WRSWorld Research StateContribute
WORLD RESEARCH STATE

Stochastic interpolants: variational velocity and score learning

Contribution grade

Continuous generative modeling · Human-designed research method or resource

Reference score155,000Provisional editorial estimate of this exact contribution; its grade is assessed separately.

What changed

Specified paths between endpoint samples and learned probability-current velocities through quadratic regression objectives; extended the construction with latent noise to relate deterministic flows, scores and tunable diffusions.

Scientific scope

Specified paths between endpoint samples and learned probability-current velocities through quadratic regression objectives; extended the construction with latent noise to relate deterministic flows, scores and tunable diffusions. Population identities assume the stated density, smoothness and integrability conditions; they do not give exact finite-network transport. Closely related Flow Matching and rectified-flow work retain separate credited precedence, not new invention here. Arbitrary interpolants are not automatically optimal transport or Schrödinger bridges; further optimization is needed.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Population identities assume the stated density, smoothness and integrability conditions; they do not give exact finite-network transport. Closely related Flow Matching and rectified-flow work retain separate credited precedence, not new invention here. Arbitrary interpolants are not automatically optimal transport or Schrödinger bridges; further optimization is needed.

Contributors

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction