EXG: Self-Evolving Agents with Experience Graphs
About
Large language model (LLM)-based agents have demonstrated strong capabilities in complex reasoning and problem solving through multi-step interactions, yet most deployed agents remain behaviorally static, with knowledge acquired during execution rarely translating into systematic improvement over time. In response, a growing line of work on self-evolving agents explores how agents can improve through experience during deployment, but most existing approaches either rely on ad hoc reflection limited to single-task correction or adopt unstructured memory that accumulates fragmented experience with delayed usability. To address this limitation, we introduce EXG, an experience graph framework for self-evolving agents that explicitly organizes accumulated successes and failures into a structured, relational representation. EXG is the first experience graph designed for self-evolving agents, supporting both online, real-time graph growth during execution for immediate cross-task experience reuse, and offline reuse of a consolidated experience graph as an external memory module. This design also enables EXG to serve as a plug-and-play component for existing self-evolving agents, organizing prior experience into a unified experience graph and improving both solution quality and resource efficiency as deployment progresses. Extensive experiments across code generation and reasoning benchmarks show that EXG attains more favorable performance-efficiency trade-offs than reflection- and memory-based baselines in both online and offline evaluations. Our results suggest that structuring experience as a graph provides a principled foundation for scalable and transferable self-evolving agent behavior.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Code Generation | EvalPlus | Pass@181.7 | 118 | |
| Code Generation | HumanEval | pass@186 | 21 | |
| Multi-hop Question Answering | MuSiQue | Pass@169.4 | 21 | |
| Multi-hop Question Answering | HotpotQA | pass@1 Success Rate65.9 | 21 | |
| Code Generation | HumanEval Online Collection | pass@182.4 | 2 | |
| Multi-hop Question Answering | HotpotQA (Online Collection Split) | Pass@161.4 | 2 | |
| Multi-hop Question Answering | HotpotQA Offline (test) | Pass@161 | 2 | |
| Code Generation | HumanEval Offline (test) | pass@187.9 | 2 |