English(EN) Less Loss, Fixed Weights: Logit Corrections for Forgetting

新方法试图修复语言模型的“遗忘”问题

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-02 05:00

研究人员开发了一种方法来缓解语言模型的“遗忘”现象，即模型在针对新指令或主题进行微调后，其预测标准文本的能力会下降。所提出的技术涉及对模型最终的 token 预测应用固定的调整，例如词汇偏差或温度修改。虽然这些固定的调整可以恢复高达 36% 在未见文本上损失的性能，但大部分原始损失仍未得到解决，特别是对于在数学数据集上进一步预训练的模型。 AI

影响这项研究探索了在微调后提高语言模型鲁棒性的方法，有望带来更稳定、更可靠的 AI 系统。

排序理由该条目是一篇研究论文，详细介绍了一种解决语言模型特定问题的新方法。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 1 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该条目是一篇研究论文，详细介绍了一种解决语言模型特定问题的新方法。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 1 days old; Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · Hanyu Yang · 2026-11-02 05:00

Less Loss, Fixed Weights: Logit Corrections for Forgetting

arXiv:2610.42000v1 Announce Type: cross Abstract: When a language model is fine-tuned to follow instructions, or further pretrained on a new subject, it often gets worse at predicting standard web text. This is known as "forgetting." We ask whether this loss can be undone by simp…

本文如何被排名

报道来源 [1]

Less Loss, Fixed Weights: Logit Corrections for Forgetting

相关实体

相关话题