Elite-Weighted Supervised Fine-tuning for Goal-Directed Molecular Optimization
We introduce Elite-Weighted Supervised Fine-tuning (EW-SFT), which passes reward through elite selection and updates the model with its own pretraining loss, yielding a single optimizer that transfers across autoregressive, masked-diffusion, and discrete-flow molecular generators.
Aug 31, 2026
COVID-19 Trend Forecasting Using State-level Mobility and Policy
We estimated the transmission rates of COVID-19 via robust regression on local mobility statistics as well as local policies.
Jan 6, 2021