List your company

Natural Language Inference on WNLI (test)

89AccuracyHNNensemble
41.88854.11966.3578.581Jul 27, 2019May 15, 2020Mar 5, 2021Dec 24, 2021Oct 14, 2022Aug 4, 2023May 24, 2024

Evaluation results

MethodLinks
HNNensembleensemble=six models2019.0789
HNN2019.0783.6
BERT_CS_largeFormat=MCQA2019.0883.6
BERT_largeFormat=MCQA2019.0878.5
BERTWiki-WSCR2019.0771.9
Kocijan et al. (2019)2019.0871.9
BERTWSCR2019.0770.5
BERT_largeTraining Objective=MTP2019.0870.5
BERTLARGE-LM2019.0765.1
FLAN-T5Model Variant=large, Parameters=780M, Evaluation Protocol=Zero-shot2023.0762
FedAvgProportion of Attackers=0%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0561.9
BREAProportion of Attackers=0%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0561.9
RFLPAProportion of Attackers=0%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0561.9
BREAProportion of Attackers=10%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0559.2
RFLPAProportion of Attackers=10%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0559.2
RFLPAProportion of Attackers=20%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0559.2
BREAProportion of Attackers=20%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0557.7
FedAvgProportion of Attackers=10%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0556.3
BREAProportion of Attackers=30%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0556.3
RFLPAProportion of Attackers=30%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0556.3
ALIGNModel Variant=large, Parameters=355M, Evaluation Protocol=Zero-shot2023.0754.9
ALIGNModel Variant=base, Parameters=125M, Evaluation Protocol=Zero-shot2023.0752.1
FLAN-T5Model Variant=base, Parameters=250M, Evaluation Protocol=Zero-shot2023.0746.5
FedAvgProportion of Attackers=20%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0543.7
FedAvgProportion of Attackers=30%, Backbone=DistilBERT, Attack Type=Gradient Manipulation2024.0543.7