Search papers, labs, and topics across Lattice.
The paper introduces Learning to Self-Evolve (LSE), a reinforcement learning framework that trains LLMs to iteratively refine their contexts at test time based on performance feedback. LSE frames multi-step context evolution as a single-step RL objective, rewarding context edits that improve downstream performance, and uses a tree-guided evolution loop. Experiments on Text-to-SQL (BIRD) and question answering (MMLU-Redux) show that a 4B-parameter model trained with LSE outperforms self-evolving policies powered by GPT-5 and Claude Sonnet 4.5, demonstrating the benefits of explicitly training for self-evolution.
Forget prompt engineering – LSE trains LLMs to self-edit their own contexts at test time, outperforming even GPT-5 and Claude Sonnet 4.5 in Text-to-SQL and question answering.
We introduce Learning to Self-Evolve (LSE), a reinforcement learning framework that trains large language models (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.
Mila