GulliBench: Measuring Skepticism in Frontier Models

18 points by rigelbm 65 days ago | 2 comments

dvaplima 65 days ago |

I liked approach to evaluating AI behavior, the fact that additional reasoning doesn’t improve performance is quite interesting!

Betaantunes11 65 days ago |

interesting approach

dvaplima 65 days ago |

I liked approach to evaluating AI behavior, the fact that additional reasoning doesn’t improve performance is quite interesting!

Betaantunes11 65 days ago |

interesting approach