English(EN) Rethinking "RL Generalizes, SFT Memorizes": The Role of SFT Data

研究发现SFT数据规模和来源对LLM泛化至关重要

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-14 05:00

一篇新的研究论文挑战了普遍认为的，在大型语言模型中，与监督微调（SFT）相比，强化学习（RL）能带来更好泛化的观点。该研究以Qwen2.5和Qwen3模型为重点，证明了精心策划和规模化的SFT数据集可以实现与RL相当甚至更优的泛化能力。研究强调数据来源和规模是关键因素，混合SFT数据集的表现优于仅由更大模型生成的数据集。通过分析训练动态并涵盖更多样化的推理信号，SFT可以有效地匹配RL的泛化性能。 AI

影响挑战了关于LLM训练的传统观念，表明通过优化数据，SFT可以匹配RL的泛化能力。

排序理由学术论文发布在arXiv上，详细介绍了关于LLM训练方法的新发现。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 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: 17 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文发布在arXiv上，详细介绍了关于LLM训练方法的新发现。[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: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Yunlong Hou, Fengzhuo Zhang, Yuan Cheng, Jiachun Pan, Xingyao Li, Zhuoran Yang · 2026-11-14 05:00

重新思考“RL泛化，SFT记忆”：SFT数据的作用

arXiv:2611.14912v1 Announce Type: cross Abstract: Large language models trained with Reinforcement Learning (RL) with verifiable rewards exhibit strong reasoning ability and broad generalization, whereas models trained with Supervised Fine-Tuning (SFT) are often viewed as more pr…

本文如何被排名

报道来源 [1]

重新思考“RL泛化，SFT记忆”：SFT数据的作用

相关实体

相关话题