PulseAugur
中
实时 20:24:36

新OnQ流水线提升量化大模型推理能力

研究人员开发了一种名为OnQ的新流水线,用于提高经过量化感知训练(QAT)的低比特大语言模型(LLMs)的推理能力。现有的QAT方法在长程推理和令牌预算限制方面常常遇到困难。OnQ通过使用on-policy蒸馏来解决这些问题,其中量化模型生成自己的前缀,并接收来自全精度教师模型的监督。此外,它采用动态KL(Dynamic KL)来平衡欠预测和过预测的校正,通过根据候选概率比率加权KL散度来实现。在2位和3位精度的Qwen3模型上进行的实验表明,OnQ在九个基准测试中显著优于现有基线,尤其是在长程推理任务中,并且产生的响应更短、重复性更低。 AI

影响 这项研究通过提高量化模型的性能,可能导致更高效、更强大的LLMs,从而降低计算成本并实现更广泛的部署。

排序理由 该集群包含一篇详细介绍改进LLM性能新方法的学术论文。[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 →

新OnQ流水线提升量化大模型推理能力

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
Tool
该集群包含一篇详细介绍改进LLM性能新方法的学术论文。[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, 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
1 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 [1]

  1. arXiv cs.CL TIER_1 English(EN) · Boyuan Zheng, Jing Xiong, Chenwei Wu, Hui Shen, Zesen Zhao, Hei Ting Una Chan, Xueshen Liu, Zhen Zhang, Liyue Shen, Zechun Liu, Z. Morley Mao ·

    Right On Q: Recovering Reasoning under Quantization via Self-Generated Trajectories

    arXiv:2610.29378v1 Announce Type: new Abstract: Quantization-aware training (QAT) aims to recover reasoning capabilities in low-bit large language models (LLMs). Existing recipes typically supervise the quantized student with the original full-precision model on fixed response se…