RE-Bench: Evaluating frontier AI R&D capabilities of language model agents against human experts

Open in new window