Variational Flow Maps: Make Some Noise for One-Step Conditional Generation
Variational Flow Maps (VFM) recasts conditional generation as a problem of learning the right initial noise distribution for a pretrained or jointly trained one-step flow map. An observation-dependent adapter transforms simple noise before the flow map sends it to data space, enforcing the measurement while retaining the learned data prior.
Jul 7, 2026
Flow Matching for Efficient and Scalable Data Assimilation
The ensemble flow filter (EnFF) is a training-free data-assimilation framework that uses flow matching to transform a forecast ensemble into samples from the filtering distribution. Its Monte Carlo flow-field estimator and localized observation guidance avoid model training while retaining the flexibility of generative flow design.
Jun 15, 2026
Flow Matching for Data Assimilation Accepted in SIAM/ASA JUQ
Our ensemble flow filter brings training-free flow matching to efficient, scalable data assimilation.
Jun 15, 2026
🔊 Make Some Noise: Variational Flow Maps for One-Step Conditional Generation
Our ICML 2026 paper learns the right initial noise for fast, calibrated conditional generation, inverse problems, and reward alignment.
Mar 10, 2026