新闻
Less Loss, Fixed Weights: Logit Corrections for Forgetting
arXiv cs.AI · 发布于 · 阅读约1分钟
30秒读懂
- 发生了什么
- Researchers tested whether fixed token-level adjustments can recover performance lost when language models are fine-tuned or retrained on new domains.
- 为何重要
- Matters for teams deploying instruction-tuned or domain-adapted models that need to maintain baseline performance on general text.
- 注意
- Simple fixes recovered at most 36% of losses; models retrained on math recovered almost nothing, suggesting the approach has fundamental limits.
- language model
- fine-tun
- token
这条新闻背后的模式
每个模式都讲清楚技术如何运作、何时值得投入，以及在哪里会失效。
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。