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

2025/02/10 by Diao, Haiwen, Li, Xiaotong, Cui, Yufeng +6 · 10 citations
#Artificial Intelligence (cs.AI) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2502.06788

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.

Cited by

• VAT: Vision Action Transformer by Unlocking Full Representation of ViT
• End-to-end Listen, Look, Speak and Act
• From Pixels to Words -- Towards Native Vision-Language Primitives at Scale
• NaViL: Rethinking Scaling Properties of Native Multimodal Large Language Models under Data Constraints
• OBS-Diff: Accurate Pruning For Diffusion Models in One-Shot
• Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training
• Catalog-Native LLM: Speaking Item-ID Dialect with Less Entanglement for Recommendation
• OneCAT: Decoder-Only Auto-Regressive Model for Unified Understanding and Generation
• MoE-TTS: Enhancing Out-of-Domain Text Understanding for Description-based TTS via Mixture-of-Experts
• Optimizing Vision-Language Consistency via Cross-Layer Regional Attention Alignment

Related

• Unveiling Encoder-Free Vision-Language Models
• Eve: Efficient Multimodal Vision Language Models with Elastic Visual Experts
• BRAVE: Broadening the visual encoding of vision-language models
• Optimizing Vision-Language Interactions Through Decoder-Only Models
• What matters when building vision-language models?