PulseAugur
中
实时 08:07:45
English(EN) Rethinking Supervision Scaling in On-Policy Distillation: From Dense Supervision to Stochastic Horizons

新的在线策略蒸馏方法提高了LLM的训练效率

研究人员正在探索新的在线策略蒸馏(OPD)方法,以提高大型语言模型训练的效率和有效性。一种方法是流式在线策略蒸馏(Streaming On-Policy Distillation),它将响应生成、监督传递和参数更新解耦,从而实现连续生成和有针对性的损失计算。另一种方法是B-OPSD,它仅从问题生成问题摘要,在不需要真实答案的情况下提供教师指导,在数学推理任务上显示出更高的准确性。第三种技术是随机视界在线策略蒸馏(SH-OPD),它通过改变滚出视界来减少冗余监督,从而在保持推理和代码生成任务具有竞争力的性能的同时,节省了大量成本。 AI

影响 这些新的蒸馏技术可能导致更高效的大型语言模型训练,降低计算成本,并可能实现更快的开发周期。

排序理由 该集群包含多篇学术论文,详细介绍了LLM在线策略蒸馏的新研究方法。

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 →

新的在线策略蒸馏方法提高了LLM的训练效率

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
Research
该集群包含多篇学术论文,详细介绍了LLM在线策略蒸馏的新研究方法。
Source corroboration
3 independent sources
Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
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
2 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 [3]

  1. arXiv cs.AI TIER_1 English(EN) · Yun Zhu Da Wang ·

    流式策略内蒸馏:解耦三个时钟以进行分阶段监督消耗

    arXiv:2611.22298v1 Announce Type: new Abstract: In on-policy distillation (OPD), response completion, supervision delivery, and parameter updates need not share the same granularity. We develop streaming OPD by decoupling these three clocks: a response repeatedly delivers newly g…

  2. arXiv cs.CL TIER_1 English(EN) · Zhi Xu, Yinqi Wu, Feng Zhang, Xiaotong Zhang, Hong Yu, Han Liu ·

    B-OPSD:基于自生成问题摘要的策略内自蒸馏

    arXiv:2611.19273v2 Announce Type: replace Abstract: On-Policy Self-Distillation (OPSD) has recently attracted interest because it provides dense teacher guidance without a separate, often larger teacher model. However, OPSD conditions its teacher on ground-truth solutions, which …

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

    重新思考在线策略蒸馏中的监督扩展:从密集监督到随机视界

    On-policy distillation (OPD) provides dense token-level supervision along student-generated trajectories, yet the learning benefits of longer trajectories may not scale proportionally with their computational cost. We investigate how supervision utility evolves with trajectory le…