Reproduce “Physics is the Best Teacher: Consistency Learning for Time-Invariant Operators of Chaotic Dynamics” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Lufang Chiang, Jiachen Yao, Thomas Y.L. Lin, Anima Anandkumar
The authors develop a time-invariant evolution operator that predicts a later state from a current state and elapsed time, using governing-equation velocity as a consistency teacher. Across five three-dimensional chaotic systems, Eulerian physics distillation achieves 0.27 versus 0.43 relative endpoint error against self-distillation at five Lyapunov times out of distribution. The evidence is limited to these systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.