PulseAugur
中
实时 13:15:37
English(EN) LAST: The Last Query Token Guides Visual Token Pruning for Edge-Cloud Collaborative MLLM Inference

新方法剪枝视觉令牌以实现高效MLLM推理 · 跟踪4个来源

作者 PulseAugur 编辑部 · [4 个来源] ·

研究人员开发了多种新方法来高效剪枝多模态大语言模型（MLLM）的视觉令牌，旨在降低推理成本和延迟。LAST框架利用最后一个查询令牌的注意力来指导剪枝，无需访问云端模型，即可在将令牌数量减少87.5%的同时保留94.5%的准确率。SFPruner将剪枝重新构建为单次前向传播，将Qwen2.5-VL的令牌选择时间从112.4毫秒显著缩短至2.5毫秒。SPARE是另一种方法，将剪枝视为子空间重建，通过最小化重建误差并纳入“反相关性”标准，在LLaVA上可移除高达94%的令牌，同时保持95%的基线性能。 AI

影响 这些视觉令牌剪枝方面的进展可以显著降低MLLM的推理延迟和计算成本，从而在边缘设备上实现更高效的部署并提高可访问性。

排序理由 多篇研究论文介绍了优化多模态大语言模型的新技术。

在 Hugging Face Daily Papers 阅读 →

AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →

新方法剪枝视觉令牌以实现高效MLLM推理 · 跟踪4个来源

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
Research
多篇研究论文介绍了优化多模态大语言模型的新技术。
Source corroboration
4 independent sources
Strong cross-source corroboration — multiple independent publishers covered this within the clustering window.
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
45 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 [4]

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

    LAST：最后一个查询令牌指导视觉令牌修剪，用于边缘-云协同MLLM推理

    Multimodal foundation models are reshaping edge-cloud visual intelligence from task-specific feature pipelines into token-based interfaces, where edge devices encode visual inputs into tokens for a general-purpose cloud MLLM. However, dense visual-token sequences increase cloud-s…

  2. arXiv cs.CV TIER_1 English(EN) · Feng Yang, Xinrui Ju, Keyang Zhang, Xiandong Meng, Rongqun Lin, Howard Leung, Shiqi Wang, Haoliang Li, Chris Xing Tian ·

    LAST：最后一个查询令牌指导视觉令牌修剪，实现边缘-云协同MLLM推理

    arXiv:2607.27952v1 Announce Type: new Abstract: Multimodal foundation models are reshaping edge-cloud visual intelligence from task-specific feature pipelines into token-based interfaces, where edge devices encode visual inputs into tokens for a general-purpose cloud MLLM. Howeve…

  3. arXiv cs.CV TIER_1 English(EN) · Jouwon Song, Woohyeong Kim, Kyeongbo Kong ·

    面向高分辨率MLLM的高效视觉Token剪枝的结构化冗余建模

    arXiv:2607.23046v1 Announce Type: new Abstract: Recent high-resolution Multimodal Large Language Models (MLLMs) generate thousands of visual tokens per input, leading to a visual token explosion that introduces severe latency bottlenecks. While token pruning mitigates this issue,…

  4. arXiv cs.CV TIER_1 English(EN) · Jaeyeon Lee, Shunjie Wen, Dong-Wan Choi ·

    超越多样性：视觉令牌剪枝作为高效VLM的子空间重建

    arXiv:2606.18681v2 Announce Type: replace Abstract: Despite their remarkable performance, Vision Language Models (VLMs) incur substantial computational overhead due to the large number of visual tokens. While diversity maximization has become a dominant strategy for token reducti…

相关实体

相关话题