{"id":"paperbench","name":"PaperBench","maker":"OpenAI","url":"https://github.com/openai/frontier-evals","repo":"openai/frontier-evals","summary":"A benchmark that asks agents to replicate the results of recent machine learning research papers.","details":"Agents read a paper and write code to reproduce its experiments, and are graded against rubrics prepared with the paper authors. It is part of OpenAI's public frontier evaluations repository, which is MIT licensed.","open_source":true,"hosted":false,"verified":"2026-11-15","category":"benchmarks","category_name":"Agent benchmarks","page":"https://modelmoot.com/tools/paperbench","reviewed":"2026-11-15","checked":{"at":"2026-11-15T15:07:43.409Z","address_answers":true},"note":"Descriptions are the site's own. Facts under \"checked\" were read automatically from the tool's address and public repository."}