Бенчмарк визуального диалога
material
ID: visualdialogbenchmark
Related Papers
Enriching Language Models with Visually-grounded Word Vectors and the Lancaster Sensorimotor Norms
AbstractLanguage models are trained only on text despite the fact that humans learn their first language in a highly interactive and multimodal environment where the first set of learned words …...