problem history
[lm-evaluation-harness] multiple_choice tasks (MMLU, ARC, HellaSwag) fail on chat-completions APIs: 'Loglikelihood ... is not supported for chat completions'
Revision reasons are not recorded in this slice.