Search papers, labs, and topics across Lattice.
University of Illinois at Urbana-Champaign, Urbana, IL, United States
2
0
4
Polynomial preconditioning can significantly enhance LLM training stability and performance without adding inference overhead.
Forget slow, fine-grained vision-language models: T-REN slashes token counts by 24x (images) and 187x (videos) while boosting performance on segmentation, retrieval, and localization.
