TAPR: A Task-Aware Prompt Rewriter That Improves LLM Performance Automatically

A new paper trains a small model to automatically rewrite your prompts into better-optimized ones, using reinforcement learning graded by an LLM judge — essentially outsourcing prompt engineering.