Learning to Self-Evolve

Xiaoyin Chen, Canwen Xu, Yite Wang, Boyi Liu, Zhewei Yao, Yuxiong He

Keywords: self-evolution, reinforcement learning, test-time adaptation, prompt optimization, large language models

TL;DR: We train LLMs with RL to improve their own contexts at test time.

Abstract: We introduce Learning to Self-Evolve (LSE), a reinforcement learning framework that trains LLMs to improve their own contexts at test time. We situate LSE in the setting of test-time self-evolution, where a model iteratively refines its context from feedback on seen problems to perform better on new ones. Existing approaches rely entirely on the inherent reasoning ability of the model and never explicitly train it for this task. LSE reduces the multi-step evolution problem to a single-step RL objective, where each context edit is rewarded by the improvement in downstream performance. We pair this objective with a tree-guided evolution loop. On Text-to-SQL generation (BIRD) and general question answering (MMLU-Redux), a 4B-parameter model trained with LSE outperforms self-evolving policies powered by GPT-5 and Claude Sonnet 4.5, as well as prompt optimization methods including GEPA and TextGrad, and transfers to guide other models without additional training. Our results highlight the effectiveness of treating self-evolution as a learnable skill.

Submission Number: 200