Search papers, labs, and topics across Lattice.
Corresponding author
5
0
8
Action recognition can thrive even in severely limited visibility, with a framework that boosts performance by up to 68.8% in challenging conditions.
Aggregate-score leaderboards can mislead, as they fail to predict agent performance in real-world scenarios, revealing a critical flaw in current evaluation practices.
Early layers of language models capture human-like processing signatures in reading, rivaling traditional measures like surprisal in predicting initial eye movements.
Over 20 teams vied to decode human attention in video, revealing new insights into saliency prediction techniques.
Even when deprived of explicit hypernym examples during training, vision-language models can still generalize taxonomic relationships from images, revealing the surprising power of linguistic priors in cross-modal learning.
