English(EN) Learning to Reason with Compressed Context: Ground-Truth-Free Adaptation of OmniLLMs via Self-Distillation

新框架在无真实标签情况下使 OmniLLMs 适应压缩上下文

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-01 04:00

研究人员开发了一种名为 CAFD（Compressed-Context Adaptation via Full-Context Distillation，通过全上下文蒸馏实现压缩上下文适应）的新型自蒸馏框架，以提高全模态大语言模型 (OmniLLMs) 的性能。该方法使 OmniLLMs 能够在不需要真实标签或奖励的情况下适应压缩的多模态令牌序列。通过将样本的全上下文视图作为特权信息，CAFD 使用全上下文自教师的软目标来训练压缩上下文学生模型。在 Qwen2.5-Omni-7B 上的实验表明，在各种压缩管道和部署预算下，准确性得到了一致的提高。 AI

影响该方法可以提高已部署的全模态 LLM 的效率和准确性，使其在实际应用中更加实用。

排序理由该集群包含一篇详细介绍大语言模型适应新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CV 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: 21 / 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大语言模型适应新方法的学术论文。[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: 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.CV TIER_1 English(EN) · Jianghao Wang, Ke Meng, Jian Li, Chi Cheng, Longyu Qi, Liyin Liang, Yifeng Qian, Chunbo Lai, Yutian Lin, Zeyu Wang · 2026-10-01 04:00

通过自蒸馏实现 OmniLLMs 的无真实标签压缩上下文推理学习适应

arXiv:2609.39953v1 Announce Type: new Abstract: Omni-modal large language models (OmniLLMs) enable unified audio-video understanding, but their long multimodal token sequences make deployment computationally expensive. Token compression reduces this cost, yet aggressive compressi…

本文如何被排名

报道来源 [1]

通过自蒸馏实现 OmniLLMs 的无真实标签压缩上下文推理学习适应

相关实体

相关话题