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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add query-guided adaptive threshold instead of fixed threshold (ablation) | MoSVE · MMVP | accuracy: 48 -> 57 | Improved |
| add query-guided adaptive threshold instead of fixed threshold (ablation) | MoSVE · CHAIR | CHAIR_S: 34.0 -> 27.1 | Improved |
| add query-guided adaptive threshold instead of fixed threshold (ablation) | MoSVE · POPE | accuracy: 85.1 -> 86.9 | Improved |
| cluster tokens instead of feeding all 1536 tokens (ablation) | MoSVE · MMVP | accuracy: 59 -> 57 | Mixed |
| add DINOv2 expert to CLIP (ablation) | LLaVA-style MLLM · MMVP | accuracy: 22 -> 38 | Improved |
| add SAM-L expert to CLIP (ablation) | LLaVA-style MLLM · MMVP | accuracy: 22 -> 36 | Improved |
| add SAM-L to CLIP+DINOv2 (ablation) | LLaVA-style MLLM · MMVP | accuracy: 38 -> 42 | Improved |
| add CRFT to CLIP+DINOv2+SAM-L (ablation) | MoSVE · MMVP | accuracy: 42 -> 57 | Improved |
| add CRFT to CLIP+DINOv2+SAM-L (ablation) | MoSVE · CHAIR | CHAIR_S: 36.8 -> 27.1 | Improved |
| add complementary routing instead of random routing (ablation) | CRFT · MMVP | accuracy: 50 -> 57 | Improved |
| add complementary routing instead of standard SFT (ablation) | CRFT · MMVP | accuracy: 54 -> 57 | Improved |
| add complementary routing instead of standard SFT (ablation) | CRFT · CHAIR | CHAIR_S: 29.2 -> 27.1 | Improved |
Closest papers by meaning
- MoVA: Adapting Mixture of Vision Experts to Multimodal Context (NeurIPS 2024 · similarity 0.91)
- Where Does Vision Meet Language? Understanding and Refining Visual Fusion in MLLMs via Contrastive Attention (CVPR 2026 · similarity 0.90)
- Perception Tokens Enhance Visual Reasoning in Multimodal Language Models (CVPR 2025 · similarity 0.90)
- VisuRiddles: Fine-grained Perception is a Primary Bottleneck for Multimodal Large Language Models in Abstract Visual Reasoning (ICLR 2026 · similarity 0.90)
- Eagle: Exploring The Design Space for Multimodal LLMs with Mixture of Encoders (ICLR 2025 · similarity 0.90)
- Lifting the Veil on Visual Information Flow in MLLMs: Unlocking Pathways to Faster Inference (CVPR 2025 · similarity 0.90)
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.
