π_0.5: a Vision-Language-Action Model with Open-World Generalization

2025/04/22 by Physical Intelligence, Kevin Black, Intelligence, Physical +70 · 1 voice · 1,508 citations
Computer Science · Social Sciences · #African history and culture analysis #Geographic Information Systems Studies #Religion and Sociopolitical Dynamics in Nigeria #cs.LG #cs.RO

paper · pdf · doi:10.48550/arxiv.2504.16054

Abstract

In order for robots to be useful, they must perform practically relevant tasks in the real world, outside of the lab. While vision-language-action (VLA) models have demonstrated impressive results for end-to-end robot control, it remains an open question how far such models can generalize in the wild. We describe π_0.5, a new model based on π₀ that uses co-training on heterogeneous tasks to enable broad generalization. π_0.5 uses data from multiple robots, high-level semantic prediction, web data, and other sources to enable broadly generalizable real-world robotic manipulation. Our system uses a combination of co-training and hybrid multi-modal examples that combine image observations, language commands, object detections, semantic subtask prediction, and low-level actions. Our experiments show that this kind of knowledge transfer is essential for effective generalization, and we demonstrate for the first time that an end-to-end learning-enabled robotic system can perform long-horizon and dexterous manipulation skills, such as cleaning a kitchen or bedroom, in entirely new homes.

Citations

• GR00T N1: An Open Foundation Model for Generalist Humanoid Robots
• HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Model
• AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems
• OTTER: A Vision-Language-Action Model with Text-Aware Visual Feature Extraction
• Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Models
• Magma: A Foundation Model for Multimodal AI Agents
• DexVLA: Vision-Language Model with Plug-In Diffusion Expert for General Robot Control
• HAMSTER: Hierarchical Action Models For Open-World Robot Manipulation
• FAST: Efficient Action Tokenization for Vision-Language-Action Models
• From Multimodal LLMs to Generalist Embodied Agents: Methods and Lessons
• NaVILA: Legged Robot Vision-Language-Action Model for Navigation
• CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation
• π₀: A Vision-Language-Action Flow Model for General Robot Control
• Data Scaling Laws in Imitation Learning for Robotic Manipulation
• RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation
• BUMBLE: Unifying Reasoning and Acting with Vision-Language Models for Building-wide Mobile Manipulation
• RACER: Rich Language-Guided Failure Recovery Policies for Imitation Learning
• TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulation
• Robot Utility Models: General Policies for Zero-Shot Deployment in New Environments
• Automating Robot Failure Recovery Using Vision-Language Models With Optimized Prompts
• Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation
• Robotic Control via Embodied Chain-of-Thought Reasoning
• PaliGemma: A versatile 3B VLM for transfer
• LLaRA: Supercharging Robot Learning Data for Vision-Language Policy
• Manipulate-Anything: Automating Real-World Robots using Vision-Language Models
• Open-vocabulary Mobile Manipulation in Unseen Dynamic Environments with 3D Semantic Maps
• Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs
• LLARVA: Vision-Action Instruction Tuning Enhances Robot Learning
• OpenVLA: An Open-Source Vision-Language-Action Model
• Octo: An Open-Source Generalist Robot Policy
• Closed-Loop Open-Vocabulary Mobile Manipulation with GPT-4V
• Yell At Your Robot: Improving On-the-Fly from Language Corrections
• LLM3:Large Language Model-based Task and Motion Planning with Motion Failure Reasoning
• 3D-VLA: A 3D Vision-Language-Action Generative World Model
• Scaling Rectified Flow Transformers for High-Resolution Image Synthesis
• Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robots
• PIVOT: Iterative Visual Prompting Elicits Actionable Knowledge for VLMs
• SPOC: Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World
• On Bringing Robots Home
• CapsFusion: Rethinking Image-Text Data at Scale
• An Unbiased Look at Datasets for Visuo-Motor Pre-Training
• RoboAgent: Generalization and Efficiency in Robot Manipulation via Semantic Augmentations and Action Chunking
• BridgeData V2: A Dataset for Robot Learning at Scale
• RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot
• ViNT: A Foundation Model for Visual Navigation
• Micrograph segmentations for DDEVD
• Segment Anything
• Where are we in the search for an Artificial Visual Cortex for Embodied Intelligence?
• PaLM-E: An Embodied Multimodal Language Model
• Open-World Object Manipulation using Pre-trained Vision-Language Models
• LLaMA: Open and Efficient Foundation Language Models
• AnyGrasp: Robust and Efficient Grasp Perception in Spatial and Temporal Domains
• RT-1: Robotics Transformer for Real-World Control at Scale
• Navigating to Objects in the Real World
• GNM: A General Navigation Model to Drive Any Robot
• Flow Matching for Generative Modeling
• Rectified Flow: A Marginal Preserving Approach to Optimal Transport
• ProgPrompt: Generating Situated Robot Task Plans using Large Language Models
• Code as Policies: Language Model Programs for Embodied Control
• Do As I Can, Not As I Say: Grounding Language in Robotic Affordances
• R3M: A Universal Visual Representation for Robot Manipulation
• Masked Visual Pre-training for Motor Control
• BNAI, NO-TOKEN, and MIND-UNITY: Pillars of a Systemic Revolution in Artificial Intelligence
• Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents
• Masked Autoencoders Are Scalable Vision Learners
• Bridge Data: Boosting Generalization of Robotic Skills with Cross-Domain Datasets
• Learning Transferable Visual Models From Natural Language Supervision
• Language Models are Few-Shot Learners
• RoboNet: Large-Scale Multi-Robot Learning
• Robot Learning in Homes: Improving Generalization and Reducing Dataset Bias
• Attention Is All You Need
• Dex-Net 2.0: Deep Learning to Plan Robust Grasps with Synthetic Point Clouds and Analytic Grasp Metrics
• Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering
• Deep Residual Learning for Image Recognition
• Microsoft COCO Captions: Data Collection and Evaluation Server
• Autonomously Learning to Visually Detect Where Manipulation Will Succeed
• Gemini Robotics: Bringing AI into the Physical World
• Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models
• OpenAI o1 System Card
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Cited by

• ArmnetBench v0.1: Parallel Real-World Evaluation of Manipulation Policies on a Low-Cost Arm Farm
• τ: Learning Touch-Augmented Vision-Language-Action Models from Future Visual Supervision
• Understanding the Impact of Geometric Foundation Models on Vision-Language-Action Models
• Embodied Robot Manipulation in the Era of Foundation Models: Planning and Learning Perspectives
• Robo-Dopamine: General Process Reward Modeling for High-Precision Robotic Manipulation
• WorldDiT: A Unified Diffusion Architecture for World and Action Modeling
• FutureRTC: Real-Time Robot Execution with Anticipatory-Conditioned Action Chunking
• VLA-Arena: An Open-Source Framework for Benchmarking Vision-Language-Action Models
• Clutter-Robust Vision-Language-Action Models through Object-Centric and Geometry Grounding
• LeapBot-WA: World-Anchor Action Models via Predictive Latent Alignments
• A Pragmatic VLA Foundation Model
• Compressing Observation History into Agent Memory: Distilling Transformers into Recurrent Transformers
• Being-H0.7: A Latent World-Action Model from Egocentric Videos
• A Few Words Go a Long Way: Language Guided Robot Policy Synthesis
• N₀-TWAM: Scaling Tactile-Native World-Action Model for Contact-Rich Manipulation
• N₀-VTLA: Scaling Vision-Tactile-Language-Action Model with Latent Tactile Tokens
• LabRobFail: A Benchmark for Robotic Failure Analysis in Chemical Self-driving Laboratory
• WCM: World-Cognition Model for Generalizable Human-Robot Interaction
• Strategy-Aware Parameter-Efficient Adaptation for LLM-based Auto-Bidding
• DeVA: Decoupled Video-Action Model with physical guidance for robot policy learning
• Unified Embodied VLM Reasoning with Robotic Action via Autoregressive Discretized Pre-training
• CAST: Game Solvers as Turn-Level Teachers for LLM Agents
• CoTinyVLA: Chain-of-Thought Distillation for a Sub-Billion-Parameter Vision-Language-Action Model
• A Causality-aware Infer-diagnose-refine Framework for Test-time Modality Adaptation in VLA Models
• RoboMME-Interference: Benchmarking Robot Memory Under Interference
• StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision
• RoboCade: Gamifying Robot Data Collection
• EVE: A Generator-Verifier System for Generative Policies
• Asynchronous Fast-Slow Vision-Language-Action Policies for Whole-Body Robotic Manipulation
• Bring My Cup! Personalizing Vision-Language-Action Models with Visual Attentive Prompting
• DeliveryBench: Can Agents Earn Profit in Real World?
• Real2Edit2Real: Generating Robotic Demonstrations via a 3D Control Interface
• Point What You Mean: Visually Grounded Instruction Policy
• LLaViDA: A Large Language Vision Driving Assistant for Explicit Reasoning and Enhanced Trajectory Planning
• AOMGen: Photoreal, Physics-Consistent Demonstration Generation for Articulated Object Manipulation
• Learning Semantic Atomic Skills for Multi-Task Robotic Manipulation
• Robotic VLA Benefits from Joint Learning with Motion Image Diffusion
• Unifying Deep Predicate Invention with Pre-trained Foundation Models
• Vidarc: Embodied Video Diffusion Model for Closed-loop Control
• PhysBrain: Human Egocentric Data as a Bridge from Vision Language Models to Physical Intelligence
• Monte Carlo Query Search: Active Capability Assessment of AI Agents
• mimic-video: Video-Action Models for Generalizable Robot Control Beyond VLAs
• MiVLA: Towards Generalizable Vision-Language-Action Model with Human-Robot Mutual Imitation Pre-training
• VLA-AN: An Efficient and Onboard Vision-Language-Action Framework for Aerial Navigation in Complex Environments
• Focus: A Streaming Concentration Architecture for Efficient Vision-Language Models
• Sample-Efficient Robot Skill Learning for Construction Tasks: Benchmarking Hierarchical Reinforcement Learning and Vision-Language-Action VLA Model
• OXE-AugE: A Large-Scale Robot Augmentation of OXE for Scaling Cross-Embodiment Policy Learning
• Spatial-Aware VLA Pretraining through Visual-Physical Alignment from Human Videos
• Motus: A Unified Latent Action World Model
• TS-DP: Reinforcement Speculative Decoding For Temporal Adaptive Diffusion Policy Acceleration
• Seeing to Act, Prompting to Specify: A Bayesian Factorization of Vision Language Action Policy
• RoboNeuron: A Middle-Layer Infrastructure for Agent-Driven Orchestration in Embodied AI
• HiF-VLA: Hindsight, Insight and Foresight through Motion Representation for Vision-Language-Action Models
• Token Expand-Merge: Training-Free Token Compression for Vision-Language-Action Models
• GLaD: Geometric Latent Distillation for Vision-Language-Action Models
• H2R-Grounder: A Paired-Data-Free Paradigm for Translating Human Interaction Videos into Physically Grounded Robot Videos
• VLSA: Vision-Language-Action Models with Plug-and-Play Safety Constraint Layer
• Mind to Hand: Purposeful Robotic Control via Embodied Reasoning
• See Once, Then Act: Vision-Language-Action Model with Task Learning from One-Shot Video Demonstrations
• Affordance Field Intervention: Enabling VLAs to Escape Memory Traps in Robotic Manipulation
• Task adaptation of Vision-Language-Action model: 1st Place Solution for the 2025 BEHAVIOR Challenge
• Training-Time Action Conditioning for Efficient Real-Time Chunking
• SIMPACT: Simulation-Enabled Action Planning using Vision-Language Models
• STARE-VLA: Progressive Stage-Aware Reinforcement for Fine-Tuning Vision-Language-Action Models
• FASTer: Toward Efficient Autoregressive Vision Language Action Modeling via Neural Action Tokenization
• SIMA 2: A Generalist Embodied Agent for Virtual Worlds
• dVLM-AD: Enhance Diffusion Vision-Language-Model for Driving via Controllable Reasoning
• Vision-Language-Action Models for Selective Robotic Disassembly: A Case Study on Critical Component Extraction from Desktops
• FALCON: Actively Decoupled Visuomotor Policies for Loco-Manipulation with Foundation-Model-Based Coordination
• ResponsibleRobotBench: Benchmarking Responsible Robot Manipulation using Multi-modal Large Language Models
• SpaceTools: Tool-Augmented Spatial Reasoning via Double Interactive RL
• PosA-VLA: Enhancing Action Generation via Pose-Conditioned Anchor Attention
• Video2Act: A Dual-System Video Diffusion Policy with Robotic Spatio-Motional Modeling
• VLA Models Are More Generalizable Than You Think: Revisiting Physical and Spatial Modeling
• Diagnose, Correct, and Learn from Manipulation Failures via Visual Symbols
• HumanoidExo: Scalable Whole-Body Humanoid Manipulation via Wearable Exoskeleton
• Much Ado About Noising: Dispelling the Myths of Generative Robotic Control
• GR-RL: Going Dexterous and Precise for Long-Horizon Robotic Manipulation
• DiG-Flow: Discrepancy-Guided Flow Matching for Robust VLA Models
• TabletopGen: Tabletop Scene Generation and Interactive Simulation for Robotic Manipulation
• Real-World Reinforcement Learning of Active Perception Behaviors
• IGen: Scalable Data Generation for Robot Learning from Open-World Images
• VLASH: Real-Time VLAs via Future-State-Aware Asynchronous Inference
• CycleManip: Enabling Cyclic Task Manipulation via Effective Historical Perception and Understanding
• Transforming Monolithic Foundation Models into Embodied Multi-Agent Architectures for Human-Robot Collaboration
• Sigma: The Key for Vision-Language-Action Models toward Telepathic Alignment
• LatBot: Distilling Universal Latent Actions for Vision-Language-Action Models
• Mechanistic Finetuning of Vision-Language-Action Models via Few-Shot Demonstrations
• Beyond Success: Refining Elegant Robot Manipulation from Mixed-Quality Data via Just-in-Time Intervention
• DIPOLE: Fusing Vision and Geometry for Robust Visuomotor Generalization
• VacuumVLA: Boosting VLA Capabilities via a Unified Suction and Gripping Tool for Complex Robotic Manipulation
• E₀: Enhancing Generalization and Fine-Grained Control in VLA Models via Continuized Discrete Diffusion
• From Observation to Action: Latent Action-based Primitive Segmentation for VLA Pre-training in Industrial Settings
• SocialNav: Training Human-Inspired Foundation Model for Socially-Aware Embodied Navigation
• When Robots Obey the Patch: Universal Transferable Patch Attacks on Vision-Language-Action Models
• Reinforcing Action Policies by Prophesying
• ArtiBench and ArtiBrain: Benchmarking Generalizable Vision-Language Articulated Object Manipulation
• GigaWorld-0: World Models as Data Engine to Empower Embodied AI
• Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning
• MAPS: Preserving Vision-Language Representations via Module-Wise Proximity Scheduling for Better Vision-Language-Action Generalization
• Unifying Perception and Action: A Hybrid-Modality Pipeline with Implicit Visual Chain-of-Thought for Robotic Action Generation
• Robot-Powered Data Flywheels: Deploying Robots in the Wild for Continual Data Collection and Foundation Model Adaptation
• SENTINEL: A Fully End-to-End Language-Action Model for Humanoid Whole Body Control
• Discover, Learn, and Reinforce: Scaling Vision-Language-Action Pretraining with Diverse RL-Generated Trajectories
• MergeVLA: Cross-Skill Model Merging Toward a Generalist Vision-Language-Action Agent
• SpeedAug: Policy Acceleration via Tempo-Enriched Policy and RL Fine-Tuning
• How to Train Your Latent Control Barrier Function: Smooth Safety Filtering Under Hard-to-Model Constraints
• EchoVLA: Synergistic Declarative Memory for VLA-Driven Mobile Manipulation
• ActDistill: General Action-Guided Self-Derived Distillation for Efficient Vision-Language-Action Models
• L1 Sample Flow for Efficient Visuomotor Learning
• Contact-Rich Robotic Assembly in Construction via Diffusion Policy Learning
• RynnVLA-002: A Unified Vision-Language-Action and World Model
• Unify Robot Actions in Camera Frame
• SPEAR-1: Scaling Beyond Robot Demonstrations via 3D Understanding
• InternData-A1: Pioneering High-Fidelity Synthetic Data for Pre-training Generalist Policy
• Mantis: A Versatile Vision-Language-Action Model with Disentangled Visual Foresight
• In-N-On: Scaling Egocentric Manipulation with in-the-wild and on-task Data
• π^*_0.6: a VLA That Learns From Experience
• AsyncVLA: Asynchronous Flow Matching for Vision-Language-Action Models
• RoboTidy : A 3D Gaussian Splatting Household Tidying Benchmark for Embodied Navigation and Action
• From Power to Precision: Learning Fine-grained Dexterity for Multi-fingered Robotic Hands
• ActiveGrasp: Information-Guided Active Grasping with Calibrated Energy-based Model
• Decoupled Action Expert: Confining Task Knowledge to the Conditioning Pathway
• SemanticVLA: Semantic-Aligned Sparsification and Enhancement for Efficient Robotic Manipulation
• RoboBenchMart: Benchmarking Robots in Retail Environment
• RGMP: Recurrent Geometric-prior Multimodal Policy for Generalizable Humanoid Robot Manipulation
• SpatialThinker: Reinforcing Scene Graph-Grounded Spatial Reasoning via Dense Rewards
• How Do VLAs Effectively Inherit from VLMs?
• ExpReS-VLA: Specializing Vision-Language-Action Models Through Experience Replay and Retrieval
• EveryDayVLA: A Vision-Language-Action Model for Affordable Robotic Manipulation
• TwinVLA: Data-Efficient Bimanual Manipulation with Twin Single-Arm Vision-Language-Action Models
• iFlyBot-VLM Technical Report
• Real-to-Sim Robot Policy Evaluation with Gaussian Splatting Simulation of Soft-Body Interactions
• TWIST2: Scalable, Portable, and Holistic Humanoid Data Collection System
• XR-1: Towards Versatile Vision-Language-Action Models via Learning Unified Vision-Motion Representations
• Learning Interactive World Model for Object-Centric Reinforcement Learning
• GenDexHand: Generative Simulation for Dexterous Hands
• Maestro: Orchestrating Robotics Modules with Vision-Language Models for Zero-Shot Generalist Robots
• iFlyBot-VLA Technical Report
• EgoMI: Learning Active Vision and Whole-Body Manipulation from Egocentric Human Demonstrations
• Learning Generalizable Visuomotor Policy through Dynamics-Alignment
• A Step Toward World Models: A Survey on Robotic Manipulation
• EBT-Policy: Energy Unlocks Emergent Physical Reasoning Capabilities
• DeepThinkVLA: Enhancing Reasoning Capability of Vision-Language-Action Models
• Running VLAs at Real-time Speed
• RoboOS-NeXT: A Unified Memory-based Framework for Lifelong, Scalable, and Robust Multi-Robot Collaboration
• Self-Improving Vision-Language-Action Models with Data Generation via Residual RL
• Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail
• Pelican-VL 1.0: A Foundation Brain Model for Embodied Intelligence
• π_RL: Online RL Fine-tuning for Flow-based Vision-Language-Action Models
• Multimodal Spatial Reasoning in the Large Model Era: A Survey and Benchmarks
• Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization
• When Does Legacy Data Start to Help? Emergent Transfer in Cross-Configuration Robot Learning
• HiFi-UMI: Learning Deployable Manipulation Policies from High-Fidelity UMI Data Alone
• Explicit Kinematic Guidance from Analytic Concepts for Vision-Language-Action Models
• Vision-TL-Action: Neuro-Symbolic Trajectory Generation from Visual Observations and Temporal Logic
• Enfold: Folding World Model Imagination into Predictive Representations for Ultra-Efficient Embodied Control
• StatePlay: State-Aware Game World Models for Mechanics-Consistent Generation
• ContactFlow: A video action conditioning that transfers across embodiments
• CheckVLA: Execution-Time Verification with Action-Conditioned World Model for Long-Horizon Mobile Manipulation
• Practice Makes Policies: Bootstrapping and Consolidating Robotic Capabilities from Zero Human Demonstrations
• VEGA: Learning Navigation VLAs from In-the-Wild Egocentric Video with Geometric Trajectory Supervision
• PACE: Phase-Aware Chunk Execution for Robot Policies with Action Chunking
• InDex: Empowering VLA Models with Intent-Conditioned Arm-Hand Coordination for Dexterous Manipulation
• Can Vision-Language-Action Models Learn from Real-World Data Continually without Forgetting?
• Nautilus: From One Prompt to Plug-and-Play Robot Learning
• TiPToP: A Modular Open-Vocabulary Robot Manipulation System That Plans
• CycleVLA: Proactive Self-Correcting Vision-Language-Action Models via Subtask Backtracking and Minimum Bayes Risk Decoding
• What Questions Should Robots Be Able to Answer? A Dataset of User Questions for Explainable Robotics
• NanoVLA: Routing Decoupled Vision-Language Understanding for Nano-sized Generalist Robotic Policies
• Do What You Say: Steering Vision-Language-Action Models via Runtime Reasoning-Action Alignment Verification
• BLM₁: A Boundless Large Model for Cross-Space, Cross-Task, and Cross-Embodiment Learning
• Endowing GPT-4 with a Humanoid Body: Building the Bridge Between Off-the-Shelf VLMs and the Physical World
• RoboOmni: Proactive Robot Manipulation in Omni-modal Context
• Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agents
• Dexbotic: Open-Source Vision-Language-Action Toolbox
• Sample By Step, Optimize By Chunk: Chunk-Level GRPO For Text-to-Image Generation
• Generalizable Hierarchical Skill Learning via Object-Centric Representation
• Scalable Vision-Language-Action Model Pretraining for Robotic Manipulation with Real-Life Human Activity Videos
• MemER: Scaling Up Memory for Robot Control via Experience Retrieval
• SutureBot: A Precision Framework & Benchmark For Autonomous End-to-End Suturing
• Learning Affordances at Inference-Time for Vision-Language-Action Models
• GigaBrain-0: A World Model-Powered Vision-Language-Action Model
• Seeing Across Views: Benchmarking Spatial Reasoning of Vision-Language Models in Robotic Scenes
• A Compositional Paradigm for Foundation Models: Towards Smarter Robotic Agents
• MoMaGen: Generating Demonstrations under Soft and Hard Constraints for Multi-Step Bimanual Mobile Manipulation
• RESample: A Robust Data Augmentation Framework via Exploratory Sampling for Robotic Manipulation
• From Spatial to Actions: Grounding Vision-Language-Action Model in Spatial Foundation Priors
• RoboChallenge: Large-scale Real-robot Evaluation of Embodied Policies
• Robobench: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models as Embodied Brain
• RDD: Retrieval-Based Demonstration Decomposer for Planner Alignment in Long-Horizon Tasks
• QDepth-VLA: Quantized Depth Prediction as Auxiliary Supervision for Vision-Language-Action Models
• RL-100: Performant Robotic Manipulation with Real-World Reinforcement Learning
• Expertise need not monopolize: Action-Specialized Mixture of Experts for Vision-Language-Action Learning
• RoboHiMan: A Hierarchical Evaluation Paradigm for Compositional Generalization in Long-Horizon Manipulation
• InternVLA-M1: A Spatially Guided Vision-Language-Action Framework for Generalist Robot Policy
• VLA-0: Building State-of-the-Art VLAs with Zero Modification
• Learning to Grasp Anything by Playing with Random Toys
• HiMaCon: Discovering Hierarchical Manipulation Concepts from Unlabeled Multi-Modal Data
• Vlaser: Vision-Language-Action Model with Synergistic Embodied Reasoning
• A Survey on Agentic Multimodal Large Language Models
• X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model
• Ctrl-World: A Controllable Generative World Model for Robot Manipulation
• SAFE: Multitask Failure Detection for Vision-Language-Action Models
• How Well do Diffusion Policies Learn Kinematic Constraint Manifolds?
• PhysToolBench: Benchmarking Physical Tool Understanding for MLLMs
• Humanoid Everyday: A Comprehensive Robotic Dataset for Open-World Humanoid Manipulation
• USIM and U0: A Vision-Language-Action Dataset and Model for General Underwater Robots
• DM1: MeanFlow with Dispersive Regularization for 1-Step Robotic Manipulation
• IntentionVLA: Generalizable and Efficient Embodied Intention Reasoning for Human-Robot Interaction
• DEAS: DEtached value learning with Action Sequence for Scalable Offline RL
• BLAZER: Bootstrapping LLM-based Manipulation Agents with Zero-Shot Data Generation
• Don't Run with Scissors: Pruning Breaks VLA Models but They Can Be Recovered
• From Data to Rewards: a Bilevel Optimization Perspective on Maximum Likelihood Estimation
• ELMUR: External Layer Memory with Update/Rewrite for Long-Horizon RL Problems
• TrackVLA++: Unleashing Reasoning and Memory Capabilities in VLA Models for Embodied Visual Tracking
• Vision-Language-Action Models for Robotics: A Review Towards Real-World Applications
• RLinf-VLA: A Unified and Efficient Framework for Reinforcement Learning of Vision-Language-Action Models
• VLA-R1: Enhancing Reasoning in Vision-Language-Action Models
• INSIGHT: INference-time Sequence Introspection for Generating Help Triggers in Vision-Language-Action Models
• MetaVLA: Unified Meta Co-training For Efficient Embodied Adaption
• ActiveUMI: Robotic Manipulation with Active Perception from Robot-Free Human Demonstrations
• Contrastive Representation Regularization for Vision-Language-Action Models
• LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorization
• GAE: Unleashing Physical Potential of VLM with Generalizable Action Expert
• Hybrid Training for Vision-Language-Action Models
• MLA: A Multisensory Language-Action Model for Multimodal Understanding and Forecasting in Robotic Manipulation
• OmniNav: A Unified Framework for Prospective Exploration and Visual-Language Navigation
• dVLA: Diffusion Vision-Language-Action Model with Multimodal Chain-of-Thought
• Data-Efficient Multitask DAgger
• Fidelity-Aware Data Composition for Robust Robot Generalization
• Unifying Agent Interaction and World Information for Multi-agent Coordination
• Training Agents Inside of Scalable World Models
• Generalizable Coarse-to-Fine Robot Manipulation via Language-Aligned 3D Keypoints
• Transferring Vision-Language-Action Models to Industry Applications: Architectures, Performance, and Challenges
• VLA-Reasoner: Empowering Vision-Language-Action Models with Reasoning via Online Monte Carlo Tree Search
• HELIOS: Hierarchical Exploration for Language-grounded Interaction in Open Scenes
• From Watch to Imagine: Steering Long-horizon Manipulation via Human Demonstration and Future Envisionment
• MimicDreamer: Aligning Human and Robot Demonstrations for Scalable VLA Training
• Actions as Language: Fine-Tuning VLMs into VLAs Without Catastrophic Forgetting
• MoWM: Mixture-of-World-Models for Embodied Planning via Latent-to-Pixel Feature Modulation
• EMMA: Generalizing Real-World Robot Manipulation via Generative Visual Transfer
• AnywhereVLA: Language-Conditioned Exploration and Mobile Manipulation
• Normalizing Flows are Capable Models for Bi-manual Visuomotor Policy
• EgoBridge: Domain Adaptation for Generalizable Imitation from Egocentric Human Data
• Agentic Scene Policies: Unifying Space, Semantics, and Affordances for Robot Action
• TacWAM: Anchor-Guided World Action Model with Mechanics-Aware Tactile Prediction
• RoboBRIDGE: A Modular Framework for Bridging Policies to Robust Real-World Robotic Agents
• World Action Planner: Generalizable Decision-Making with Action-Conditioned World Models
• Cross-Embodiment Transfer via Behavior-Aligned Representations
• RL²-VLA: Adaptive RL Latent Compositional Steering with Test-Time Scaling for Vision-Language-Action Models
• A scalar per patch from pre-trained ViTs enables fast moving navigation in the real world
• SeedPolicy: Horizon Scaling via Self-Evolving Diffusion Policy for Robot Manipulation
• System 0/1/2/3: Quad-Process Theory for Multitimescale Embodied Collective Cognitive Systems
• MolmoB0T: Large-Scale Simulation Enables Zero-Shot Manipulation
• ROPA: Synthetic Robot Pose Generation for RGB-D Bimanual Data Augmentation
• Pure Vision Language Action (VLA) Models: A Comprehensive Survey
• Eva-VLA: Evaluating Vision-Language-Action Models' Robustness Under Real-World Physical Variations
• N2M: Bridging Navigation and Manipulation by Learning Pose Preference from Rollout
• Growing with Your Embodied Agent: A Human-in-the-Loop Lifelong Code Generation Framework for Long-Horizon Manipulation Skills
• VGGT-DP: Generalizable Robot Control via Vision Foundation Models
• Residual Off-Policy RL for Finetuning Behavior Cloning Policies
• Latent Action Pretraining Through World Modeling
• Ratatouille: Imitation Learning Ingredients for Real-world Social Robot Navigation
• IDfRA: Self-Verification for Iterative Design in Robotic Assembly
• FILIC: Dual-Loop Force-Guided Imitation Learning with Impedance Torque Control for Contact-Rich Manipulation Tasks
• TranTac: Leveraging Transient Tactile Signals for Contact-Rich Robotic Manipulation
• A Vision-Language-Action-Critic Model for Robotic Real-World Reinforcement Learning
• Self-Improving Embodied Foundation Models
• Ask-to-Clarify: Resolving Instruction Ambiguity through Multi-turn Dialogue
• Embodied Arena: A Comprehensive, Unified, and Evolving Evaluation Platform for Embodied AI
• RealMirror: A Comprehensive, Open-Source Vision-Language-Action Platform for Embodied AI
• VLA-LPAF: Lightweight Perspective-Adaptive Fusion for Vision-Language-Action to Enable More Unconstrained Robotic Manipulation
• Dual-Actor Fine-Tuning of VLA Models: A Talk-and-Tweak Human-in-the-Loop Approach
• GeoAware-VLA: Implicit Geometry Aware Vision-Language-Action Model
• Robust Online Residual Refinement via Koopman-Guided Dynamics Modeling
• The Better You Learn, The Smarter You Prune: Towards Efficient Vision-language-action Models via Differentiable Token Pruning
• Embodied Navigation Foundation Model
• Deceptive Risk Minimization: Out-of-Distribution Generalization by Deceiving Distribution Shift Detectors
• EgoMem: Lifelong Memory Agent for Full-duplex Omnimodal Models
• Igniting VLMs toward the Embodied Space
• Enhancing Generalization in Vision-Language-Action Models by Preserving Pretrained Representations
• SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning
• VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Model
• NinA: Normalizing Flows in Action. Training VLA Models with Normalizing Flows
• PegasusFlow: Parallel Rolling-Denoising Score Sampling for Robot Diffusion Planner Flow Matching
• RoboMatch: A Unified Mobile-Manipulation Teleoperation Platform with Auto-Matching Network Architecture for Long-Horizon Tasks
• RoboChemist: Long-Horizon and Safety-Compliant Robotic Chemical Experimentation
• TA-VLA: Elucidating the Design Space of Torque-aware Vision-Language-Action Models
• EMMA: Scaling Mobile Manipulation via Egocentric Human Data
• Align-Then-stEer: Adapting the Vision-Language Action Models through Unified Latent Guidance
• Galaxea Open-World Dataset and G0 Dual-System VLA Model
• TReF-6: Inferring Task-Relevant Frames from a Single Demonstration for One-Shot Skill Generalization
• Prompt-to-Product: Generative Assembly via Bimanual Manipulation
• CogVLA: Cognition-Aligned Vision-Language-Action Model via Instruction-Driven Routing & Sparsification
• Learning Primitive Embodied World Models: Towards Scalable Robotic Learning
• Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies
• Survey of Vision-Language-Action Models for Embodied Manipulation
• Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Survey
• Human Centric General Physical Intelligence for Agile Manufacturing Automation
• KDPE: A Kernel Density Estimation Strategy for Diffusion Policy Trajectory Selection
• Physical Autoregressive Model for Robotic Manipulation without Action Pretraining
• Shortcut Learning in Generalist Robot Policies: The Role of Dataset Diversity and Fragmentation
• Information-Theoretic Graph Fusion with Vision-Language-Action Model for Policy Reasoning and Dual Robotic Control
• H-RDT: Human Manipulation Enhanced Bimanual Robotic Manipulation
• Being-H0: Vision-Language-Action Pretraining from Large-Scale Human Videos
• Exploring the Link Between Bayesian Inference and Embodied Intelligence: Toward Open Physical-World Embodied AI Systems
• GR-3 Technical Report
• VITA: Vision-to-Action Flow Matching Policy
• RobotValues: Evaluating Household Robots When Human Values Conflict
• Hierarchical Planning with Latent World Models
• Confidence Calibration in Vision-Language-Action Models
• VLA-Touch: Enhancing Vision-Language-Action Models with Dual-Level Tactile Feedback
• Towards Human-level Intelligence via Human-like Whole-Body Manipulation
• InstructVLA: Vision-Language-Action Instruction Tuning from Understanding to Manipulation
• From Intention to Execution: Probing the Generalization Boundaries of Vision-Language-Action Models
• Vidar: Embodied Video Diffusion Model for Generalist Manipulation
• EgoVLA: Learning Vision-Language-Action Models from Egocentric Human Videos
• V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning
• Vision Language Action Models in Robotic Manipulation: A Systematic Review
• FreqPolicy: Efficient Flow-based Visuomotor Policy via Frequency Consistency
• ULC: A Unified and Fine-Grained Controller for Humanoid Loco-Manipulation
• ST-WAM: Semantic-Temporal World Action Model for Robust Manipulation under Visual Distribution Shifts
• FBFM: A Training-Free Asynchronous Feedback Mechanism for Flow-Matching in World-Action Models Execution
• CLIFT: Turning Gemini Robotics On-Device into Humanoid Specialists via Non-Invasive Closed-Loop Iterative Fine-Tuning
• A Careful Examination of Large Behavior Models for Multitask Dexterous Manipulation
• DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge
• Dynamic Execution Commitment of Vision-Language-Action Models
• Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs
• TriVLA: A Triple-System-Based Unified Vision-Language-Action Model with Episodic World Modeling for General Robot Control
• RoboEval: Where Robotic Manipulation Meets Structured and Scalable Evaluation
• Safe and Performant Deployment of Autonomous Systems via Model Predictive Control and Hamilton-Jacobi Reachability Analysis
• Hierarchical Vision-Language Planning for Multi-Step Humanoid Manipulation
• Embodied AI Agents: Modeling the World
• Parallels Between VLA Model Post-Training and Human Motor Learning: Progress, Challenges, and Trends
• HRIBench: Benchmarking Vision-Language Models for Real-Time Human Perception in Human-Robot Interaction
• DemoDiffusion: One-Shot Human Imitation using pre-trained Diffusion Policy
• Thinking vs. Doing: Agents that Reason by Scaling Test-Time Interaction
• BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models
• CronusVLA: Towards Efficient and Robust Manipulation via Multi-Frame Vision-Language-Action Modeling
• Real-Time Execution of Action Chunking Flow Policies
• RoboArena: Distributed Real-World Evaluation of Generalist Robot Policies
• RLRC: Reinforcement Learning-based Recovery for Compressed Vision-Language-Action Models
• General-Purpose Robotic Navigation via LVLM-Orchestrated Perception, Reasoning, and Acting
• TensorTouch: Calibration of Tactile Sensors for High Resolution Stress Tensor and Deformation for Dexterous Manipulation
• Human2LocoMan: Learning Versatile Quadrupedal Manipulation with Human Pretraining
• LeVERB: Humanoid Whole-Body Control with Latent Vision-Language Instruction
• SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Acceleration
• AntiGrounding: Lifting Robotic Actions into VLM Representation Space for Decision Making
• NeSyPack: A Neuro-Symbolic Framework for Bimanual Logistics Packing
• Benchmarking Vision, Language, & Action Models in Procedurally Generated, Open Ended Action Environments
• A Smooth Sea Never Made a Skilled SAILOR: Robust Imitation via Learning to Search
• Mixture-of-Experts Meets In-Context Reinforcement Learning
• Vision-Language-Action (VLA) Models: Concepts, Progress, Applications and Challenges
• LoHoVLA: A Unified Vision-Language-Action Model for Long-Horizon Embodied Tasks
• GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Data
• TrackVLA: Embodied Visual Tracking in the Wild
• Knowledge Insulating Vision-Language-Action Models: Train Fast, Run Fast, Generalize Better
• ChatVLA-2: Vision-Language-Action Model with Open-World Embodied Reasoning from Pretrained Knowledge
• 3DLLM-Mem: Long-Term Spatial-Temporal Memory for Embodied 3D Large Language Model
• ForceVLA: Enhancing VLA Models with a Force-aware MoE for Contact-rich Manipulation
• ACTIVE-o3: Empowering MLLMs with Active Perception via Pure Reinforcement Learning
• Embodied AI with Foundation Models for Mobile Service Robots: A Systematic Review
• What Can RL Bring to VLA Generalization? An Empirical Study
• SelfWAM: A Self-Grounded Unified World Action Model for Fast Robot Control
• EgoZero: Robot Learning from Smart Glasses
• WorldEval: World Model as Real-World Robot Policies Evaluator
• Uncovering and Mitigating Positional Blind Spots in Vision-Language-Action Models
• What Do You Need for Compositional Generalization in Diffusion Planning?
• ChainVLA: Chaining Vision-Language-Action Queries through a Unified Execution State for Long-Horizon Manipulation
• ReTouch: Empowering Contact-Rich Dexterous Manipulation with Online-Refined Tactile Prediction
• ScanBot: Towards Intelligent Surface Scanning in Embodied Robotic Systems
• Faster-WAM: Do World Action Models Need Deep Action Modules?
• 3D Equivariant Visuomotor Policy Learning via Spherical Projection
• CoMo: Learning Continuous Latent Motion from Internet Videos for Scalable Robot Learning
• DriveMoE: Mixture-of-Experts for Vision-Language-Action Model in End-to-End Autonomous Driving
• Interactive Post-Training for Vision-Language-Action Models
• Learning Panorama-Aware VLA for Mobile Manipulation with Whole-Body Teleoperation
• From Grounding to Manipulation: Case Studies of Foundation Model Integration in Embodied Robotic Systems
• Hermite Curves as Trajectory Priors for Vision-Language-Action Models
• Flattening Hierarchies with Policy Bootstrapping
• Toward Embodied AGI: A Review of Embodied AI and the Road Ahead
• EndoWAM: A Grounded World-Action Model for Generalizable Endoscopic Navigation
• AutoBio: A Simulation and Benchmark for Robotic Automation in Digital Biology Laboratory
• DreamGen: Unlocking Generalization in Robot Learning through Video World Models
• SG-WAM: Self-Guided World Modeling in Geometry-Aware Policy Space
• Demystifying When and Why VLAs Fail in Contact-Rich Tasks and How to Fix Them
• DreamTrajectory: Trajectory-Guided Action Generation with World Model Alignment for Mobile Manipulation
• OneTwoVLA: A Unified Vision-Language-Action Model with Adaptive Reasoning
• DynamicWAM: Dual-Path Motion Conditioning for World-Action Models in Dynamic Manipulation
• FabriVLA: A Lightweight Vision-Language-Action Model with Conformal Action Chunk Uncertainty
• Towards General Language-Conditioned Latent Safety Filters
• Action Chunk Scheduling for Batched Robot Policy Serving
• Unfettered Forceful Skill Acquisition with Physical Reasoning and Coordinate Frame Labeling
• Spline Policy: A Structured Representation for Robot Policies
• Freeform Preference Learning for Robotic Manipulation
• Pixel Motion as Universal Representation for Robot Control
• X-Sim: Cross-Embodiment Learning via Real-to-Sim-to-Real
• Towards Embodiment Scaling Laws in Robot Locomotion
• Long-Term Memory for VLA-based Agents in Open-World Task Execution
• Multi-agent Embodied AI: Advances and Future Directions
• Interleave-VLA: Enhancing Robot Manipulation with Interleaved Image-Text Instructions
• UniDriveVLA: Unifying Understanding, Perception, and Action Planning for Autonomous Driving
• InSight: Self-Guided Skill Acquisition via Steerable VLAs
• Visual Verification Enables Inference-time Steering and Autonomous Policy Improvement
• Geometric Action Model for Robot Policy Learning
• DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation
• vla-eval: A Unified Evaluation Harness for Vision-Language-Action Models
• Xiaomi-Robotics-U0: Unified Embodied Synthesis with World Foundation Model
• B-spline Policy: Accelerating Manipulation Policies via B-spline Action Representations
• DexVerse: A Modular Benchmark for Multi-Task, Multi-Embodiment Dexterous Manipulation
• RynnWorld-Teleop: An Action-Conditioned World Model for Digital Teleoperation
• From Fixed to Free Cameras: Calibration-Free View-Robust Vision-Language-Action Model
• AnoleVLA: Lightweight Vision-Language-Action Model with Deep State Space Models for Mobile Manipulation
• ReMem-VLA: Empowering Vision-Language-Action Model with Memory via Dual-Level Recurrent Queries
• ENPIRE: Agentic Robot Policy Self-Improvement in the Real World
• Qwen-RobotManip Technical Report: Alignment Unlocks Scale for Robotic Manipulation Foundation Models
• μ₀: A Scalable 3D Interaction-Trace World Model
• AR-VLA: True Autoregressive Action Expert for Vision-Language-Action Models
• MAIN-VLA: Modeling Abstraction of Intention and eNvironment for Vision-Language-Action Models
• RoboCasa365: A Large-Scale Simulation Framework for Training and Benchmarking Generalist Robots
• π-StepNFT: Wider Space Needs Finer Steps in Online RL for Flow-based VLAs
• Unleashing the Potential of Diffusion Models for End-to-End Autonomous Driving
• VLANeXt: Recipes for Building Strong VLA Models
• Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots
• VTAM: Video-Tactile-Action Models for Complex Physical Interaction Beyond VLAs
• World Action Models are Zero-shot Policies
• WoVR: World Models as Reliable Simulators for Post-Training VLA Policies with RL
• ActQuant: Sub-4-bit Action-Guided Quantization for Vision-Language-Action Models
• PhysBrain 1.0 Technical Report
• Long-Horizon Q-Learning: Accurate Value Learning via n-Step Inequalities
• OmniTacTune: Policy-Agnostic Real-World RL for Tactile Residual Adaptation of Visual Policies
• VLK: Learning Humanoid Loco-Manipulation from Synthetic Interactions in Reconstructed Scenes
• Learning Action Priors for Cross-embodiment Robot Manipulation
• WEAVER, Better, Faster, Longer: An Effective World Model for Robotic Manipulation
• Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models
• VLA Foundry: A Unified Framework for Training Vision-Language-Action Models
• Veo-Act: How Far Can Frontier Video Models Advance Generalizable Robot Manipulation?
• Contact-Anchored Policies: Contact Conditioning Creates Strong Robot Utility Models
• Unmasking the Illusion of Embodied Reasoning in Vision-Language-Action Models
• StarVLA-α: Reducing Complexity in Vision-Language-Action Systems
• χ₀: Resource-Aware Robust Manipulation via Taming Distributional Inconsistencies
• Fauna Sprout: A lightweight, approachable, developer-ready humanoid robot
• DRIFT: Derailing Denoising Trajectories of Flow-Matching VLAs with Adversarial Patch Attack
• How Should Vision-Language-Action Models Use Proprioceptive State?
• VLM4VLA: Revisiting Vision-Language-Models in Vision-Language-Action Models
• InternVLA-A1: Unifying Understanding, Generation and Action for Robotic Manipulation
• Continue or Replan? Bernoulli-Continuation Policy Learning for Adaptive Horizon Execution
• Unified Visuomotor Targets: Supervising VLAs Beyond Physical Actions
• ValueFormer: A Causal Transformer Value Function with Stage-Aware Labels for Semi-Autonomous Vision-Language-Action Policies
• CofactVLA: Deconfounding Vision-Language-Action Models via Counterfactual Intervention
• Faster-WAM: Efficient Inference-Time Future Conditioning for Robust World Action Models
• MobileWAM: Bridging World Action Models to Mobile Manipulation with Chain-of-Foresight
• Suppression Sticks, Locality Is Fragile: A Closed-Loop Target-and-Control Audit of Task-Vector Negation in VLA Policies
• Explicit Language Memory for Long-Horizon Planning in Vision-Language-Action Models
• DreamWAM: Beyond RGB Future Prediction for World Action Models
• BridgeVLA++: A Data-Efficient, Generalizable, and Memory-Augmented Vision-Language-Action Framework for 3D Manipulation
• SiMDex: Mining Similar Egocentric Videos for Cross-Embodiment Dexterous Manipulation
• SAFECAST: Robust Failure Detection for VLA Policies with Contrast-Set Training and Calibration
• Text-to-Decision Agent: Offline Meta-Reinforcement Learning from Natural Language Supervision
• In-Context VLA: Endowing Vision-Language-Action Models with Language via In-Context Post-Training and Agentic Tool Use
• SkillMemo: Expert-guided Skill Memory Framework for Compositional Embodied Manipulation
• World-to-Wrist: Task-Conditioned Future Wrist Modeling for Fine-Grained Robot Manipulation
• DFM-VLA: Iterative Action Refinement for Robot Manipulation via Discrete Flow Matching
• MMaDA-VLA: Large Diffusion Vision-Language-Action Model with Unified Multi-Modal Instruction and Generation

Discussions

• これだけ学習してもまだ足りないものはなんなんだろうな arxiv.org/abs/2504.16054 [bsky, 0 points, 0 comments]

Related

• π₀: A Vision-Language-Action Flow Model for General Robot Control
• RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
• ObjectVLA: End-to-End Open-World Object Manipulation Without Demonstration
• CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation
• DexVLA: Vision-Language Model with Plug-In Diffusion Expert for General Robot Control