Frontier AI models reportedly beat expert consensus on earnings predictions
A researcher says the test used a point-in-time gate to prevent information leakage.
TLDR
A researcher says Opus 5.5, Fable 5.1 and Astra, using Samaya’s harness, outperformed expert consensus on earnings predictions, including revenue and gross margin. The researcher says the team gave models access to the same information available to experts and found that only the most recent frontier models beat a bias-corrected consensus baseline.
Combined views
655
1 Source, first seen ago
8 likes2 comments7 saves
