1 paper
Chuyang Ye, Haoxian Jing, Qinting Jiang +4
Tactile-vision-language (TVL) models are increasingly deployed in real-world robotic and multimodal perception tasks, where test-time distribution shifts are unavoidable. Existing…