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

ArXi:2605.10064v1 Announce Type: new 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 s a frozen weak backbone at inference time. This paper