AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Mixture of Specialized Vision Experts: Unlocking Complementary Visual Insights for Faithful MLLM Reasoning

Yifei Gao, Liangliang You, Jiye Xie, changwei wang, Kexue Fu, Jingyi Liu, Rongtao Xu, Zhiqiang Kou, Haoran Xu, Longxiang Gao, Yu Zhang

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Instruction-tuned multimodal assistants · 0 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 query-guided adaptive threshold instead of fixed threshold (ablation)MoSVE · MMVPaccuracy: 48 -> 57Improved
add query-guided adaptive threshold instead of fixed threshold (ablation)MoSVE · CHAIRCHAIR_S: 34.0 -> 27.1Improved
add query-guided adaptive threshold instead of fixed threshold (ablation)MoSVE · POPEaccuracy: 85.1 -> 86.9Improved
cluster tokens instead of feeding all 1536 tokens (ablation)MoSVE · MMVPaccuracy: 59 -> 57Mixed
add DINOv2 expert to CLIP (ablation)LLaVA-style MLLM · MMVPaccuracy: 22 -> 38Improved
add SAM-L expert to CLIP (ablation)LLaVA-style MLLM · MMVPaccuracy: 22 -> 36Improved
add SAM-L to CLIP+DINOv2 (ablation)LLaVA-style MLLM · MMVPaccuracy: 38 -> 42Improved
add CRFT to CLIP+DINOv2+SAM-L (ablation)MoSVE · MMVPaccuracy: 42 -> 57Improved
add CRFT to CLIP+DINOv2+SAM-L (ablation)MoSVE · CHAIRCHAIR_S: 36.8 -> 27.1Improved
add complementary routing instead of random routing (ablation)CRFT · MMVPaccuracy: 50 -> 57Improved
add complementary routing instead of standard SFT (ablation)CRFT · MMVPaccuracy: 54 -> 57Improved
add complementary routing instead of standard SFT (ablation)CRFT · CHAIRCHAIR_S: 29.2 -> 27.1Improved

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