List your company

Backdoor Defense on MNLI (test)

91.8Clean AccuracyDARE
86.0887.56589.0590.535Apr 8, 2025

Evaluation results

MethodLinks
DAREBackbone=RoBERTa-large, Scenario=Combined: InsertSent + Hidden2025.0491.8—90.8—40.265.5
WAGBackbone=RoBERTa-large, Scenario=Combined: InsertSent + Hidden2025.0491.5—36.6—46.941.8
DAREBackbone=RoBERTa-large, Scenario=Combined: InsertSent + LWS2025.0491.4—88.840.2—64.5
DAREBackbone=RoBERTa-large, Scenario=Combined: BadNet + InsertSent2025.0491.311.690.6——51.1
DAREBackbone=RoBERTa-large, Scenario=Combined: InsertSent + Benign2025.0491.3—42.3——42.3
OursBackbone=RoBERTa-large, Scenario=Combined: InsertSent + Hidden2025.0491.1—24.3—56.140.2
OursBackbone=RoBERTa-large, Scenario=Combined: InsertSent + LWS2025.0491—24.862.5—43.7
DAREBackbone=RoBERTa-large, Scenario=Combined: LWS + Hidden2025.0491——41.588.765.1
DAREBackbone=RoBERTa-large, Scenario=Combined: Benign + BadNet2025.049141.8———41.8
TIESBackbone=RoBERTa-large, Scenario=Combined: InsertSent + Hidden2025.0490.9—65.1—55.260.2
DAREBackbone=RoBERTa-large, Scenario=Combined: BadNet + Hidden2025.0490.933.4——81.857.6
DAREBackbone=RoBERTa-large, Scenario=Combined: Hidden + Benign2025.0490.9———63.363.3
WAGBackbone=RoBERTa-large, Scenario=Combined: InsertSent + LWS2025.0490.6—36.162.6—49.4
OursBackbone=RoBERTa-large, Scenario=Combined: BadNet + InsertSent2025.0490.524.822.5——23.7
DAREBackbone=RoBERTa-large, Scenario=Combined: BadNet + LWS2025.0490.571.7—56.4—64.1
OursBackbone=RoBERTa-large, Scenario=Combined: InsertSent + Benign2025.0490.5—18.3——18.3
WAGBackbone=RoBERTa-large, Scenario=Combined: InsertSent + Benign2025.0490.4—23.2——23.2
TIESBackbone=RoBERTa-large, Scenario=Combined: InsertSent + Benign2025.0490.4—40.6——40.6
WAGBackbone=RoBERTa-large, Scenario=Combined: BadNet + InsertSent2025.0490.339.827.6——33.7
TIESBackbone=RoBERTa-large, Scenario=Combined: BadNet + InsertSent2025.0490.373.656.1——64.9
TIESBackbone=RoBERTa-large, Scenario=Combined: InsertSent + LWS2025.0490.3—6065.3—62.7
WAGBackbone=RoBERTa-large, Scenario=Combined: Hidden + Benign2025.0490.3———4747
OursBackbone=RoBERTa-large, Scenario=Combined: BadNet + Hidden2025.0490.232.5——59.345.9
WAGBackbone=RoBERTa-large, Scenario=Combined: Benign + BadNet2025.0490.247.8———47.8
OursBackbone=RoBERTa-large, Scenario=Combined: BadNet + LWS2025.0490.145.1—68.9—57
TIESBackbone=RoBERTa-large, Scenario=Combined: LWS + Hidden2025.0490.1——73.859.166.5
OursBackbone=RoBERTa-large, Scenario=Combined: Benign + BadNet2025.0490.143.3———43.3
DAREBackbone=RoBERTa-large, Scenario=Combined: LWS + Benign2025.0490.1——48.9—48.9
TIESBackbone=RoBERTa-large, Scenario=Combined: BadNet + LWS2025.049087.3—73.1—80.2
TIESBackbone=RoBERTa-large, Scenario=Combined: BadNet + Hidden2025.049089.4——6476.7
WAGBackbone=RoBERTa-large, Scenario=Combined: BadNet + Hidden2025.0489.961.6——51.756.7
OursBackbone=RoBERTa-large, Scenario=Combined: LWS + Hidden2025.0489.9——70.357.363.8
WAGBackbone=RoBERTa-large, Scenario=Combined: BadNet + LWS2025.0489.859.3—69.3—64.3
WAGBackbone=RoBERTa-large, Scenario=Combined: LWS + Hidden2025.0489.8——70.255.162.7
TIESBackbone=RoBERTa-large, Scenario=Combined: Benign + BadNet2025.0489.864.9———64.9
TIESBackbone=RoBERTa-large, Scenario=Combined: LWS + Benign2025.0489.8——69.3—69.3
TIESBackbone=RoBERTa-large, Scenario=Combined: Hidden + Benign2025.0489.8———54.354.3
VictimBackbone=RoBERTa-large, Scenario=Baseline2025.0489.51001009699.999
OursBackbone=RoBERTa-large, Scenario=Combined: Hidden + Benign2025.0489.4———47.947.9
OursBackbone=RoBERTa-large, Scenario=Combined: LWS + Benign2025.0489.3——64.1—64.1
Z-DefBackbone=RoBERTa-large, Scenario=Baseline2025.0489.211.111.692.250.641.4
WAGBackbone=RoBERTa-large, Scenario=Combined: LWS + Benign2025.0489——65.6—65.6
BenignBackbone=RoBERTa-large, Scenario=Baseline2025.0487.612.312.626.436.922.1
ONIONBackbone=RoBERTa-large, Scenario=Baseline2025.0486.364.398.68998.887.7