English(EN) GrowMTP: Can RL Grow Its Own Draft Head?

GrowMTP在RL循环中训练草稿头，加速LLM训练

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-16 04:00

研究人员开发了GrowMTP，一种在强化学习（RL）循环内完全训练用于推测性解码的草稿头的新颖方法。这种方法无需单独预训练草稿头，从而降低了训练成本。GrowMTP利用RL训练期间产生的窄回滚分布和连续监督信号来优化草稿头。在Qwen3-4B、MiMo-7B-SFT和Qwen3.5-4B-Base模型上的实验表明，回滚生成和端到端训练时间均显著加快。 AI

影响通过将草稿头优化直接集成到RL过程中，加速LLM训练，减少计算开销。

排序理由该集群描述了一篇关于训练AI模型的新颖方法的最新研究论文。[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: 15 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群描述了一篇关于训练AI模型的新颖方法的最新研究论文。[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, infra; 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.CL TIER_1 English(EN) · Minghua He, Lingzhe Zhang, Yuan Liu, Xiao Zhou, Aiwei Liu · 2026-09-16 04:00

GrowMTP：强化学习能否“自我培养”其草稿头？

arXiv:2609.16648v1 Announce Type: cross Abstract: Reinforcement learning (RL) post-training drives the frontier capabilities of large language models, with its wall-clock dominated by autoregressive rollout generation. Speculative decoding is an established remedy for this bottle…

本文如何被排名

报道来源 [1]

GrowMTP：强化学习能否“自我培养”其草稿头？

相关实体

相关话题