Mitigating Coordinate Prediction Bias from Positional Encoding Failures

Xingjian Tao, Yiwei Wang, Yujun Cai +3

While Multimodal Large Language Models (MLLMs) excel at general vision-language tasks, precise coordinate prediction remains a significant challenge, particularly as high-resolutio…