AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens

Yiming Qin, Bomin Wei, Jiaxin Ge, Konstantinos Kallidromitis, Stephanie Fu, Trevor Darrell, XuDong Wang

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 80 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add CoVT with 3 visual tokensQwen2.5-VL-7B · CV-Benchaccuracy: 74.5 -> 80.0Improved
add CoVT with 3 visual tokensQwen2.5-VL-7B · CV-Bench Depthaccuracy: 72.8 -> 86.8Improved
add CoVT with 3 visual tokensQwen2.5-VL-7B · MME-RealWorldaccuracy: 68.6 -> 72.9Improved
add CoVT with 3 visual tokensQwen2.5-VL-7B · HRBench-8Kaccuracy: 64.9 -> 69.4Improved
add edge token to make four visual tokensCoVT · CV-Bench Depthaccuracy: 86.8 -> 89.2Improved
add edge token to make four visual tokensCoVT · MMVPaccuracy: 58.7 -> 56.7Hurt
add segmentation tokens (8) instead of 0 token fine-tuning (ablation)CoVT · CV-Benchaccuracy: 76.6 -> 80.0Improved
use 16 empty tokens without visual alignment (ablation)CoVT · CV-Benchaccuracy: 76.6 -> 75.7Hurt
increase segmentation tokens from 8 to 32 (ablation)CoVT · CV-Benchaccuracy: 80.0 -> 73.9Hurt
use only 1 segmentation token instead of 8 (ablation)CoVT · CV-Benchaccuracy: 80.0 -> 78.9Hurt
align tokens through task decoders instead of direct feature MSE (ablation)CoVT Seg · BLINKaccuracy: 55.2 -> 57.4Improved
align tokens through task decoders instead of direct feature MSE (ablation)CoVT Depth · CV-Benchaccuracy: 77.0 -> 78.7Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy