Inspect AI pilot evaluating evidence interpretation in synthetic AMR studies, with custom scoring, answer-order comparisons, and token-limit failure analysis. - KyleCard/amr-evidence-audit ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results