List your company

Counterfactual Generation on SNLI Hypothesis

83LFRiFlip-Conf
8.6427.94547.2566.555Jan 4, 2026

Evaluation results

MethodLinks
iFlip-ConfExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B, Feedback Type=Confidence2026.018379.947.43
iFlipBackbone LLM=OLMo2-7B, Feedback Type=Confidence2026.0182.68740.65
FIZLEExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B2026.0182.192.942.39
iFlip-NLExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B, Feedback Type=Natural Language2026.0181.581.552.91
iFlipBackbone LLM=OLMo2-7B, Feedback Type=AttnLRP2026.018186.940.31
iFlip-SHAPExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B, Feedback Type=Feature Attribution2026.018079.544.69
iFlipBackbone LLM=OLMo2-7B, Feedback Type=Grad×Input2026.0178.287.141.81
iFlipBackbone LLM=OLMo2-7B, Feedback Type=Natural Language2026.017790.143.81
iFlipBackbone LLM=OLMo2-7B, Feedback Type=SHAP2026.017686.742.39
iFlipBackbone LLM=Qwen3-32B, Feedback Type=Natural Language2026.0175.490.647.2
Causal What-IfsExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B2026.017489.243.63
iFlipBackbone LLM=OLMo2-7B, Feedback Type=LIME2026.0173.687.541.02
CGGExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B2026.016894.454.23
iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=Natural Language2026.016886.169.45
BAEExplained Model=RoBERTa-base2026.0166.591.492.1
iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=Confidence2026.0164.284.343.88
iFlipBackbone LLM=Qwen3-32B, Feedback Type=AttnLRP2026.016186.744.77
iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=LIME2026.0160.882.542.91
iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=SHAP2026.0160.682.841.09
iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=Grad×Input2026.0160.382.539.59
iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=AttnLRP2026.0160.282.239.92
iFlipBackbone LLM=Qwen3-32B, Feedback Type=SHAP2026.0158.686.645.04
iFlipBackbone LLM=Qwen3-32B, Feedback Type=LIME2026.01578744.22
iFlipBackbone LLM=Qwen3-32B, Feedback Type=Confidence2026.0156.287.147.03
iFlipBackbone LLM=Qwen3-32B, Feedback Type=Grad×Input2026.0154.286.545.6
FIZLEBackbone LLM=LLaMA3.3-70B2026.0153.492.942.39
FIZLEBackbone LLM=Qwen3-32B2026.0152.491.835.97
Causal What-IfsBackbone LLM=Qwen3-32B2026.0151.89234.18
Causal What-IfsBackbone LLM=OLMo2-7B2026.0148.791.135.56
Causal What-IfsBackbone LLM=LLaMA3.3-70B2026.014290.441.29
FIZLEBackbone LLM=OLMo2-7B2026.0140.488.274.15
Polyjuice2026.013573.3114.59
CGGBackbone LLM=Qwen3-32B2026.0128.393.847.94
CGGBackbone LLM=LLaMA3.3-70B2026.0127.193.746.58
PolyjuiceExplained Model=RoBERTa-base2026.012673.3114.59
CGGBackbone LLM=OLMo2-7B2026.0125.394.853.92
BAE2026.0111.597.981