UNICORN on RAINBOW: A Universal Commonsense Reasoning Model on a New Multitask Benchmark

Benchmark Model Rank Results
common-sense-reasoning-on-commonsenseqaUnicorn 11B (fine-tuned)#6Accuracy: 79.3
common-sense-reasoning-on-winograndeUnicorn 11B (fine-tuned)#2Accuracy: 91.3
question-answering-on-piqaUnicorn 11B (fine-tuned)#1Accuracy: 90.1
question-answering-on-social-iqaUnicorn 11B (fine-tuned)#1Accuracy: 83.2
sentence-completion-on-hellaswagUnicorn 11B (fine-tuned)#8Accuracy: 93.9