Modality-Aware Expert Pruning for MoE-Based Multimodal Large Language Models
cs.CV, cs.CL
Submitted: 2026-11-04
Updated: 2026-11-04
Code: https://github.com/nota-github/MAEP
Terminology
Sources
- ShareGPT4V: Improving Large Multi-Modal Models with Better Captions
- Are We on the Right Way for Evaluating Large Vision-Language Models?
- Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
- DeepSeek-V3 Technical Report
- MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models
- GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning
- AirCache: Activating Inter-modal Relevancy KV Cache Compression for Efficient Large Vision-Language Model Inference
- Mixture Compressor for Mixture-of-Experts LLMs Gains More
- MoDES: Accelerating Mixture-of-Experts Multimodal Large Language Models via Dynamic Expert Skipping
- Mixtral of Experts
- See What You Are Told: Visual Attention Sink in Large Multimodal Models
- ChartNet: A Million-Scale, High-Quality Multimodal Dataset for Robust Chart Understanding
- Sub-MoE: Efficient Mixture-of-Expert LLMs Compression via Subspace Expert Merging
- Merge, Then Compress: Demystify Efficient SMoE with Hints from Its Routing Policy
- Mostly Text, Smart Visuals: Asymmetric Text-Visual Pruning for Large Vision-Language Models
- Evaluating Object Hallucination in Large Vision-Language Models
- MoE-LLaVA: Mixture of Experts for Large Vision-Language Models
- Not All Experts are Equal: Efficient Expert Pruning and Skipping for Mixture-of-Experts Large Language Models
- Multi-Cue Adaptive Visual Token Pruning for Large Vision-Language Models
- MergeMoE: Efficient Compression of MoE Models via Expert Output Merging
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs