factual
Test-time compute scaling shows consistent performance gains
When reasoning models' performance is plotted against test-time compute, performance on basically any reasoning benchmark rises clearly the longer the model thinks, analogous to people performing better on the SAT with five hours than with five minutes.
factualpending
Speaker
Noam BrownEvidence Quote
“you see a very clear pattern where the longer these models take to think about their answer, the better they do.”
Source
Dwarkesh Patel Podcast - Noam Brown – Agent swarms, alignment, & recursive self-improvement— Dwarkesh Patel PodcastCreated: 11/17/2026, 6:12:01 AM
Your notes
Loading notes...