NewtonBench: Benchmarking Generalizable Scientific Law Discovery in LLM Agents

Submitted to Arxiv, 2025