GEPAA framework for optimizing textual system components using LLM-based reflection and Pareto-efficient evolutionary search.https://gepa-ai.github.io/gepa/ https://github.com/gepa-ai/gepaen Wed, 13 May 2026 17:13:17 -0000 Wed, 13 May 2026 17:13:17 -0000 1440 MkDocs RSS plugin - v1.19.0 None GEPA https://gepa-ai.github.io/gepa/ Introducing the GEPA Blog Lakshya A Agrawal Research updates, applications, and community case studies from the GEPA ecosystem. https://gepa-ai.github.io/gepa/blog/2026/02/13/introducing-the-gepa-blog/ Wed, 13 May 2026 17:09:32 +0000 GEPAhttps://gepa-ai.github.io/gepa/blog/2026/02/13/introducing-the-gepa-blog/ Automatically Learning Skills for Coding Agents Shangyin Tan Lakshya A Agrawal Rohit Sandadi Dan Klein Koushik Sen Alexandros G. Dimakis Matei Zaharia Introducing gskill, a fully automated pipeline that uses GEPA and SWE-smith to learn repository-specific skills for coding agents. https://gepa-ai.github.io/gepa/blog/2026/02/18/automatically-learning-skills-for-coding-agents/ Wed, 13 May 2026 17:09:32 +0000 GEPAhttps://gepa-ai.github.io/gepa/blog/2026/02/18/automatically-learning-skills-for-coding-agents/ optimize_anything: A Universal API for Optimizing any Text Parameter Lakshya A Agrawal Donghyun Lee Wenjie Ma Karim Elmaaroufi Shangyin Tan Sanjit A. Seshia Koushik Sen Dan Klein Ion Stoica Joseph E. Gonzalez Omar Khattab Alexandros G. Dimakis Matei Zaharia GEPA's new API setting state-of-the-art results on optimizing any text parameter: code, prompts, agent architectures, and more. If you can measure it, you can optimize it. https://gepa-ai.github.io/gepa/blog/2026/02/18/introducing-optimize-anything/ Wed, 13 May 2026 17:09:32 +0000 GEPAhttps://gepa-ai.github.io/gepa/blog/2026/02/18/introducing-optimize-anything/ Confidence-Aware Prompt Optimization for LLM Classification Rodolfo Nobrega ConfidenceAdapter uses token-level log-probabilities to score prompt candidates on a continuous scale instead of binary correct/wrong, producing better disambiguation rules and higher-accuracy prompts for classification tasks. https://gepa-ai.github.io/gepa/blog/2026/03/17/confidence-adapter-benchmark/ Wed, 13 May 2026 17:09:32 +0000 GEPAhttps://gepa-ai.github.io/gepa/blog/2026/03/17/confidence-adapter-benchmark/ Scaling GEPA with Combee: Parallel Prompt Learning for Self-Improving Agents Hanchen Li Runyuan He Qizheng Zhang Changxiu Ji Qiuyang Mang Xiaokun Chen Lakshya A Agrawal Wei-Liang Liao Eric Yang Alvin Cheung James Zou Kunle Olukotun Ion Stoica Joseph E. Gonzalez Exploring how to scale GEPA's prompt learning across many parallel agents using the Combee framework, achieving up to 17× speedup with no quality loss. https://gepa-ai.github.io/gepa/blog/2026/04/09/gepa-at-scale-with-combee/ Wed, 13 May 2026 17:09:32 +0000 GEPAhttps://gepa-ai.github.io/gepa/blog/2026/04/09/gepa-at-scale-with-combee/ Learning, Fast and Slow: LLMs That Adapt Continually Rishabh Tiwari Kusha Sareen Lakshya A Agrawal Joseph E. Gonzalez Matei Zaharia Kurt Keutzer Inderjit S Dhillon Rishabh Agarwal Devvrit Khatri Fast-Slow Training (FST) interleaves prompt optimization with reinforcement learning, treating the prompt as fast weights and parameters as slow weights — improving data efficiency, performance ceiling, plasticity, and continual learning. https://gepa-ai.github.io/gepa/blog/2026/05/11/learning-fast-and-slow/ Wed, 13 May 2026 17:09:32 +0000 GEPAhttps://gepa-ai.github.io/gepa/blog/2026/05/11/learning-fast-and-slow/