Papers with Code paper Sep 30

Soft Spatial Reasoning

Large Vision-Language Models (LVLMs) commonly perform spatial reasoning through chain-of-thought (CoT), encoding intermediate reasoning as autoregressive sequences of discrete lang...