English(EN) Later Is Better: Token Reduction for ViTs Under Distribution Shift

新的 ViT Token 缩减策略提升了分布外准确率

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-07 04:00

研究人员为 Vision Transformer (ViTs) 开发了一种新的 Token 缩减策略，可在分布偏移下提高性能。这种“后期集中式”策略在后续层中移除更多 Token，与标准的“平坦式”策略相比，始终能提高分布外准确率。该方法以计算量的一小部分恢复了大部分原始准确率，在 ImageNet-C 和其他偏移数据集上，跨不同骨干网络和模态均显示出显著的提升，且无需进行每输入调优。 AI

影响这项研究有望带来更鲁棒、更高效的 Vision Transformer 模型，尤其是在数据分布偏移常见的实际应用中。

排序理由详细介绍一种提高模型性能的新方法的学术论文。[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 →

本文如何被排名

Signal score: 15 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍一种提高模型性能的新方法的学术论文。[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.AI TIER_1 English(EN) · Hyeongheon Cha, Hyungjun Yoon, Sung-Ju Lee · 2026-10-07 04:00

迟到者更优：分布偏移下ViT的Token缩减

arXiv:2610.07758v1 Announce Type: cross Abstract: Training-free token reduction accelerates vision transformers by removing redundant tokens across layers, recovering most of the original accuracy at a fraction of the compute. These methods, however, are designed and evaluated pr…

本文如何被排名

报道来源 [1]

迟到者更优：分布偏移下ViT的Token缩减

相关实体

相关话题