PhD student, KAIST
2 papers at NeurIPS 2025
Inference-time scaling for flow models
We propose a novel regularization loss that enforces standard Gaussianity, encouraging samples to align with a standard Gaussian distribution.