MAGE: Multi-Agent Self-Evolution with Co-Evolutionary Knowledge Graphs
On the map: LLM agents and tool use → General LLM agent frameworks and benchmarks
Abstract (arXiv)
Self-evolving language-model agents must decide what to learn next and how to preserve what they have learned across iterations. Existing systems typically carry this cross-iteration knowledge as natural-language feedback, flat episodic memory, or implicit reinforcement signals, none of which cleanly supports a frozen weak backbone at inference time. This paper introduces MAGE (Multi-Agent Graph-guided Evolution), a framework that externalizes self-knowledge into a four-subgraph co-evolutionary knowledge graph. Its experience subgraph stores both teacher-written failure corrections and the learner's own past correct reasoning traces, which are retrieved as task-conditioned guidance for a frozen execution model. During evolution, the graph, a task-level search bandit, and a skill-level routing bandit are updated from the same reward stream, while the learner's backbone remains unchanged. We further provide structural analysis showing how append-only memory growth, bounded curriculum coverage, and task-filtered retrieval together support stable improvement of the retrieval substrate for frozen-learner evolution. Across nine benchmarks spanning mathematical reasoning, multi-hop and open-domain question answering, spatio-temporal analysis, financial numerical reasoning, medical multiple-choice, an open-world survival game, and web navigation, MAGE achieves strong performance against prompt-based frozen-backbone baselines. Ablations show that self-harvested success traces and teacher-written corrections are complementary, with success memories contributing most on reasoning-template-heavy tasks and corrective memories supporting harder composition and interaction settings. Code is available in https://github.com/RuiyiYang01/mage-alg
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use. Builds on 30 of its 33 references that are in the map. First posted on arXiv 11 May 2026.
Builds on
30 of 33 references.
Closest ancestors · 8
Same direction, newest first
- 2026Experiential Reflective Learning for Self-Improving LLM Agents25 Mar
- 2026SAGE: Multi-Agent Self-Evolution for LLM Reasoning16 Mar
- 2025Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning20 Nov
- 2025AgentEvolver: Towards Efficient Self-Evolving Agent System13 Nov
- 2025EvolveR: Self-Evolving LLM Agents through an Experience-Driven Lifecycle17 Oct
- 2025InfiAgent: Self-Evolving Pyramid Agent Framework for Infinite Scenarios26 Sep
- 2025SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience6 Aug
- 2025SiriuS: Self-improving Multi-agent Systems via Bootstrapped Reasoning7 Feb
Other references · 22
- 2021Training Verifiers to Solve Math Word Problemscited by 9,743
- 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
- 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
- 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
- 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,686
- 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
- 2024From Local to Global: A Graph RAG Approach to Query-Focused Summarizationcited by 1,411
- 2020What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Examscited by 1,324
- 2022WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agentscited by 1,197
- 2025A-MEM: Agentic Memory for LLM Agentscited by 1,000
- 2021FinQA: A Dataset of Numerical Reasoning over Financial Datacited by 552
- 2025RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learningcited by 357
10 more
- 2025Absolute Zero: Reinforced Self-play Reasoning with Zero Datacited by 323
- 2021Benchmarking the Spectrum of Agent Capabilitiescited by 193
- 2025Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memorycited by 129
- 2026MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memorycited by 113
- 2025Multi-Agent Evolve: LLM Self-Improve through Co-evolutioncited by 58
- 2025RealMath: A Continuous Benchmark for Evaluating Language Models on Research-Level Mathematicscited by 30
- 2024STBench: Assessing the Ability of Large Language Models in Spatio-Temporal Analysiscited by 11
- 2025AGENTiGraph: A Multi-Agent Knowledge Graph Framework for Interactive, Domain-Specific LLM Chatbotscited by 9
- 2025Divide by Question, Conquer by Agent: SPLIT-RAG with Question-Driven Graph Partitioningcited by 2
- 2025Agentic-KGR: Co-evolutionary Knowledge Graph Construction through Multi-Agent Reinforcement Learningcited by 1
Built on by
Citing papers per year: 2026: 2.
2 papers.
Similar, not cited
Close in wording, no citation either way
- 2026CoEvoKG: Co-Evolving Knowledge Graphs with Self-Evolving Search Agents
- 2025A Survey of Self-Evolving Agents: What, When, How, and Where to Evolve on the Path to Artificial Super Intelligence
- 2026Search-E1: Self-Distillation Drives Self-Evolution in Search-Augmented Reasoning
- 2025SE-Agent: Self-Evolution Trajectory Optimization in Multi-Step Reasoning with LLM-Based Agents
- 2025The Path of Self-Evolving Large Language Models: Achieving Data-Efficient Learning via Intrinsic Feedback
- 2026SEARL: Joint Optimization of Policy and Tool Graph Memory for Self-Evolving Agents
- 2026SkillEvo: Self-Renewing Evolution Gradients from Multi-Turn Interaction Feedback
- 2026DemoEvolve: Overcoming Sparse Feedback in Agentic Harness Evolution with Demonstrations