Researcher, Benchmark Reviews

• Develop and publish detailed reviews evaluating new AI benchmarks and their methodologies.
• Assess new benchmarks every two weeks, analyzing their methodology and what strong performance implies about AI capabilities.
• Write public-facing reports on benchmarks and update them periodically as new versions release and models progress.
• Examine individual benchmark tasks in detail using coding agents to assist analysis and critical evaluation.

This role helps the AI research community understand the quality and implications of AI benchmarks, promoting rigorous evaluation and informed progress in AI capabilities.

• Strong analytical skills for evaluating AI benchmarks and methodologies.
• Ability to write clear, public-facing reports.
• Experience with coding agents or similar tools for analysis.
• Familiarity with AI research and benchmark design.

Compensation and benefits not specified in the posting.

Epoch AI is an organization focused on analyzing and forecasting trends in AI development, providing research to inform AI safety and governance.