EVEv2: Improved Baselines for Encoder-Free Vision-Language Models

TL;DR

EVEv2 introduces an improved encoder-free vision-language model that achieves competitive performance with encoder-based models through novel training strategies and hierarchical modality association, emphasizing efficiency and reasoning.

Contribution

The paper presents a new decoder-only architecture for vision-language models with effective training methods, reducing modality interference and enhancing performance.

Findings

EVEv2 outperforms previous encoder-free models in vision-language tasks.

Hierarchical association of vision and language improves model effectiveness.

Efficient training strategies enable competitive performance with less data.

Abstract

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.