Mufan Li

Assistant Professor, University of Waterloo

1 paper at NeurIPS 2025

Homepage· OpenReview· Semantic Scholar· Google Scholar

Poster Session 2

Wednesday, December 3, 2025 · 4:30 PM → 7:30 PM

Don't be lazy: CompleteP enables compute-efficient deep transformers

#4000 · Nolan Simran Dey, Bin Claire Zhang, Lorenzo Noci, Mufan Li, Blake Bordelon, Shane Bergsma, Cengiz Pehlevan, Boris Hanin, Joel Hestness

We introduce CompleteP, which offers depth-wise HP transfer, FLOP savings when training deep models, and a larger range of compute-efficient width/depth ratios.