Random chance baseline

Reported on 6 benchmarks across 3 tasks · 3 papers

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing6 results

Question AnsweringonCOPA
Accuracy· 2021-04-16
50
best: 100 (PaLM 540B (finetuned) )
Back to Square One: Artifact Detection, Training and Commonsense Disentanglement in the Winograd Schema arXiv:2104.08161
Question AnsweringonPIQA
Accuracy· 2021-04-16
50
best: 90.1 (Unicorn 11B (fine-tuned))
Back to Square One: Artifact Detection, Training and Commonsense Disentanglement in the Winograd Schema arXiv:2104.08161
Coreference ResolutiononWinograd Schema Challenge
Accuracy· 2021-04-16
50
best: 100 (PaLM 540B (fine-tuned))
Back to Square One: Artifact Detection, Training and Commonsense Disentanglement in the Winograd Schema arXiv:2104.08161
Question AnsweringonOpenBookQA
Accuracy· 2019-05-19
25
best: 95.9 (GPT-4 + knowledge base)
HellaSwag: Can a Machine Really Finish Your Sentence?arXiv:1905.07830
Sentence CompletiononHellaSwag
Accuracy· 2019-05-19
25
best: 96.1 (CompassMTL 567M with Tailor)
HellaSwag: Can a Machine Really Finish Your Sentence?arXiv:1905.07830
Question AnsweringonSIQA
Accuracy· 2019-04-22
33.3
best: 83.2 (Unicorn 11B (fine-tuned))
SocialIQA: Commonsense Reasoning about Social Interactions arXiv:1904.09728