Tag
1 articles
A bias-controlled fine-tuning method improves LLM self-knowledge and generalizes across unseen data, languages, and new facts.
