Data Pruning Method Boosts LLM Fact Retention 1.3x, Matches 10x Larger Model

Saturday, April 11, 2026