TriviaQA (closed-book, filtered dev)
General QA · 2017-05-09
Closed-book TriviaQA on the filtered development set used by Gopher, Chinchilla and LLaMA: the model answers trivia questions without the evidence documents and is scored by exact match against the answer aliases. It is kept apart from the unfiltered test set and from the Wikipedia validation subset, which give different scores for the same model.
Top models (higher is better)
| Model | Score |
|---|---|
| LLaMA 65B | 68.2 |
| LLaMA 33B | 65.1 |
| LLaMA-13B | 56.6 |
| Chinchilla | 55.4 |
| LLaMA 7B | 50.0 |
| Gopher (280B) | 43.5 |