RLPE: Reinforcement Learned Prompt Engineering via Error Trace Analysis

RLPE trains a model to analyze task model failure traces and write targeted prompts to address those specific errors. Unlike current methods, the generated prompts are optimized for performance and can be reused without running the optimization model at inference time.