作者位次 #2/6ACL 2026Poster· Natural Language Processing
JW - SVD : Bridging the Cross-Modal Mismatch in Post-Training MLLM Compression
Runchao Li · Yao Fu · Mu Sheng · Haotian Yu · Xianxuan Long · Kenneth A. Loparo
摘要
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...