List your company

Bias Defense on Average of four datasets (test)

92.38AccuracyNo Defense (Before)
84.413686.481888.5590.6182May 8, 2023

Evaluation results

MethodLinks
No Defense (Before)Backbone=RoBERTa, Attack Type=Bias Word2023.0592.3843.75
Before (No Defense)Backbone=ROBERTa, Attack=BiasWord2023.0592.3843.75
No Defense (Before)Backbone=RoBERTa, Attack Type=Average2023.0592.2643.75
NADBackbone=RoBERTa, Attack Type=Bias Word2023.0592.1743.91
No Defense (Before)Backbone=RoBERTa, Attack Type=BiasSent2023.0592.1443.75
Before (No Defense)Backbone=ROBERTa, Attack=BiasSent2023.0592.1443.75
NADBackbone=RoBERTa, Attack Type=Average2023.0592.1343.84
NADBackbone=RoBERTa, Attack Type=BiasSent2023.0592.0843.78
KDBackbone=RoBERTa, Attack Type=Bias Word2023.0592.0143.75
KDBackbone=RoBERTa, Attack Type=Average2023.0591.9743.75
KDBackbone=RoBERTa, Attack Type=BiasSent2023.0591.9343.75
NADBackbone=BERT, Attack Type=BiasSent2023.0591.6644.65
Fine-tuningBackbone=ROBERTa, Attack=BiasSent2023.0591.644.06
No Defense (Before)Backbone=BERT, Attack Type=BiasSent2023.0591.4443.75
Before (No Defense)Backbone=BERT, Attack=BiasSent2023.0591.4443.75
NADBackbone=BERT, Attack Type=Average2023.0591.4244.73
No Defense (Before)Backbone=BERT, Attack Type=Average2023.0591.3543.75
Fine-tuningBackbone=ROBERTa, Attack=BiasWord2023.0591.343.75
No Defense (Before)Backbone=BERT, Attack Type=Bias Word2023.0591.2743.75
Before (No Defense)Backbone=BERT, Attack=BiasWord2023.0591.2743.75
NADBackbone=BERT, Attack Type=Bias Word2023.0591.1844.82
KDBackbone=BERT, Attack Type=BiasSent2023.0591.0343.75
Fine-tuningBackbone=BERT, Attack=BiasWord2023.0590.8443.75
Fine-tuningBackbone=BERT, Attack=BiasSent2023.0590.8343.75
KDBackbone=BERT, Attack Type=Average2023.0590.843.76
KDBackbone=BERT, Attack Type=Bias Word2023.0590.5743.76
RAPBackbone=RoBERTa, Attack Type=Bias Word2023.0589.9888.73
RAPBackbone=RoBERTa, Attack Type=Average2023.0589.5571.67
RAPBackbone=BERT, Attack Type=BiasSent2023.0589.466.47
RAPBackbone=RoBERTa, Attack Type=BiasSent2023.0589.1354.61
RAPBackbone=BERT, Attack Type=Average2023.0589.0977.58
RAPBackbone=BERT, Attack Type=Bias Word2023.0588.7988.69
ONIONBackbone=BERT, Attack Type=BiasSent2023.0587.8258.65
ONIONBackbone=BERT, Attack Type=Average2023.0587.566.89
ONIONBackbone=BERT, Attack Type=Bias Word2023.0587.1275.14
Fine-pruningBackbone=ROBERTa, Attack=BiasWord2023.0587.0964.65
Fine-purifyingBackbone=RoBERTa, Attack Type=BiasSent2023.0586.7184.11
Fine-purifyingBackbone=ROBERTa, Attack=BiasSent2023.0586.7184.11
Fine-pruningBackbone=ROBERTa, Attack=BiasSent2023.0586.6976.43
Fine-purifyingBackbone=RoBERTa, Attack Type=Average2023.0586.5685.2
ONIONBackbone=RoBERTa, Attack Type=Bias Word2023.0586.4276.8
Fine-purifyingBackbone=RoBERTa, Attack Type=Bias Word2023.0586.4286.3
Fine-purifyingBackbone=ROBERTa, Attack=BiasWord2023.0586.4286.3
ONIONBackbone=RoBERTa, Attack Type=Average2023.0586.463.53
ONIONBackbone=RoBERTa, Attack Type=BiasSent2023.0586.3750.26
Fine-pruningBackbone=BERT, Attack=BiasWord2023.0586.0561.57
Fine-mixingBackbone=ROBERTa, Attack=BiasSent2023.0586.0277.73
Fine-mixingBackbone=ROBERTa, Attack=BiasWord2023.0585.9281.79
Fine-purifyingBackbone=BERT, Attack Type=BiasSent2023.0585.6384.03
Fine-purifyingBackbone=BERT, Attack=BiasSent2023.0585.6384.03
Fine-purifyingBackbone=BERT, Attack Type=Average2023.0585.584.55
Fine-pruningBackbone=BERT, Attack=BiasSent2023.0585.4864.38
Fine-purifyingBackbone=BERT, Attack Type=Bias Word2023.0585.3885.06
Fine-purifyingBackbone=BERT, Attack=BiasWord2023.0585.3885.06
Fine-mixingBackbone=BERT, Attack=BiasSent2023.0584.8175.26
Fine-mixingBackbone=BERT, Attack=BiasWord2023.0584.7276.45