AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Paying More Attention to Visual Tokens in Self-Evolving Large Multimodal Models

Shravan Venkatraman, Ritesh Thawkar, Omkar Thawakar, Rao M Anwer, Hisham Cholakkal, Salman Khan, Fahad Shahbaz Khan

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 4 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 geometric invariance reward only (ablation)VISE on Qwen3-VL-2B · COCO CaptionCIDEr: 21.54 -> 26.37Improved
add semantic invariance (ghosting) reward only (ablation)VISE on Qwen3-VL-2B · COCO CaptionCIDEr: 21.54 -> 35.53Improved
combine geometric reward with semantic reward (ablation)VISE on Qwen3-VL-2B · COCO CaptionCIDEr: 35.53 -> 38.39Improved
add geometric reward to semantic reward (ablation)VISE on Qwen3-VL-2B · COCO Cap Chair-IChair-I: 9.06 -> 8.21Improved
train with VISE on unlabeled COCO imagesQwen3-VL-2B · COCO CaptionCIDEr: 21.54 -> 38.39Improved
train with VISEQwen3-VL-2B · COCO Cap Chair-IChair-I: 13.21 -> 8.21Improved
train with VISEQwen3-VL-2B · POPEaccuracy: 89.01 -> 90.03Improved
train with self-consistency self-evolution (EvoLMM)Qwen3-VL-2B · COCO CaptionCIDEr: 21.54 -> 20.84Hurt
train with EvoLMMQwen3-VL-2B · OK-VQAaccuracy: 40.76 -> 38.03Hurt
apply VISE to larger modelsVISE · COCO CaptionCIDEr gain: +16.85 at 2B to +8.72 at 32BMixed
apply VISE to four backbone familiesVISE · COCO CaptionCIDEr: Gemma3-12B 22.18 -> 29.83; Llama-3.2-11B 16.37 -> 22.81Improved
add semantic invariance on 8B model (ablation)VISE on Qwen3-VL-8B · COCO CaptionCIDEr: 29.01 -> 35.27 (R_sem only)Improved

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