EVEv2: Improved Baselines for Encoder-Free Vision-Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909702936330240 |
|---|---|
| author | Diao, Haiwen Li, Xiaotong Cui, Yufeng Wang, Yueze Deng, Haoge Pan, Ting Wang, Wenxuan Lu, Huchuan Wang, Xinlong |
| author_facet | Diao, Haiwen Li, Xiaotong Cui, Yufeng Wang, Yueze Deng, Haoge Pan, Ting Wang, Wenxuan Lu, Huchuan Wang, Xinlong |
| contents | Existing encoder-free vision-language models (VLMs) are rapidly narrowing the performance gap with their encoder-based counterparts, highlighting the promising potential for unified multimodal systems with structural simplicity and efficient deployment. We systematically clarify the performance gap between VLMs using pre-trained vision encoders, discrete tokenizers, and minimalist visual layers from scratch, deeply excavating the under-examined characteristics of encoder-free VLMs. We develop efficient strategies for encoder-free VLMs that rival mainstream encoder-based ones. After an in-depth investigation, we launch EVEv2.0, a new and improved family of encoder-free VLMs. We show that: (i) Properly decomposing and hierarchically associating vision and language within a unified model reduces interference between modalities. (ii) A well-designed training strategy enables effective optimization for encoder-free VLMs. Through extensive evaluation, our EVEv2.0 represents a thorough study for developing a decoder-only architecture across modalities, demonstrating superior data efficiency and strong vision-reasoning capability. Code is publicly available at: https://github.com/baaivision/EVE. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_06788 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | EVEv2: Improved Baselines for Encoder-Free Vision-Language Models Diao, Haiwen Li, Xiaotong Cui, Yufeng Wang, Yueze Deng, Haoge Pan, Ting Wang, Wenxuan Lu, Huchuan Wang, Xinlong Computer Vision and Pattern Recognition Artificial Intelligence Existing encoder-free vision-language models (VLMs) are rapidly narrowing the performance gap with their encoder-based counterparts, highlighting the promising potential for unified multimodal systems with structural simplicity and efficient deployment. We systematically clarify the performance gap between VLMs using pre-trained vision encoders, discrete tokenizers, and minimalist visual layers from scratch, deeply excavating the under-examined characteristics of encoder-free VLMs. We develop efficient strategies for encoder-free VLMs that rival mainstream encoder-based ones. After an in-depth investigation, we launch EVEv2.0, a new and improved family of encoder-free VLMs. We show that: (i) Properly decomposing and hierarchically associating vision and language within a unified model reduces interference between modalities. (ii) A well-designed training strategy enables effective optimization for encoder-free VLMs. Through extensive evaluation, our EVEv2.0 represents a thorough study for developing a decoder-only architecture across modalities, demonstrating superior data efficiency and strong vision-reasoning capability. Code is publicly available at: https://github.com/baaivision/EVE. |
| title | EVEv2: Improved Baselines for Encoder-Free Vision-Language Models |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2502.06788 |