New benchmark tests how well AI models discover physics

Original: Very excited to release DiscoverPhysics, a new benchmark and evaluation pipeline for experimentation and discovery in LLMs.

Source: sampsonml.github.io ↗