Tag
This paper conducts a causal-inspired analysis of automated prompt optimization across frameworks, LLMs, and tasks, identifying that specific edit types (e.g., complexity-increasing, meta-instructional) have systematic negative or positive effects depending on task characteristics, explaining generalization failures.
Introduces Reflective Prompt Tuning (RPT), a framework that uses LLM function-calling to iteratively diagnose and revise prompts based on systematic error patterns, improving reasoning task performance and calibration.