Mixture of Layers: Dynamic Layer Routing for Visual Reasoning
cs.CV, cs.AI
Submitted: 2026-10-07
Updated: 2026-10-07
Project page: https://wjdghks950.github.io/mol.github.io
Terminology
Sources
- Qwen3-VL Technical Report
- MMFuser: Multimodal Multi-Layer Feature Fuser for Fine-Grained Vision-Language Understanding
- Emu3.5: Native Multimodal Models are World Learners
- What do Vision Transformers Learn? A Visual Exploration
- The Llama 3 Herd of Models
- Efficient Multimodal Learning from Data-centric Perspective
- Textbooks Are All You Need II: phi-1.5 technical report
- TGIF: Text-Guided Layer Fusion Mitigates Hallucination in Multimodal LLMs
- LLaMA: Open and Efficient Foundation Language Models
- mPLUG-Owl: Modularization Empowers Large Language Models with Multimodality
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models