A machine–to–corpus–to–machine pipeline where knowledge is transferred from a large, general model to a compact commonsense model, through a commonsense corpus.
A model that learns real-world conversation from a large-scale video-based dialogue dataset.
A novel framework on inductive knowledge distillation that combines NeuroLogic decoding and self-imitation learning.
On the limits of social intelligence in large language models.
Contextual reasoning about effects and harms of offensive statements.
On the limits of transformers on compositionality.
An interactive system that learns to ask clarification questions in order to elicit additional salient contexts of a social or moral situation.
An open, billion-scale corpus of images interleaved with text.
Million-scale dialogue dataset with social commonsense contextualization.
A reinforced knowledge introspector, that learns to generate contextually relevant knowledge in response to given questions.
To measure progress in machine common sense, we're developing a suite of benchmark datasets.
Knowledge graphs provide useful semi-structured representations of commonsense. Currently, we're developing and exploring how to better use such graphs with current models.
Photo credits: AI image generator available at hotpot.ai/art-generator.