AI RESEARCH
Forgetting: A New Mechanism Towards Better Large Language Model Fine-tuning
arXiv CS.LG
•
ArXi:2508.04329v5 Announce Type: replace Supervised fine-tuning (SFT) plays a critical role for pretrained large language models (LLMs), notably enhancing their capacity to acquire domain-specific knowledge while preserving or potentially augmenting their general-purpose capabilities. However, the efficacy of SFT hinges on data quality as well as data volume, otherwise it may result in limited performance gains or even degradation relative to the associated baselines.