ReRoute aims to predict scientific what-ifs using data and partial physics knowledge

A researcher introducing ReRoute says the framework combines observational data with partial physical knowledge to predict unseen interventions without new controlled experiments. They report up to 31.8% lower prediction error under severe CO₂ distribution shifts in climate emulators and up to 89.6% lower counterfactual error in controlled advection–diffusion experiments. They also describe a causal identification result under explicit structural assumptions, with a machine-checked proof in Lean.