PulseAugur
中
实时 22:57:34
English(EN) Deferred Audio Pruning with Local Audio-Visual Dynamics for Omni-LLMs

新的 A-PACK 框架通过延迟音频剪枝大幅降低全模态大语言模型的成本

作者 PulseAugur 编辑部 · [1 个来源] ·

研究人员推出了一种新颖的两阶段框架 A-PACK，旨在降低全模态大语言模型（omni-LLM）相关的计算成本。该方法将音频剪枝推迟到查询条件下的多模态交互变得相关时进行，解决了处理长多模态序列的显著预填充和 KV 缓存费用问题。A-PACK 优先考虑音频信息密度，并利用局部视听动态进行更有效的视觉选择，从而在 Qwen2.5-Omni 模型上的基准测试中大幅减少计算操作并提高解码吞吐量。 AI

影响 降低了全模态大语言模型的计算成本，可能能够更有效地处理长多模态序列。

排序理由 该集群包含一篇详细介绍提高大语言模型效率新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

新的 A-PACK 框架通过延迟音频剪枝大幅降低全模态大语言模型的成本

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
Tool
该集群包含一篇详细介绍提高大语言模型效率新方法的论文。[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
58 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

  1. arXiv cs.AI TIER_1 English(EN) · Kyeongyoon Lee, Hongyeob Kim, Youngeun Kim, Sungeun Hong ·

    面向全能大模型的延迟音频剪枝与局部视听动态

    arXiv:2608.08794v1 Announce Type: new Abstract: Omni-modal LLMs jointly process audio, video, and text, but long multimodal sequences incur substantial prefill and KV-cache costs. Existing omni-modal compression methods primarily focus on pre-LLM token reduction, leaving modality…

相关实体

相关话题