Prompt Optimization for LLM Code Generation via Reinforcement Learning

Collection of recent SE/ML research papers focused on improving reliability, correctness, and maintainability of code and software engineering workflows. Highlights include a PPO-based reinforcement learning framework for iterative prompt optimization that materially improves LLM code-generation Pass@1 on standard benchmarks; multi-agent LLM pipelines for knowledge-graph extraction to automate test-case generation from technical manuals; a MAPE-K self-healing framework with AutoFix-like recovery for web apps; a mutation-guided multi-agent approach (MuMuTestUp) to automatically update tests in/