VISTAQA: Benchmarking Joint Visual Question Answering and Pixel-Level Evidence

Mozhgan Nasr Azadani, Yimu Wang, Yongpeng Zhu +5

Establishing a clear link between model predictions and the visual evidence that supports them is critical for transparency and reliability in multimodal reasoning, yet current mul…