oMeBench is a large-scale, expert-curated benchmark containing over 10,000 annotated mechanistic steps to test LLM chemical reasoning. It utilizes the oMeS dynamic scoring framework to evaluate whether models can maintain consistency across multi-step organic reaction pathways and generate chemically valid intermediates.
HOW THIS AFFECTS YOU
●
researcherYou can use this to measure if your model possesses genuine chemical reasoning rather than pattern matching.
●
healthThis benchmark is critical for evaluating LLMs intended for molecular and reaction design tasks.