Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning

Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning explores PACT fine-tuning framework stabilizes LLM safety by constraining confidence on safety tokens, preventing alignment drift during downstream tasks.. Commercial viability score: 7/10 in LLM Safety.