paper-with-me

Papers

Reconstructive Visual Instruction Tuning

2024-10-12 · Haochen Wang, Anlin Zheng, Yucheng Zhao, Tiancai Wang, Zheng Ge, Xiangyu Zhang, Zhaoxiang Zhang

This paper introduces reconstructive visual instruction tuning (ROSS), a family of Large Multimodal Models (LMMs) that exploit vision-centric supervision signals. In contrast to conventional visual instruction tuning approaches that exclusively supervise text outputs, ROSS prompts LMMs to supervise visual outputs via reconstructing input images. By doing so, it capitalizes on the inherent richness and detail present within input images themselves, which are often lost in pure text supervision. However, producing meaningful feedback from natural images is challenging due to the heavy spatial redundancy of visual signals. To address this issue, ROSS employs a denoising objective to reconstruct latent representations of input images, avoiding directly regressing exact raw RGB values. This intrinsic activation design inherently encourages LMMs to maintain image detail, thereby enhancing their fine-grained comprehension capabilities and reducing hallucinations. Empirically, ROSS consistently brings significant improvements across different visual encoders and language models. In comparison with extrinsic assistance state-of-the-art alternatives that aggregate multiple visual experts, ROSS delivers competitive performance with a single SigLIP visual encoder, demonstrating the efficacy of our vision-centric supervision tailored for visual outputs.

📄 PDF Abstract BibTeX arXiv:2410.09575

Code (1)

haochen-wang409/ross pytorch

Tasks

Denoising

Similar Papers 제목 키워드 기반

Ross3D: Reconstructive Visual Instruction Tuning with 3D-Awareness

2025-04-02 · Haochen Wang, Yucheng Zhao, Tiancai Wang, Haoqiang Fan 외

The rapid development of Large Multimodal Models (LMMs) for 2D images and videos has spurred efforts to adapt these models for interpreting 3D scenes. However, the absence of large-scale 3D vision-language datasets has p…

Scene Understanding

VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction

2025-05-26 · Zhiwen Fan, Jian Zhang, Renjie Li, Junge Zhang 외

The rapid advancement of Large Multimodal Models (LMMs) for 2D images and videos has motivated extending these models to understand 3D scenes, aiming for human-like visual-spatial intelligence. Nevertheless, achieving de…

3D ReconstructionSpatial Reasoning

Reg3D: Reconstructive Geometry Instruction Tuning for 3D Scene Understanding

2025-09-03 · Hongpei Zheng, Lintao Xiang, Qijun Yang, Qian Lin 외 arxiv

The rapid development of Large Multimodal Models (LMMs) has led to remarkable progress in 2D visual understanding; however, extending these capabilities to 3D scene understanding remains a significant challenge. Existing…

Scene UnderstandingSpatial Reasoning

Toward Graph-Tokenizing Large Language Models with Reconstructive Graph Instruction Tuning

2026-03-02 · Zhongjian Zhang, Xiao Wang, Mengmei Zhang, Jiarui Tan 외 arxiv

The remarkable success of large language models (LLMs) has motivated researchers to adapt them as universal predictors for various graph-related tasks, with the ultimate goal of developing a graph foundation model that g…

Video-XL-Pro: Reconstructive Token Compression for Extremely Long Video Understanding

2025-03-24 · Xiangrui Liu, Yan Shu, Zheng Liu, Ao Li 외

Despite advanced token compression techniques, existing multimodal large language models (MLLMs) still struggle with hour-long video understanding. In this work, we propose Video-XL-Pro, an efficient method for extremely…

8kGPUSelf-Supervised LearningVideo Understanding