GulliBench: Measuring Skepticism in Frontier Models

2 comments

• dvaplima 52 minutes ago
  I liked approach to evaluating AI behavior, the fact that additional reasoning doesn’t improve performance is quite interesting!
• Betaantunes11 27 minutes ago
  interesting approach