LLM New Method Speeds Up LLM Data Valuation with Shapley Values Researchers propose efficient Shapley value approximation using language model arithmetic to determine which training data samples matter most for LLM fine-tuning.
