List your company
ResearchFormerly WizWand

Social Commonsense Reasoning on SIQA (test)

83.3AccuracyUL20B
45.75655.50365.2574.997May 10, 2022Jul 21, 2022Oct 1, 2022Dec 13, 2022Feb 23, 2023May 6, 2023Jul 18, 2023
Updated 8mo ago

Evaluation results

MethodLinks
UL20Bmode=supervised finetuning, leaderboard submission=true2022.0583.3
UNICORN 11BModel=UNICORN 11B2022.1083.2
Lourie et al.status=previous state-of-the-art2022.0583.2
UnifiedQA 11BModel=UnifiedQA 11B2022.1081.4
ScoreBackbone=DeBERTa Large2022.1080.18
TEAMBackbone=DeBERTa Large2022.1079.89
ScoreBackbone=RoBERTa Large2022.1076.7
TEAMBackbone=RoBERTa Large2022.1075.96
LLAMA 1Size=65B2023.0752.3
LLAMA 2Size=34B2023.0750.9
LLAMA 2Size=70B2023.0750.7
LLAMA 1Size=13B2023.0750.4
LLAMA 1Size=33B2023.0750.4
LLAMA 2Size=13B2023.0750.3
FalconSize=40B2023.0750.1
MPTSize=30B2023.0748.9
LLAMA 1Size=7B2023.0748.9
MPTSize=7B2023.0748.5
LLAMA 2Size=7B2023.0748.3
FalconSize=7B2023.0747.2