PhD student, University of California, Los Angeles
1 paper at NeurIPS 2025
ADMN enables layer-wise allocation of resources among modalities at inference time according to modality quality for a fixed resource budget.
