Breaking Frozen Priors: Teaching Language Models to Discover Conservation Laws from Numerical Simulation
Breaking Frozen Priors: Teaching Language Models to Discover Conservation Laws from Numerical Simulation
Language models exhibit frozen priors — uniform rejection patterns that prevent recognition of valid conservation laws. We demonstrate a three-phase pipeline (diagnose → break → supervise) that transforms a frozen classifier into a physics-aware ranking engine, achieving Spearman rho = 0.893 on held-out conservation laws. The pipeline has been validated across 24 domains (275/275 facts, 100%) spanning physics, genetics, and mathematics.
automated scientific discovery, frozen priors, LLM evaluation, Kirchhoff equations, logit-space adapters, conservation laws, point-vortex dynamics, vortex dynamics, physics-informed AI
automated scientific discovery, frozen priors, LLM evaluation, Kirchhoff equations, logit-space adapters, conservation laws, point-vortex dynamics, vortex dynamics, physics-informed AI
selected citations These citations are derived from selected sources.This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).0 popularity This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.Average influence This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).Average impulse This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.Average
Citations provided by BIP!Popularity provided by BIP!