AI agent teams reportedly outperform their strongest member across five math and physics benchmarks

A post describing a Stanford and Together AI paper says agent teams learned to divide roles and challenge one another. Across five math and physics benchmarks, they averaged 66.7% accuracy, versus 48.8% for their strongest member alone and 58.7% when that member had the same compute budget. The post says the teams also beat a system that picked the correct answer whenever any member found it independently. These are the authors’ results and await independent replication.