Skip to content

Evaluation

Closed-book question answering (NaturalQuestions, TriviaQA)

Exact-match performance on NaturalQuestions and TriviaQA in a closed-book setting (no evidence documents)

Core capabilities Meta

How it was run

Open a release to see the setup, scoring, and other details that the lab published. Only fields the lab actually disclosed are shown.

Llama 1
Meta

partial detail 12 method fields published

See how this test was run