Iterative Refinement of Large Language Model Outputs Using Fine-Grained Actionable Feedback
LLMRefine, an inference-time optimization method, iteratively refines the output of large language models using a learned fine-grained feedback model to pinpoint defects and guide the refinement process.
