Towards Agentic RAG with Deep Reasoning: A Survey of RAG-Reasoning Systems in LLMs

2025/07/13 by Li, Yangning, Zhang, Weizhi, Yang, Yuyao +17 · 16 citations
#Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2507.09477

Abstract

Retrieval-Augmented Generation (RAG) lifts the factuality of Large Language Models (LLMs) by injecting external knowledge, yet it falls short on problems that demand multi-step inference; conversely, purely reasoning-oriented approaches often hallucinate or mis-ground facts. This survey synthesizes both strands under a unified reasoning-retrieval perspective. We first map how advanced reasoning optimizes each stage of RAG (Reasoning-Enhanced RAG). Then, we show how retrieved knowledge of different type supply missing premises and expand context for complex inference (RAG-Enhanced Reasoning). Finally, we spotlight emerging Synergized RAG-Reasoning frameworks, where (agentic) LLMs iteratively interleave search and reasoning to achieve state-of-the-art performance across knowledge-intensive benchmarks. We categorize methods, datasets, and open challenges, and outline research avenues toward deeper RAG-Reasoning systems that are more effective, multimodally-adaptive, trustworthy, and human-centric. The collection is available at https://github.com/DavidZWZ/Awesome-RAG-Reasoning.

Citations

• From Web Search towards Agentic Deep Research: Incentivizing Search with Reasoning Agents
• ZeroSearch: Incentivize the Search Capability of LLMs without Searching
• LLM-Independent Adaptive RAG: Let the Question Speak for Itself
• SymbioticRAG: Enhancing Document Intelligence Through Human-LLM Symbiotic Collaboration
• WebThinker: Empowering Large Reasoning Models with Deep Research Capability
• BrowseComp-ZH: Benchmarking Web Browsing Ability of Large Language Models in Chinese
• Generative to Agentic AI: Survey, Conceptualization, and Challenges
• DualRAG: A Dual-Process Approach to Integrate Reasoning and Retrieval for Multi-Hop Question Answering
• Credible Plan-Driven RAG Method for Multi-Hop Question Answering
• BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agents
• ARise: Towards Knowledge-Augmented Reasoning via Risk-Adaptive Search
• HM-RAG: Hierarchical Multi-Agent Multimodal Retrieval Augmented Generation
• Collab-RAG: Boosting Retrieval-Augmented Generation for Complex Question Answering via White-Box and Black-Box LLM Collaboration
• ReaRAG: Knowledge-guided Reasoning Enhances Factuality of Large Reasoning Models with Iterative Retrieval Augmented Generation
• MCTS-RAG: Enhancing Retrieval-Augmented Generation with Monte Carlo Tree Search
• Open Deep Search: Democratizing Search with Open-source Reasoning Agents
• Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models
• SWEET-RL: Training Multi-Turn LLM Agents on Collaborative Reasoning Tasks
• MDocAgent: A Multi-Modal Multi-Agent Framework for Document Understanding
• RAG-RL: Advancing Retrieval-Augmented Generation via RL and Curriculum Learning
• Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey
• RAG-KG-IL: A Multi-Agent Hybrid Framework for Reducing Hallucinations and Enhancing LLM Reasoning through RAG and Incremental Knowledge Graph Learning Integration
• SurgRAW: Multi-Agent Workflow with Chain-of-Thought Reasoning for Surgical Intelligence
• Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning
• OpenRAG: Optimizing RAG End-to-End via In-Context Retrieval Learning
• LightThinker: Thinking Step-by-Step Compression
• TokenSkip: Controllable Chain-of-Thought Compression in LLMs
• Self-Training Large Language Models for Tool-Use Without Demonstrations
• ReARTeR: Retrieval-Augmented Reasoning with Trustworthy Process Rewarding
• WebWalker: Benchmarking LLMs in Web Traversal
• Search-o1: Agentic Search-Enhanced Large Reasoning Models
• Agent Laboratory: Using LLM Agents as Research Assistants
• Tree-based RAG-Agent Recommendation System: A Case Study in Medical Test Data
• Cold-Start Recommendation towards the Era of Large Language Models (LLMs): A Comprehensive Survey and Roadmap
• LongDocURL: a Comprehensive Multimodal Long Document Benchmark Integrating Understanding, Reasoning, and Locating
• MINTQA: A Multi-Hop Question Answering Benchmark for Evaluating LLMs on New and Tail Knowledge
• PERC: Plan-As-Query Example Retrieval for Underrepresented Code Generation
• RAG-Star: Enhancing Deliberative Reasoning with Retrieval Augmented Verification and Refinement
• A Collaborative Multi-Agent Approach to Retrieval-Augmented Generation Across Diverse Data
• SLA Management in Reconfigurable Multi-Agent RAG: A Systems Approach to Question Answering
• Benchmarking Multimodal Retrieval Augmented Generation with Dynamic VQA Dataset and Self-adaptive Planning Agent
• Eliciting Critical Reasoning in Retrieval-Augmented Language Models via Contrastive Explanations
• CoPS: Empowering LLM Agents with Provable Cross-Task Experience Sharing
• RuleRAG: Rule-Guided Retrieval-Augmented Generation with Language Models for Question Answering
• StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization
• GIVE: Structured Reasoning of Large Language Models with Knowledge Graph Inspired Veracity Extrapolation
• Retrieval-Augmented Decision Transformer: External Memory for In-context RL
• ALR²: A Retrieve-then-Reason Framework for Long-context Question Answering
• Retrieval Augmented Generation (RAG) and Beyond: A Comprehensive Survey on How to Make your LLMs use External Data More Wisely
• Fact, Fetch, and Reason: A Unified Evaluation of Retrieval-Augmented Generation
• SFR-RAG: Towards Contextually Faithful LLMs
• Retrieval-Augmented Hierarchical in-Context Reinforcement Learning and Hindsight Modular Reflections for Task Planning with LLMs
• The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
• Re-Invoke: Tool Invocation Rewriting for Zero-Shot Tool Retrieval
• Improving Retrieval Augmented Language Model with Self-Reasoning
• MindSearch: Mimicking Human Minds Elicits Deep AI Searcher
• MMLongBench-Doc: Benchmarking Long-context Document Understanding with Visualizations
• BeamAggR: Beam Aggregation Reasoning over Multi-source Knowledge for Multi-hop Question Answering
• UDA: A Benchmark Suite for Retrieval Augmented Generation in Real-world Document Analysis
• Learning to Plan for Retrieval-Augmented Large Language Models from Knowledge Graphs
• Retrieval Meets Reasoning: Dynamic In-Context Editing for Long-Text Understanding
• AvaTaR: Optimizing LLM Agents for Tool Usage via Contrastive Reasoning
• VeraCT Scan: Retrieval-Augmented Fake News Detection with Justifiable Reasoning
• RATT: A Thought Structure for Coherent and Correct LLM Reasoning
• Chain of Agents: Large Language Models Collaborating on Long-Context Tasks
• MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark
• GNN-RAG: Graph Neural Retrieval for Large Language Model Reasoning
• M-RAG: Reinforcing Large Language Model Performance through Retrieval-Augmented Generation with Multiple Partitions
• DALK: Dynamic Co-Augmentation of LLMs and KG to answer Alzheimer's Disease Questions with Scientific Literature
• Recall, Retrieve and Reason: Towards Better In-Context Relation Extraction
• CBR-RAG: Case-Based Reasoning for Retrieval Augmented Generation in LLMs for Legal Question Answering
• Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity
• RAFT: Adapting Language Model to Domain Specific RAG
• Re-Search for The Truth: Multi-round Retrieval-augmented Large Language Models are Strong Fake News Detectors
• LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
• RAT: Retrieval Augmented Thoughts Elicit Context-Aware Reasoning in Long-Horizon Generation
• MATHSENSEI: A Tool-Augmented Large Language Model for Mathematical Reasoning
• SciAgent: Tool-augmented Language Models for Scientific Reasoning
• RAP: Retrieval-Augmented Planning with Contextual Memory for Multimodal LLM Agents
• MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries
• INTERS: Unlocking the Power of Large Language Models in Search with Instruction Tuning
• IAG: Induction-Augmented Generation Framework for Answering Reasoning Questions
• GAIA: a benchmark for General AI Assistants
• GPQA: A Graduate-Level Google-Proof Q&A Benchmark
• Chain-of-Note: Enhancing Robustness in Retrieval-Augmented Language Models
• JARVIS-1: Open-World Multi-task Agents with Memory-Augmented Multimodal Language Models
• A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions
• Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection
• InstructRetro: Instruction Tuning post Retrieval-Augmented Pretraining
• GROVE: A Retrieval-augmented Complex Story Generation Framework with A Forest of Evidence
• Making Retrieval-Augmented Language Models Robust to Irrelevant Context
• RA-DIT: Retrieval-Augmented Dual Instruction Tuning
• Knowledge Graph Prompting for Multi-Document Question Answering
• Counterfactually Auditable Lifecycle Certification for Autonomous Agents
• Think-on-Graph: Deep and Responsible Reasoning of Large Language Model on Knowledge Graph
• Dr.ICL: Demonstration-Retrieved In-context Learning
• Making Language Models Better Tool Learners with Execution Feedback
• ToolkenGPT: Augmenting Frozen Language Models with Massive Tools via Tool Embeddings
• UPRISE: Universal Prompt Retrieval for Improving Zero-Shot Evaluation
• Toolformer: Language Models Can Teach Themselves to Use Tools
• Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions
• Measuring and Narrowing the Compositionality Gap in Language Models
• Distributing Accountability, Not Capability: Phase Separation and the LLM Workflow Quadrant in Autonomous AI Agent Architectures
• Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering
• WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents
• Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models
• WebGPT: Browser-assisted question-answering with human feedback
• QuALITY: Question Answering with Long Input Texts, Yes!
• TopiOCQA: Open-domain Conversational Question Answering with Topic Switching
• CREAK: A Dataset for Commonsense Reasoning over Entity Knowledge
• MiniF2F: a cross-system benchmark for formal Olympiad-level mathematics
• QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering
• Measuring Mathematical Problem Solving With the MATH Dataset
• Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategies
• Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps
• IIRC: A Dataset of Incomplete Information Reading Comprehension Questions
• Explainable Automated Fact-Checking for Public Health Claims
• ALFWorld: Aligning Text and Embodied Environments for Interactive Learning
• RefactoringMiner 2.0
• BIGPATENT: A Large-Scale Dataset for Abstractive and Coherent Summarization
• PullNet: Open Domain Question Answering with Iterative Retrieval on Knowledge Bases and Text
• HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering
• Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional\n Neural Networks for Extreme Summarization
• CrisisMMD: Multimodal Twitter Datasets from Natural Disasters
• The Web as a Knowledge-base for Answering Complex Questions
• FEVER: a large-scale dataset for Fact Extraction and VERification
• DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset
• Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problems
• TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for\n Reading Comprehension
• Retrieval is Not Enough: Enhancing RAG Reasoning through Test-Time Critique and Optimization
• Talk to Right Specialists: Iterative Routing in Multi-agent Systems for Question Answering
• PersonaAgent: Bridging Memory and Action for Personalized LLM Agents
• Supervising the search process produces reliable and generalizable information-seeking agents

Cited by

• Detecting Hallucinations in Graph Retrieval-Augmented Generation via Attention Patterns and Semantic Alignment
• ReasonBENCH: Benchmarking the (In)Stability of LLM Reasoning
• AdmTree: Compressing Lengthy Context with Adaptive Semantic Trees
• MultivationBench: A Benchmark for Multimodal Sequential Motivation Reasoning
• Embracing Trustworthy Brain-Agent Collaboration as Paradigm Extension for Intelligent Assistive Technologies
• AutoGraph-R1: End-to-End Reinforcement Learning for Knowledge Graph Construction
• MADIAVE: Multi-Agent Debate for Implicit Attribute Value Extraction
• Atomic Thinking of LLMs: Decoupling and Exploring Mathematical Reasoning Abilities
• HeDA: An Intelligent Agent System for Heatwave Risk Discovery through Automated Knowledge Graph Construction and Multi-layer Risk Propagation Analysis
• From Deferral to Learning: Online In-Context Knowledge Distillation for LLM Cascades
• LexSemBridge: Fine-Grained Dense Representation Enhancement through Token-Aware Embedding Augmentation
• Reinforcement Learning Foundations for Deep Research Systems: A Survey
• Test-time Corpus Feedback: From Retrieval to RAG
• Structuring the Unstructured: A Systematic Review of Text-to-Structure Generation for Agentic AI with a Universal Evaluation Framework
• When Iterative RAG Beats Ideal Evidence: A Diagnostic Study in Scientific Multi-hop Question Answering
• XToM: Exploring the Multilingual Theory of Mind for Large Language Models
• RACT: Retrieval Augmented Column-Table Learning and Prediction for Multi-Table Schema Matching

Related

• Reasoning RAG via System 1 or System 2: A Survey on Reasoning Agentic Retrieval-Augmented Generation for Industry Challenges
• Synergizing RAG and Reasoning: A Systematic Review
• RAGCap-Bench: Benchmarking Capabilities of LLMs in Agentic Retrieval Augmented Generation Systems
• Open-RAG: Enhanced Retrieval-Augmented Reasoning with Open-Source Large Language Models
• RAG+: Enhancing Retrieval-Augmented Generation with Application-Aware Reasoning