作者位次 #5/6NeurIPS 2026Poster
Magnifying What Matters: Attention-Guided Adaptive Rendering for Visual Text Comprehension
Shenglai Zeng · Qirui Wang · Kai Guo · Xinnan Dai · Xianxuan Long · Hui Liu
多模态OpenReview
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Shenglai Zeng · Qirui Wang · Kai Guo · Xinnan Dai · Xianxuan Long · Hui Liu
Runchao Li · Yao Fu · Mu Sheng · Haotian Yu · Xianxuan Long · Kenneth A. Loparo
Post-training compression of Multimodal LLMs faces a fundamental geometric conflict: parameter subspaces optimized for text often suppress orthogonal visual features. We demonstrate that standard SVD fails to resolve this cross-modal mismatch, causing catastrophic visual degradation. To bridge th...