Reproduce “COSMOS: Soft Mechanism Mixtures with Verifiable Routing for Long-Horizon PDE Forecasting” 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
Anupam Rawat, Manikandan Padmanaban, Jagabondhu Hazra
COSMOS tests whether dense soft routing improves PDE forecasting for mixtures of physical processes. It combines outputs from four process-biased specialists at every step and adds a full-resolution residual to recover detail lost in a coarse latent. On a synthetic compositional holdout, soft routing achieves roll@20 0.137 ± 0.065, versus hard top-1 routing’s 0.301 ± 0.006. The external diffusion–reaction audit found no dependable accuracy advantage.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.