ViCToR: Improving Visual Comprehension via Token Reconstruction for Pretraining LMMs

Large Multimodal Models (LMMs) often face a modality representation gap during pretraining: while language embeddings remain stable, visual representations are highly sensitive to contextual noise (e.g., background clutter). To address this issue, we introduce a visual comprehension stage, which we call ViCToR (Visual Comprehension via Token Reconstruction), a novel pretraining framework for LMMs. ViCToR employs a learnable visual token pool and utilizes the Hungarian matching algorithm to select semantically relevant tokens from this pool for visual token replacement. Furthermore, by integrating a visual token reconstruction loss with dense semantic supervision, ViCToR can learn tokens which retain high visual detail, thereby enhancing the large language model's (LLM's) understanding of visual information. After pretraining on 3 million publicly accessible images and captions, ViCToR achieves state-of-the-art results, improving over LLaVA-NeXT-8B by 10.4%, 3.2%, and 7.2% on the MMStar, SEEDI, and RealWorldQA benchmarks, respectively.

Authors

• Yin Xie
• Kaicheng Yang DeepGlint
• Peirou Liang University of Science and Technology of China
• Xiang An DeepGlint
• Yongle Zhao DeepGlint
• Yumeng Wang DeepGlint
• Ziyong Feng DeepGlint
• Roy Miles Imperial College London
• Ismail Elezi Huawei Noah's Ark Lab, London, UK
• Jiankang Deng Imperial College London

Keywords

No keywords are indexed for this paper.