List your company
ResearchFormerly WizWand

Natural Language Inference on ANLI Round 2

66.5AccuracyLMSI
29.37239.01148.6558.289Apr 14, 2022Aug 20, 2022Dec 26, 2022May 4, 2023Sep 9, 2023Jan 15, 2024May 23, 2024
Updated 8mo ago

Evaluation results

MethodLinks
LMSIPrompting Method=Self-Consistency2022.1066.5
LMSIPrompting Method=CoT-Prompting2022.1065.3
Wang et al. (2022a)description=Previous SOTA2022.1064.9
LMSIPrompting Method=Standard-Prompting2022.1064.8
Self-ConsistencyLMSI=false2022.1064.5
FLAN-T5Model Variant=xlarge, Model Parameters=3B2023.0760.6
CoT-PromptingLMSI=false2022.1058.9
Standard-PromptingLMSI=false2022.1055.8
ALIGNModel Variant=large, Model Parameters=355M2023.0752.4
Ties-MergingValidation=true2024.0551.3
Traditional MTL2024.0549.8
Task ArithmeticValidation=true2024.0549.4
ALIGNModel Variant=base, Model Parameters=125M2023.0748.7
FLAN-T5Model Variant=large, Model Parameters=780M2023.0748.7
Task ArithmeticValidation=false2024.0547.5
Individual2024.0546.5
Ties-MergingValidation=false2024.0546.5
FLAN 137BEvaluation protocol=0-shot, Training strategy=leave-one-category-out2022.1243.9
OPT-IML 175BEvaluation protocol=few-shot, k=52022.1243.8
EMR-MERGINGValidation=false2024.0543.8
OPT-IML 175BEvaluation protocol=0-shot2022.1243.5
FLAN 137BEvaluation protocol=few-shot, k=variable, Training strategy=leave-one-category-out2022.1241.6
Fisher MergingValidation=true2024.0541
T5(3B) + PE w/ ROE (ORC.)Backbone=T5 (3B), Expert type=Oracle Retrieval-of-Expert, Additional Parameters=100M2023.0240.11
LaMDA-PT 137BEvaluation protocol=0-shot2022.1239.9
RegMeanValidation=true2024.0539.2
Weight AveragingValidation=false2024.0539.2
T0-11BParameters=11B2023.0238.68
GPT-3Model Variant=DaVinci, Zero-shot=true2022.0437.5
LaMDA-PT 137BEvaluation protocol=few-shot, k=variable2022.1237.5
OPT-IML 30BEvaluation protocol=0-shot2022.1237.5
OPT-IML 30BEvaluation protocol=few-shot, k=52022.1237
T5(3B) + Cos PEBackbone=T5 (3B), Expert type=Prompt Expert trained on COSMOS-QA, Additional Parameters=100M2023.0236.11
GPT-3 (175B)Parameters=175B2023.0235.4
FairSeqNumber of Parameters=355M, Evaluation Protocol=5-shot2022.0435
OPT 175BEvaluation protocol=few-shot, k=52022.1235
FairSeqNumber of Parameters=1.3B, Evaluation Protocol=5-shot2022.0434.7
T5(3B) + PE w/ ROEBackbone=T5 (3B), Expert type=Retrieval-of-Expert, Additional Parameters=100M2023.0234.64
FairSeqNumber of Parameters=125M, Evaluation Protocol=5-shot2022.0434.5
BloombergGPTEvaluation Protocol=1-shot2023.0334.4
GPT-NeoXModel Size=20B, Zero-shot=true2022.0434.3
GPT-3Model Variant=Ada, Zero-shot=true2022.0434.2
OPT-66BEvaluation Protocol=1-shot2023.0334.2
FairSeqNumber of Parameters=6.7B, Evaluation Protocol=5-shot2022.0434
GPT-JModel Size=6B, Zero-shot=true2022.0434
FairSeqNumber of Parameters=2.7B, Zero-Shot=true2022.0433.9
GPT-3Evaluation Protocol=1-shot2023.0333.9
FairSeqNumber of Parameters=13B, Evaluation Protocol=5-shot2022.0433.8
GPT-3Model Variant=Curie, Zero-shot=true2022.0433.8
GPT-NeoXEvaluation Protocol=1-shot2023.0333.8
BLOOM-176BEvaluation Protocol=1-shot2023.0333.8
FairSeqNumber of Parameters=125M, Zero-Shot=true2022.0433.6
OPT 30BEvaluation protocol=few-shot, k=52022.1233.6
FairSeqNumber of Parameters=1.3B, Zero-Shot=true2022.0433.4
FairSeqNumber of Parameters=2.7B, Evaluation Protocol=5-shot2022.0433.3
OPT 30BEvaluation protocol=0-shot2022.1233.3
OPT 175BEvaluation protocol=0-shot2022.1233.3
T0-3BParameters=3B2023.0233.27
GPT-JParameters=6B, Evaluation=Five-shot2022.0433.1
FairSeqNumber of Parameters=13B, Zero-Shot=true2022.0433
GPT-NeoXParameters=20B, Evaluation=Five-shot2022.0432.9
FairSeqNumber of Parameters=6.7B, Zero-Shot=true2022.0432.2
FairSeqNumber of Parameters=355M, Zero-Shot=true2022.0431.2
GPT-3Model Variant=Babbage, Zero-shot=true2022.0430.8