New SPLIT method speeds up Vision Transformers by learning token resolution

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-05 05:00

研究人员开发了一种名为Split Policy Learned Image Tokenization (SPLIT)的新方法，以提高Vision Transformers (ViTs)的效率。SPLIT根据模型对任务相关信息的解释来学习分配token分辨率，而不是依赖原始像素统计。这种方法产生的token序列更短，同时保持了准确性，在分类任务和COCO及ADE20K等密集预测任务上取得了有竞争力的结果，同时使用的token减少了高达30%。 AI

影响该方法有望实现更快、更高效的图像分析AI模型。

排序理由该集群包含一篇学术论文，详细介绍了一种提高AI模型效率的新方法。[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 →

New SPLIT method speeds up Vision Transformers by learning token resolution

本文如何被排名

Signal score: 26 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇学术论文，详细介绍了一种提高AI模型效率的新方法。[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.AI TIER_1 English(EN) · Aditya Gupta, Jean Sebastien Dandurand, Kai Qiu, Rohan Choudhury, Laszlo A. Jeni · 2026-11-05 05:00

学习分层 Patch 拆分策略以加速 Vision Transformers

arXiv:2611.05742v1 Announce Type: cross Abstract: Vision Transformers (ViTs) are bottlenecked by the length of their input token sequence. Standard patchification fixes this length with a single patch size applied across the image, regardless of regional content. Prior work on ad…

本文如何被排名

报道来源 [1]

学习分层 Patch 拆分策略以加速 Vision Transformers

相关实体

相关话题