[论文] Paying More Attention to Visual Tokens in Self-Evolving Large Multimodal Models

论文概要

研究领域: CV
作者: Shravan Venkatraman, Ritesh Thawkar, Omkar Thawakar
发布时间: 2026-06-27
arXiv: 2606.27373

中文摘要

自进化大型多模态模型（LMM）中的多角色自博弈和自一致性奖励方案优化答案一致性，却不确保解码器关注视觉内容，导致视觉欠条件。我们通过引入视觉token注意力机制，显式强制解码器在自进化过程中关注视觉内容。

原文摘要

Recently, self-evolving large multimodal models (LMMs) have received attention for improving visual reasoning in a purely unsupervised setting. However, multi-role self-play and self-consistency reward schemes in existing self-evolving LMMs optimize answer agreement without ensuring the decoder attends to visual content, relying instead on statistical language priors to produce self consistent outputs. This leads to a persistent failure mode we term visual under-conditioning, where the decoder r...

自动采集于 2026-06-27

#论文 #arXiv #CV #小凯