MAGE: Multi-Agent Self-Evolution with Co-Evolutionary Knowledge Graphs

TL;DR

MAGE introduces a co-evolutionary knowledge graph framework for self-evolving language agents, enabling stable knowledge retention and task-specific guidance without altering the core model.

Contribution

The paper presents a novel multi-agent framework that externalizes self-knowledge into a co-evolving graph, improving self-evolution of language models across diverse benchmarks.

Findings

MAGE outperforms prompt-based frozen-backbone baselines on nine benchmarks.

Self-harvested success traces and teacher corrections are complementary.

Structural analysis shows stable improvement through memory growth and retrieval strategies.

Abstract

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.