A Strong Baseline for Evaluating Vision Encoders in Multimodal Large Language Models
cs.CV, cs.AI
Submitted: 2026-10-04
Updated: 2026-10-04
Comments: Code is available at https://github.com/JuntaoTang/MLLM-VisionEncoder-Eval
Code: https://github.com/JuntaoTang/MLLM-VisionEncoder-Eval
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- Aggregating Deep Convolutional Features for Image Retrieval
- Training Deep Nets with Sublinear Memory Cost
- Gaussian Error Linear Units (GELUs)
- A Study of BFLOAT16 for Deep Learning Training
- Are Bigger Encoders Always Better in Vision Large Models?
- TokLIP: Marry Visual Tokens to CLIP for Multimodal Comprehension and Generation
- All-but-the-Top: Simple and Effective Postprocessing for Word Representations
- DINOv3
- Improved Baselines with Representation Autoencoders
- Whitening Sentence Representations for Better Semantics and Faster Retrieval
- SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
- TokBench: Evaluating Your Visual Tokenizer before Visual Generation
- Qwen2.5 Technical Report
- Qwen3 Technical Report
- Efficient Universal Perception Encoder
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