RealUID is a universal distillation framework for matching models that incorporates real-data supervision without GANs and supplies a theoretical foundation covering diffusion, flow matching, bridge matching, and stochastic interpolants.
First, we use (Wang et al., 2023, Lemma 1) which says that, for any two distributions p, q∈ P(R D) and pointx∈R D, we have δDKL(q||p) δq [x] = logq(x)−logp(x) + 1
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Universal Inverse Distillation for Matching Models with Real-Data Supervision (No GANs)
RealUID is a universal distillation framework for matching models that incorporates real-data supervision without GANs and supplies a theoretical foundation covering diffusion, flow matching, bridge matching, and stochastic interpolants.