Track genomics, neuroscience and molecular biology.
Search papers, labs, and topics across Lattice.
University of Illinois at Urbana-Champaign
2
1
5
3
A single global topology predictor can unlock substantial performance gains in Transformers, outperforming layer-wise routing and achieving better results with less data.
Current video generation models struggle with visual reasoning, with the best achieving only 51% accuracy on a new benchmark designed to probe their capabilities.