作者位次 #2/10ACL 2026Findings· Findings of ACL
T iny A lign: Boosting Lightweight Vision-Language Models by Mitigating Modal Alignment Bottlenecks
Yuanze Hu · Xinyu Wang · Zhichao Yang · Gen Li · Ye Qiu · Zhaoxin Fan · Yifan Sun · wenjun wu · Jin Dong · Xiaotie Deng
摘要
Lightweight Vision-Language Models (VLMs) are indispensable for resource-constrained applications. The prevailing approach to aligning vision and language models involves freezing both the vision encoder and the language model while training small connector modules. However, this strategy heavily...