作者位次 #2/12ACL 2026Poster· Natural Language Processing
U ni C orn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision
Zhen Fang · Ruiyan Han · XinYu Sun · Yuchen Ma · Ziheng Wang · Yu Zeng · Zehui Chen · Lin Chen · Wenxuan Huang · Wei-Jie Xu · Yi Cao · Feng Zhao
摘要
While Unified Multimodal Models (UMMs) have achieved remarkable success in cross-modal comprehension, a significant gap persists in their ability to leverage such internal knowledge for high-quality generation. We formalize this discrepancy as Conduction Aphasia, a phenomenon where models accurat...