GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning
This addresses the problem of high computational cost in fine-tuning LLMs for researchers and practitioners, offering a more efficient alternative to RL methods.
The paper tackles the inefficiency of reinforcement learning methods like GRPO for adapting large language models to tasks by introducing GEPA, a prompt optimizer that uses natural language reflection to learn from trial and error. It shows that GEPA outperforms GRPO by 10% on average and up to 20% with up to 35x fewer rollouts, and beats MIPROv2 by over 10% across tasks.
Large language models (LLMs) are increasingly adapted to downstream tasks via reinforcement learning (RL) methods like Group Relative Policy Optimization (GRPO), which often require thousands of rollouts to learn new tasks. We argue that the interpretable nature of language can often provide a much richer learning medium for LLMs, compared with policy gradients derived from sparse, scalar rewards. To test this, we introduce GEPA (Genetic-Pareto), a prompt optimizer that thoroughly incorporates natural language reflection to learn high-level rules from trial and error. Given any AI system containing one or more LLM prompts, GEPA samples system-level trajectories (e.g., reasoning, tool calls, and tool outputs) and reflects on them in natural language to diagnose problems, propose and test prompt updates, and combine complementary lessons from the Pareto frontier of its own attempts. As a result of GEPA's design, it can often turn even just a few rollouts into a large quality gain. Across four tasks, GEPA outperforms GRPO by 10% on average and by up to 20%, while using up to 35x fewer rollouts. GEPA also outperforms the leading prompt optimizer, MIPROv2, by over 10% across two LLMs, and demonstrates promising results as an inference-time search strategy for code optimization.