Sentence-pair classification on QNLI

92.45Accuracy

LeRaC

Updated 4d ago

Evaluation Results

Method
LeRaC 2022.05	92.45	-
conventional 2022.05	92.13	-
BERT (Full Clean Training) 2022.10	91.56	-
Backdoored Model (No Defense) 2022.10	91.56	96.23
CBS 2022.05	91.47	-
Backdoored Model (No Defense) 2022.10	90.79	99.98
Fine-mixing 2022.10	84.29	86.02
Fine-mixing 2022.10	84.07	30.68
Fine-mixing 2022.10	82.83	29.52
Fine-mixing 2022.10	82.83	31.4
Fine-mixing 2022.10	82.47	77.23
Fine-mixing 2022.10	81.9	60.74
Fine-mixing 2022.10	81.68	21.77
Fine-mixing 2022.10	81.37	22.73
Fine-mixing 2022.10	80.3	21.85
Initial BERT (Clean fine-tuned) 2022.10	67.27	-
LeRaC 2022.05	59.98	-
CBS 2022.05	59.89	-
conventional 2022.05	59.42	-
Initial BERT (Clean fine-tuned) 2022.10	49.95	-