English(EN) Why Constrained Updates Still Forget: A Functional-Path Analysis of Continual SFT

新分析揭示了AI模型在更新过程中为何会遗忘

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-23 04:00

研究人员开发了一种新的函数路径分析方法，以理解语言模型的监督微调（SFT）为何会导致遗忘先前获得的行为。该分析将旧损失的变化分解为受控方向、梯度陈旧性和回放有效性等因素。研究发现，QLoRA干预措施相对于Adam等优化器的位置，对保留率有显著影响，而检查点生成有助于区分终点恢复和路径准确性。研究结果表明，显式的函数锚点、轨迹评估以及对旧任务和新任务的联合审计，对于减轻持续学习场景中的遗忘至关重要。 AI

影响提供了对模型遗忘的更深入理解，可能导致改进持续学习和模型更新的方法。

排序理由学术论文，详细介绍了AI模型持续学习的新分析方法。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 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: 4 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文，详细介绍了AI模型持续学习的新分析方法。[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: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Qingsen Ma, Dianyun Wang, Jiahong Liu, Irwin King · 2026-10-23 04:00

Why Constrained Updates Still Forget: A Functional-Path Analysis of Continual SFT

arXiv:2610.30191v1 Announce Type: cross Abstract: Supervised fine-tuning adapts language models to specialized tasks but can disrupt previously acquired behavior. Existing approaches constrain parameters, gradients, or replay objectives, yet the connection between the protected o…

本文如何被排名

报道来源 [1]

Why Constrained Updates Still Forget: A Functional-Path Analysis of Continual SFT

相关实体

相关话题