CoFFT: Chain of Foresight-Focus Thought for Visual Language Models

Xinyu Zhang, Yuxuan Dong, Lingling Zhang +5

Despite significant advances in Vision Language Models (VLMs), they remain constrained by the complexity and redundancy of visual input. When images contain large amounts of irrele…