Duke/Princeton Paper: Parallel Small Agent Teams Scale Test-Time Compute Better

A paper from Duke, Princeton, Modal Labs and NYU shows that splitting 100 agents into 10 parallel small teams and keeping the best team's result beats a single large team or a single agent, scaling test-time compute more efficiently with both high mean performance and controlled variance.

2026-11-11 ~ 2026-11-11 · 2 related posts

• Duke and Princeton paper: 10 small parallel agent teams beat best-of-N and one giant team — rohanpaul_ai · 2026-11-11

1 near-duplicate retellings: rohanpaul_ai