PulseAugur
中
实时 07:39:52
English(EN) Beyond Passive Imitation: On-Policy Mid-Training Distillation for Language Models

新的蒸馏方法改进了LLM的领域适应性

研究人员开发了一种名为策略内中期蒸馏(OPMD)的新方法,以改进大型语言模型(LLM)在专业领域的适应性。与平等对待所有训练数据的传统方法不同,OPMD会主动从不断发展的模型中采样轨迹,并使用教师模型的奖励,从而解决了训练-生成不匹配的问题。这种方法在Qwen等模型的数学基准测试中表现出改进,优于标准的下一个词预测方法。OPMD可以集成到现有的训练流程中,而不会增加推理时间。 AI

影响 这项研究可能导致LLM更高效、更有效的领域适应性,从而可能提高它们在专业任务上的性能。

排序理由 该集群包含两篇学术论文,详细介绍了一种新的语言模型训练方法及其对训练-推理不匹配的容忍度。

在 Hugging Face Daily Papers 阅读 →

AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →

新的蒸馏方法改进了LLM的领域适应性

本文如何被排名

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Research
该集群包含两篇学术论文,详细介绍了一种新的语言模型训练方法及其对训练-推理不匹配的容忍度。
Source corroboration
3 independent sources
Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
Topics
model release, paper
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
8 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.
Coverage growth since scoring
+1 source(s) since last score
New sources have picked up this story since our last re-score. Score will update on the next scoring pass.

完整方法见我们的编辑标准。

报道来源 [3]

  1. arXiv cs.CL TIER_1 English(EN) · Ruize Zheng, Xuyang Liu, Siyu Ren, Xiaobing Tu, Yinggui Wang, Jinkui Ren, Xiantao Zhang, Linfeng Zhang ·

    超越被动模仿:语言模型的在线策略中期蒸馏

    arXiv:2611.13855v1 Announce Type: new Abstract: Continual pre-training (CPT) adapts large language models (LLMs) to specialized domains, yet its standard next-token prediction (NTP) objective treats every corpus token as a fixed target, regardless of its utility to the current mo…

  2. Hugging Face Daily Papers TIER_1 English(EN) ·

    超越被动模仿:语言模型的在线策略中期蒸馏

    Continual pre-training (CPT) adapts large language models (LLMs) to specialized domains, yet its standard next-token prediction (NTP) objective treats every corpus token as a fixed target, regardless of its utility to the current model. In a controlled study on the same 1B-token …

  3. Hugging Face Daily Papers TIER_1 English(EN) ·

    On-Policy Distillation 的训练-推理不匹配容忍度有多大?

    On-policy distillation tolerates substantial training-inference mismatch from quantized inference and asynchronous rollouts, with model- and layer-specific limits, and a lightweight admission check plus FP8 recipes can exploit this tolerance to improve training efficiency.