FAPO: Fully Autonomous Prompt Optimization of Multi-Step LLM Pipelines

FAPO (Fully Autonomous Prompt Optimization) is a new framework designed to optimize multi-step LLM pipelines by evaluating and diagnosing failures across retrieval, reasoning, and formatting steps. It outperforms the baseline GEPA in 15 out of 18 model-benchmark comparisons, achieving a mean performance gain of +14.1 percentage points, and excels in security tasks, enhancing accuracy on the CTIBench-RCM dataset for models like GPT-5 and Foundation-Sec-8B. This framework's ability to autonomously propose and validate both prompt edits and structural changes makes it a significant advancement for practitioners aiming to improve LLM pipeline efficiency and effectiveness.