through Game Play
• ADRD: LLM-Driven Autonomous Driving Based on Rule-based Decision Systems
• Explainable Detection of Implicit Influential Patterns in Conversations via Data Augmentation
• Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMs
• Expectation Confirmation Preference Optimization for Multi-Turn Conversational Recommendation Agent
• Causes in neuron diagrams, and testing causal reasoning in Large Language Models. A glimpse of the future of philosophy?
• Reasoning with Exploration: An Entropy Perspective
• Recognition through Reasoning: Reinforcing Image Geo-localization with Large Vision-Language Models
• From Generation to Generalization: Emergent Few-Shot Learning in Video Diffusion Models
• Xolver: Multi-Agent Reasoning with Holistic Experience Learning Just Like an Olympiad Team
• Overclocking LLM Reasoning: Monitoring and Controlling Thinking Path Lengths in LLMs
• RadFabric: Agentic AI System with Reasoning Capability for Radiology
• Adaptive Guidance Accelerates Reinforcement Learning of Reasoning Models
• We Should Identify and Mitigate Third-Party Safety Risks in MCP-Powered Agent Systems
• Perception-R1: Advancing Multimodal Reasoning Capabilities of MLLMs via Visual Perception Reward
• SeqPE: Transformer with Sequential Position Encoding
• IGD: Token Decisiveness Modeling via Information Gain in LLMs for Personalized Recommendation
• CTDGSI: A comprehensive exploitation of instance selection methods for automatic text classification. VII Concurso de Teses, Dissertações e Trabalhos de Graduação em SI -- XXI Simpósio Brasileiro de Sistemas de Informação
• Rethinking Test-Time Scaling for Medical AI: Model and Task-Aware Strategies for LLMs and VLMs
• AceReason-Nemotron 1.1: Advancing Math and Code Reasoning through SFT and RL Synergy
• Metis-RISE: RL Incentivizes and SFT Enhances Multimodal Reasoning Model Learning
• AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning
• VL-GenRM: Enhancing Vision-Language Verification via Vision Experts and Iterative Training
• LocationReasoner: Evaluating LLMs on Real-World Site Selection Reasoning
• EvolvTrip: Enhancing Literary Character Understanding with Temporal Theory-of-Mind Graphs
• Discrete Diffusion in Large Language and Multimodal Models: A Survey
• Document-Level Tabular Numerical Cross-Checking: A Coarse-to-Fine Approach
• xbench: Tracking Agents Productivity Scaling with Profession-Aligned Real-World Evaluations
• TimeMaster: Training Time-Series Multimodal LLMs to Reason via Reinforcement Learning
• ExtendAttack: Attacking Servers of LRMs via Extending Reasoning
• Position: Pause Recycling LoRAs and Prioritize Mechanisms to Uncover Limits and Effectiveness
• MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention
• FinLMM-R1: Enhancing Financial Reasoning in LMM through Scalable Data and Reward Design
• AdaVideoRAG: Omni-Contextual Adaptive Retrieval-Augmented Efficient Long Video Understanding
• Socratic RL: A Novel Framework for Efficient Knowledge Acquisition through Iterative Reflection and Viewpoint Distillation
• Multipole Attention for Efficient Long Context Reasoning
• Interpretable and Reliable Detection of AI-Generated Images via Grounded Reasoning in MLLMs
• Chain-of-Code Collapse: Reasoning Failures in LLMs via Adversarial Prompting in Code Generation
• STRuCT-LLM: Unifying Tabular and Graph Reasoning with Reinforcement Learning for Semantic Parsing
• Reasoning Model Unlearning: Forgetting Traces, Not Just Answers, While Preserving Reasoning Skills
• Scaling Test-time Compute for LLM Agents
• MCTS-Refined CoT: High-Quality Fine-Tuning Data for LLM-Based Repository Issue Resolution
• Wide-Horizon Thinking and Simulation-Based Evaluation for Real-World LLM Planning with Multifaceted Constraints
• Bridging the Digital Divide: Small Language Models as a Pathway for Physics and Photonics Education in Underdeveloped Regions
• QiMeng-Attention: SOTA Attention Operator is generated by SOTA Attention Algorithm
• SCGAgent: Recreating the Benefits of Reasoning Models for Secure Code Generation with Agentic Workflows
• Efficient Reasoning Through Suppression of Self-Affirmation Reflections in Large Reasoning Models
• Performance Plateaus in Inference-Time Scaling for Text-to-Image Diffusion Without External Models
• Advances in LLMs with Focus on Reasoning, Adaptability, Efficiency and Ethics
• Holistic approach and Advanced Color Singlet Identification for physics measurements at high energy frontier
• HOI-PAGE: Zero-Shot Human-Object Interaction Generation with Part Affordance Guidance
• Bias Amplification in RAG: Poisoning Knowledge Retrieval to Steer LLMs
• TongSearch-QR: Reinforced Query Reasoning for Retrieval
• VGR: Visual Grounded Reasoning
• Schema-R1: A reasoning training approach for schema linking in Text-to-SQL Task
• RAG+: Enhancing Retrieval-Augmented Generation with Application-Aware Reasoning
• How Far Are We from Optimal Reasoning Efficiency?
• Training-free LLM Verification via Recycling Few-shot Examples
• From Emergence to Control: Probing and Modulating Self-Reflection in Language Models
• Towards Universal Offline Black-Box Optimization via Learning Language Model Embeddings
• VFaith: Do Large Multimodal Models Really Reason on Seen Images Rather than Previous Memories?
• QForce-RL: Quantized FPGA-Optimized Reinforcement Learning Compute Engine
• The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity
• LLM Embedding-based Attribution (LEA): Quantifying Source Contributions to Generative Model's Response for Vulnerability Analysis
• What Users Value and Critique: Large-Scale Analysis of User Feedback on AI-Powered Mobile Apps
• KnowCoder-V2: Deep Knowledge Analysis
• Motion-R1: Enhancing Motion Generation with Decomposed Chain-of-Thought and RL Binding
• Formalizing Learning from Language Feedback with Provable Guarantees
• WGSR-Bench: Wargame-based Game-theoretic Strategic Reasoning Benchmark for Large Language Models
• OIBench: Benchmarking Strong Reasoning Models with Olympiad in Informatics
• LogiPlan: A Structured Benchmark for Logical Planning and Relational Reasoning in LLMs
• Specification and Evaluation of Multi-Agent LLM Systems -- Prototype and Cybersecurity Applications
• PAG: Multi-Turn Reinforced LLM Self-Correction with Policy as Generative Verifier
• OPT-BENCH: Evaluating LLM Agent on Large-Scale Search Spaces Optimization Problems
• Discovering Hierarchical Latent Capabilities of Language Models via Causal Representation Learning
• VideoExplorer: Think With Videos For Agentic Long-Video Understanding
• TeleMath: A Benchmark for Large Language Models in Telecom Mathematical Problem Solving
• Vision-EKIPL: External Knowledge-Infused Policy Learning for Visual Reasoning
• Poutine: Vision-Language-Trajectory Pre-Training and Reinforcement Learning Post-Training Enable Robust End-to-End Autonomous Driving
• Causal Sufficiency and Necessity Improves Chain-of-Thought Reasoning
• Revisiting Visual Understanding in Multimodal Reasoning through a Lens of Image Perturbation
• KG-Infused RAG: Augmenting Corpus-Based RAG with External Knowledge Graphs
• Athena: Enhancing Multimodal Reasoning with Data-efficient Process Reward Models
• Understanding and Mitigating Numerical Sources of Nondeterminism in LLM Inference
• DivScore: Zero-Shot Detection of LLM-Generated Text in Specialized Domains
• Boosting LLM Reasoning via Spontaneous Self-Correction
• STARFlow: Scaling Latent Normalizing Flows for High-resolution Image Synthesis
• Table-r1: Self-supervised and Reinforcement Learning for Program-based Table Reasoning in Small Language Models
• Loki's Dance of Illusions: A Comprehensive Survey of Hallucination in Large Language Models
• Route-and-Reason: Scaling Large Language Model Reasoning with Reinforced Model Router
• BioMol-MQA: A Multi-Modal Question Answering Dataset For LLM Reasoning Over Bio-Molecular Interactions
• SPRINT: Enabling Interleaved Planning and Parallelized Execution in Reasoning Models
• Saffron-1: Safety Inference Scaling
• PuzzleWorld: A Benchmark for Multimodal, Open-Ended Reasoning in Puzzlehunts
• Evolutionary Perspectives on the Evaluation of LLM-Based AI Agents: A Comprehensive Survey
• Information Bargaining: Bilateral Commitment in Bayesian Persuasion
• Being Strong Progressively! Enhancing Knowledge Distillation of Large Language Models through a Curriculum Learning Framework
• Cartridges: Lightweight and general-purpose long context representations via self-study
• AV-Reasoner: Improving and Benchmarking Clue-Grounded Audio-Visual Counting for MLLMs
• A Smooth Sea Never Made a Skilled SAILOR: Robust Imitation via Learning to Search
• Resisting Contextual Interference in RAG via Parametric-Knowledge Reinforcement
• PCDVQ: Enhancing Vector Quantization for Large Language Models via Polar Coordinate Decoupling
• Beyond Accuracy: Dissecting Mathematical Reasoning for LLMs Under Reinforcement Learning
• Towards Holistic Visual Quality Assessment of AI-Generated Videos: A LLM-Based Multi-Dimensional Evaluation Model
• Reshaping Reasoning in LLMs: A Theoretical Analysis of RL Training Dynamics through Pattern Selection
• Normative Conflicts and Shallow AI Alignment
• Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning
• TreeRPO: Tree Relative Policy Optimization
• ProRefine: Inference-Time Prompt Refinement with Textual Feedback
• Mixture-of-Experts Meets In-Context Reinforcement Learning
• Reason-to-Recommend: Using Interaction-of-Thought Reasoning to Enhance LLM Recommendation
• Dissecting Logical Reasoning in LLMs: A Fine-Grained Evaluation and Supervision Study
• Reasoning-Aligned Perception Decoupling for Scalable Multi-modal Reasoning
• ScaleRTL: Scaling LLMs with Reasoning Data and Test-Time Compute for Accurate RTL Code Generation
• Joint Evaluation of Answer and Reasoning Consistency for Hallucination Detection in Large Reasoning Models
• When Models Know More Than They Can Explain: Quantifying Knowledge Transfer in Human-AI Collaboration
• Evaluation is All You Need: Strategic Overclaiming of LLM Reasoning Capabilities Through Evaluation Design
• Confidence Is All You Need: Few-Shot RL Fine-Tuning of Language Models
• A Reasoning-Based Approach to Cryptic Crossword Clue Solving
• VideoConviction: A Multimodal Benchmark for Human Conviction and Stock Market Recommendations
• Plugging Schema Graph into Multi-Table QA: A Human-Guided Framework for Reducing LLM Reliance
• Cascadia: An Efficient Cascade Serving System for Large Language Models
• RedRFT: A Light-Weight Benchmark for Reinforcement Fine-Tuning-Based Red Teaming

Discussions

• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL [hn, 1351 points, 1056 comments]
• DeepSeek R1 is an open-source LLM comparable in size to gemini that exceeds o1 on reasoning tasks with lower training costs. It combines reinforcement learning & inference time compute. "DeepSeek-R1: [bsky, 73 points, 2 comments]
• Major reasoning models trained w RL so far with technical reports: 2025-01-22 — DeepSeek R1 — arxiv.org/abs/2501.12948 2025-01-22 — Kimi 1.5 — arxiv.org/abs/2501.12599 2025-03-31 — Open-Reasoner-Zero [bsky, 48 points, 2 comments]
• DeepSeek recently updated their R1 technical report with a bunch of new appendices, including a safety report arxiv.org/abs/2501.12948 [bsky, 18 points, 2 comments]
• Warum 200 Euro pro Monat für ChatGPT Pro ausgeben, wenn man bei DeepSeek nichts dafür bezahlt. Wer sich für die technischen Hintergründe bzw. Vergleiche interessiert, wird hier fündig: arxiv.org/pdf/2 [bsky, 16 points, 1 comments]
• DeepSeek-AI showcases a reinforcement learning method that boosts logical reasoning in language models without massive supervised data. By guiding chain-of-thought and distilling results to smaller mo [bsky, 13 points, 0 comments]
• the deepseek paper is surprisingly readable and, at times, kind of cute? arxiv.org/pdf/2501.12948 [bsky, 11 points, 1 comments]
• People claim anyone can now repro DeepSeek R1 and by proxy OpenAI's o1-3. Has anyone actually read the paper? Do people not understand, that there's infinitely more to training such a model than 2.5 [bsky, 10 points, 3 comments]
• I'm going throught he paper like this: paper on left, Claude on right (with PDF in context). Paper's here: arxiv.org/abs/2501.12948 [bsky, 8 points, 1 comments]
• Skepticism is healthy and I encourage you to follow your curuosity. Here's the paper for free. A friend made a very astute remark. The media is making it a West Vs China issue but this is actually a [bsky, 7 points, 1 comments]
• The paper behind the Deepseek approach: DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org/abs/2501.12948 [bsky, 6 points, 0 comments]
• DeepSeek R1 paperet er vild læsning! De er fuldstændig åbne omkring hvordan de har trænet R1 og især det de beskriver, som “aha moment”, hvor modellen begynder på selvrefleksion er virkelig fascineren [bsky, 4 points, 0 comments]
• this is based on a quip @jakehofman.bsky.social made but it's no joke about the arxiv paper arxiv.org/abs/2501.12948 [bsky, 4 points, 0 comments]
• I read the DeepSeek R1 paper and so should you. They used RL and set the model loose on verifiable coding tasks, among other things. arxiv.org/abs/2501.12948 [bsky, 4 points, 0 comments]
• They have published a paper on how they do it arxiv.org/abs/2501.12948 [bsky, 4 points, 1 comments]
• It’s missing one pretty important thing though - the Jan 2025 inflection point correlates strongly with arxiv.org/abs/2501.12948 which demonstrated that efficient gpu optimization combined with aggres [bsky, 3 points, 0 comments]
• For this week's @milanlp.bsky.social reading group, Emanuele Moscato presented "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning" by @deepseek.bsky.social Paper: arx [bsky, 3 points, 0 comments]
• Lots of panic as well as excitement about DeepSeek, the new open source LLM from China Seems to me a very clever, if not ingenious, approach, and one that undercuts the idea that we need bigger model [bsky, 3 points, 0 comments]
• Here's the paper that tanked the market #Followers #neuroskyence #DeepSeek #AI arxiv.org/pdf/2501.12948 [bsky, 3 points, 0 comments]
• The R1 paper arxiv.org/abs/2501.12948 does not have dollar values, but they have pretty specific numbers for the amount of data and training they did. An expert should be able to get a very reasonable [bsky, 2 points, 0 comments]
• yep! the main paper from deepseek itself is here: arxiv.org/pdf/2501.12948 and it's open source, so there's plenty of explainers as a result: www.analyticsvidhya.com/blog/2025/01... [bsky, 2 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model train [bsky, 2 points, 0 comments]
• "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL" There's a new smart model that's better than older ones, making others seem less important. This shift is making big tech nervous abou [bsky, 2 points, 0 comments]
• @viajandoconciencia.bsky.social @pcoffeebreak.bsky.social @emulenews.bsky.social Creo que esto de DeepSeek merece una mención especial. Hablan de rendimiento similar con mucho menos consumo energét [bsky, 2 points, 1 comments]
• I still don’t have enough concrete evidence to explain why reinforcement learning in GenAI enables the emergence of advanced reasoning patterns, such as mathematical reasoning. It might simply be a ma [bsky, 2 points, 1 comments]
• Kiina on vuosia tuottanut paljon tekoäly tutkimusta, joten voinee ajatella, että kaikki palvelevat Kiinan strategiaa. Tämä applikaatio mm. Kuten sanoit. Sinänsä, että mallin arkkitehtuuri on avoin ta [bsky, 2 points, 0 comments]
• “DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning”: arxiv.org/abs/2501.12948 [bsky, 2 points, 1 comments]
• DeepSeek-R1 seems to have learnt to "revisit and reevaluate its previous steps" on its own but that part of the training process might not generalize as well beyond things like math where the results [bsky, 1 points, 0 comments]
• 🧪 the accessibility and simplicity of the concepts in this paper that lead to such a large improvement represents a significant moment in the world of post-training reasoning models with R1 performin [bsky, 1 points, 1 comments]
• Maybe some reason. arxiv.org/abs/2501.12948 [bsky, 1 points, 0 comments]
• It lies largely outside my areas of expertise, but it seems to involve quite a bit more than that if one may believe DeepSeek's recent paper arxiv.org/abs/2501.12948 DeepSeek-R1: Incentivizing reasoni [bsky, 1 points, 0 comments]
• Totalmente Daniel y lo que más se llama la atención sobre todo es como han sido capaces de entrenarlos en menos tiempo y de forma mucho más barata que OpenAI. El paper no tiene desperdicio arxiv.org/p [bsky, 1 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL https://arxiv.org/abs/2501.12948 (https://news.ycombinator.com/item?id=42823568) [bsky, 1 points, 0 comments]
• Deepseek's ability is on par to slightly better than the other big boy LLMs (and tbh one should put decreasing stock in these benchmark values as they become targets for the models)--it isn't a GPT3-t [bsky, 1 points, 1 comments]
• einfach andere methodik soweit ich das verstehe arxiv.org/abs/2501.12948 [bsky, 1 points, 0 comments]
• arxiv.org/abs/2501.12948 [bsky, 1 points, 0 comments]
• This paper explains the process well: arxiv.org/abs/2501.12948 [bsky, 1 points, 1 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL (arxiv.org) Main Link | Discussion [bsky, 1 points, 0 comments]
• DeepSeek-R1 article #DeepSeek #ReinforcementLearning #LLM #Reasoning arxiv.org/abs/2501.12948 [bsky, 1 points, 0 comments]
• I didn't quite understand the first part of this sentence? By the way, coincidentally the target of your criticism saw an extensive update just a few days ago, after almost a year: arxiv.org/abs/2501. [bsky, 1 points, 3 comments]
• This isn't defending Chinese AI strategy. Their policies maximize profits for annotation companies. They don't care about working conditions. DeepSeek, after open-sourcing its model, won't even disclo [bsky, 1 points, 1 comments]
• Here is the DeepSeek-R1 paper explaining the use of reinforcement learning for training large reasoning models. The model is also open sourced for further research and development. #AI #RL #LLM #reinf [bsky, 1 points, 0 comments]
• #InIAMode Pues parece que los modelos grandes del lenguaje pueden entrenar (con datos sintéticos) a los modelos pequeños del lenguaje. Y, la gracia, así entrenados, los peques salen "más razonadores" [bsky, 1 points, 0 comments]
• We're following the terminology of the DeepSeek-R1 paper that introduced this model: arxiv.org/abs/2501.12948 Whether it's really the best metaphor is certainly worth asking! I can see pros and cons f [bsky, 1 points, 1 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL view on hacker news [bsky, 1 points, 0 comments]
• No data is sent to the servers. OK, the most powerful model can't be employed locally, but it can still do something. And progress only goes forward. Then maybe there's no big need AI factories? What [bsky, 1 points, 1 comments]
• DeepSeekの論文（ arxiv.org/abs/2501.12948 ）おもろすぎる。R1は「メタ思考」を獲得してるらしい。 回答途中で自分の誤りに気づいて（赤字の部分）、「ちょっと待て、ここで何かおかしいぞ！」って言ったあとに回答を修正し始めてる。自分の回答を客観的に見直してるっていう。（1枚目） Aha momentと名づけられていた。 ChatGPTに「お前はこれできる？」って聞 [bsky, 1 points, 0 comments]
• Queréis papers? Tengo papers [bsky, 1 points, 0 comments]
• arxiv.org/pdf/2501.12948 [bsky, 1 points, 0 comments]
• Dieses neuatige Reinforcment Learning der Reasoning-LLM wurde jüngst mit dem Open-Weight-Modell DeepSeek R1 aus 🇨🇳 offengelegt und man kann davon ausgehen, dass das Reinforcement Learning während de [bsky, 0 points, 1 comments]
• In natural language reasoning, we don't have such clarity. The R1 paper explicitly alludes to this problem, having (unsuccessfully) tried MCTS. MCTS did improve performance, but the search space just [bsky, 0 points, 1 comments]
• The DeepSeek-R1 paper (arxiv.org/abs/2501.12948) is one of the landmark papers in #AI and one reason why is that "Open"AI never published a #Technical report on o1. [bsky, 0 points, 0 comments]
• Excited to see the potential of DeepSeek! Promising results in language model development. Check out their official paper for a deep dive into the research #DeepLearning #AI #DeepSeek #LanguageModel # [bsky, 0 points, 0 comments]
• 🤯 DeepSeek-R1 Chinese breakthrough in LLM arxiv.org/pdf/2501.12948 [bsky, 0 points, 0 comments]
• My bad not to add: arxiv.org/pdf/2501.12948 [bsky, 0 points, 0 comments]
• I don't think they've put out the technical report for r1 but there is a paper and it talks about how many steps, training examples they use: arxiv.org/pdf/2501.12948 [bsky, 0 points, 0 comments]
• Deep Seek - arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL https://arxiv.org/abs/2501.12948 (https://news.ycombinator.com/item?id=42823568) [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL https://arxiv.org/abs/2501.12948 (https://news.ycombinator.com/item?id=42823568) [bsky, 0 points, 0 comments]
• arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• This is the report. If you read it, keep in mind this was a small team, which did its work in a very short period of time. arxiv.org/pdf/2501.12948 [bsky, 0 points, 1 comments]
• 赞： arxiv.org/pdf/2501.12948 [bsky, 0 points, 0 comments]
• arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning The 22-page paper, released last week by a scrappy Chinese A.I. start-up called DeepSeek arxiv.org/pdf/2501.12948 [bsky, 0 points, 0 comments]
• The outputs of reasoning models are growing rapidly – between 2-13x/year at the current pace, although this varies with the specific benchmark. This isn't too surprising. DeepSeek's R1 paper already s [bsky, 0 points, 1 comments]
• Leitura sugerida para o fim de semana... "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning" arxiv.org/pdf/2501.12948 [bsky, 0 points, 0 comments]
• arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL #HackerNews arxiv.org/abs/... [bsky, 0 points, 0 comments]
• arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL https://arxiv.org/abs/2501.12948 https://news.ycombinator.com/item?id=42823568 [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL https://arxiv.org/abs/2501.12948 arxiv.org [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL https://arxiv.org/abs/2501.12948 [comments] [188 points] [bsky, 0 points, 0 comments]
• arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• This may be the first paper in the arxiv.org depository that has significant and economic consequences. [bsky, 0 points, 1 comments]
• arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• That trillion dollar open source paper... arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• 🌐DeepSeek-R1: RL による LLM の推論能力の促進 https://arxiv.org/abs/2501.12948 via #HackerNews [bsky, 0 points, 0 comments]
• Le dernier papier des développeurs de DeepSeek "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning" arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• The technique is pretty standard now. DeepSeek or DeepMind use it too. But it effectively manages to nudge models towards longer and more sophisticated chains of reasoning (illustration from the DeepS [bsky, 0 points, 1 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL https://arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. See arxiv.org/abs/2501.12948 [bsky, 0 points, 0 comments]
• https://bsky.app/profile/news.ycombinator.com.web.brid.gy/post/3lgnrwzjg7342 [bsky, 0 points, 0 comments]
• Looking at the recent deep seek paper, definitely a good use of RL, something similar is for sure being done by the other big labs behind the scenes tho I suspect deepseek distilled it down more effec [bsky, 0 points, 0 comments]
• in the top 3 of the modern AI papers arxiv.org/pdf/2501.12948 [bsky, 0 points, 0 comments]
• Pity that deepseek discussions are more often about censorship/challenges to US. Less focus on the improvement of a rational basis to output that they've added. Who knows, this could be the right time [bsky, 0 points, 0 comments]
• Odd response. arxiv.org/abs/2501.12948 Which of these folks do you think are really stupid? And can you quote the part of the paper that you think is dumb? [bsky, 0 points, 0 comments]
• "A particularly intriguing phenomenon observed during the training of DeepSeek-R1-Zero is the occurrence of an “aha moment”. DeepSeek-AI et al., DeepSeek-R1: Incentivizing Reasoning Capability in LLM [bsky, 0 points, 1 comments]
• This basically describes certain forms of current reinforcement learning. I dont neccsarily think it is a good test, but effectively certain LLM are already are behaving in that way. youtu.be/fFL7la73 [bsky, 0 points, 0 comments]
• The question remains whether Europe can learn from DeepSeek R1 and emerge from the shadows, with political strategies creating an environment that supports this advancement yet to be seen. arxiv.org/ [bsky, 0 points, 0 comments]
• Incentivizing Reasoning Capability in LLMs via RL [DeepSeek-AI, 2025] R1-Zero acquired o1-level reasoning capability by directly applying GRPO, which rewards correct answers, to the base model. The te [bsky, 0 points, 0 comments]
• Check out DeepSeek's papers on arXiv to get a deeper understanding of how they are approaching model creation and training. arxiv.org/abs/2501.12948 #DeepSeek #AI #innovation [bsky, 0 points, 0 comments]
• #arxiv paper on #Deepseek [bsky, 0 points, 0 comments]

Related

• A Survey of Reinforcement Learning for Large Reasoning Models
• Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?
• RL of Thoughts: Navigating LLM Reasoning with Inference-time Reinforcement Learning
• General-Reasoner: Advancing LLM Reasoning Across All Domains
• Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMs