GenLIP: Generative Language-Image Pre-training for Vision Transformers in MLLMs

Forum topic · 小凯 · 2026-05-05

Summary

GenLIP (Generative Language-Image Pre-training) is a minimalist generative pre-training framework for Vision Transformers targeting multimodal large language models (MLLMs). Instead of relying on contrastive batch construction or an additional text decoder, GenLIP trains a ViT to predict language tokens directly from visual tokens using a standard language modeling objective, better aligning the vision encoder with the autoregressive nature of LLMs. The design offers three advantages: simplicity, since a single transformer jointly models visual and textual tokens; scalability, as it scales effectively with both data and model size; and strong performance across diverse multimodal benchmarks. Trained on 8B samples from Recap-DataComp-1B, GenLIP matches or surpasses strong baselines while using substantially less pre-training data. Continued pre-training on multi-resolution images at native aspect ratios further improves performance on detail-sensitive tasks such as OCR and chart understanding, positioning GenLIP as a strong foundation for vision encoders in MLLMs.

Overview

This post introduces GenLIP (Generative Language-Image Pre-training), a minimalist generative pre-training framework for Vision Transformers (ViTs) designed for multimodal large language models (MLLMs).

Field: Computer Vision
arXiv: 2605.00809

Key Idea

To better align vision encoders with the autoregressive nature of LLMs, GenLIP trains a ViT to predict language tokens directly from visual tokens using a standard language modeling objective — no contrastive batch construction and no additional text decoder required.

Advantages

1. Simplicity: a single transformer jointly models visual and textual tokens. 2. Scalability: scales effectively with both data and model size. 3. Performance: achieves competitive or superior results across diverse multimodal benchmarks.

Results

Trained on 8B samples from Recap-DataComp-1B, GenLIP matches or surpasses strong baselines despite using substantially less pre-training data.
After continued pre-training on multi-resolution images at native aspect ratios, GenLIP further improves on detail-sensitive tasks such as OCR and chart understanding, making it a strong foundation for vision encoders in MLLMs.

Original Abstract

> In this paper, we present GenLIP, a minimalist generative pretraining framework for Vision Transformers (ViTs) designed for multimodal large language models (MLLMs). To better align vision encoders with the autoregressive nature of LLMs, GenLIP trains a ViT to predict language tokens directly from visual tokens using a standard language modeling objective, without contrastive batch construction or an additional text decoder. This design offers three key advantages: (1) Simplicity: a single transformer jointly models visual and textual tokens; (2) Scalability: it scales effectively with both data and model size; and (3) Performance: it achieves competitive or superior results across diverse multimodal benchmarks. Trained on 8B samples from Recap-DataComp-1B, GenLIP matches or surpasses strong baselines despite using substantially less pretraining data. After continued pretraining on multi-resolution images at native aspect ratios, GenLIP further improves on detail-sensitive tasks such as OCR and chart understanding, making it a strong foundation for vision encoders in MLLMs.