PaperBench: Evaluating AI’s Ability to Replicate AI Research

9 months ago 7
We introduce PaperBench, a benchmark evaluating the ability of AI agents to replicate state-of-the-art AI research.
Read Entire Article