Papers
arxiv:2511.21150

LLaVA-UHD v3: Progressive Visual Compression for Efficient Native-Resolution Encoding in MLLMs

Published on Nov 26
Authors:
,
,
,
,
,
,
,
,

Abstract

Progressive Visual Compression (PVC) integrates into Vision Transformer (ViT) for efficient native-resolution encoding, reducing TTFT in multi-modal large language models (MLLMs) while maintaining performance.

AI-generated summary

Visual encoding followed by token condensing has become the standard architectural paradigm in multi-modal large language models (MLLMs). Many recent MLLMs increasingly favor global native- resolution visual encoding over slice-based methods. To investigate this trend, we systematically compare their behavior on vision-language understanding and attention patterns, revealing that global encoding enhances overall capability but at the expense of greater computational overhead. To address this issue, we present LLaVA-UHD v3, an MLLM centered upon our proposed Progressive Visual Compression (PVC) method, which can be seamlessly integrated into standard Vision Transformer (ViT) to enable efficient native-resolution encoding. The PVC approach consists of two key modules: (i) refined patch embedding, which supports flexible patch-size scaling for fine-grained visual model- ing, (ii) windowed token compression, hierarchically deployed across ViT layers to progressively aggregate local token representations. Jointly modulated by these two modules, a widely pretrained ViT can be reconfigured into an efficient architecture while largely preserving generality. Evaluated across extensive benchmarks, the transformed ViT, termed ViT-UHD, demonstrates competitive performance with MoonViT while reducing TTFT (time-to-first-token) by 2.4x, when developed within an identical MLLM architecture. Building upon ViT-UHD, LLaVA-UHD v3 also achieves competitive performance to Qwen2-VL, while further reducing TTFT by 1.9x. We will release all code and checkpoints to support future research on efficient MLLMs.

Community

Sign up or log in to comment

Models citing this paper 1

Datasets citing this paper 1

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2511.21150 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.