Counterfactual Generation on IMDb
100LFRiFlip-Conf
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| iFlip-ConfExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B, Feedback Type=Confidence2026.01 | 100 | 89.5 | 37.48 | |
| iFlip-SHAPExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B, Feedback Type=Feature Attribution2026.01 | 100 | 89.5 | 36.19 | |
| iFlip-NLExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B, Feedback Type=Natural Language2026.01 | 100 | 89.3 | 36.48 | |
| iFlipBackbone LLM=Qwen3-32B, Feedback Type=AttnLRP2026.01 | 100 | 85.8 | 33.19 | |
| iFlipBackbone LLM=Qwen3-32B, Feedback Type=LIME2026.01 | 99.8 | 86.3 | 32.57 | |
| iFlipBackbone LLM=Qwen3-32B, Feedback Type=Confidence2026.01 | 99.6 | 85.5 | 33.3 | |
| iFlipBackbone LLM=Qwen3-32B, Feedback Type=Grad×Input2026.01 | 99.6 | 86.5 | 33.65 | |
| iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=Confidence2026.01 | 99.6 | 87.7 | 34.74 | |
| iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=SHAP2026.01 | 99.6 | 87.8 | 34.32 | |
| iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=Grad×Input2026.01 | 99.6 | 87.8 | 34.8 | |
| iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=LIME2026.01 | 99.6 | 87.7 | 34.54 | |
| iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=Natural Language2026.01 | 99.6 | 89.5 | 36.9 | |
| iFlipBackbone LLM=Qwen3-32B, Feedback Type=SHAP2026.01 | 99.4 | 85.7 | 33.7 | |
| iFlipBackbone LLM=LLaMA3.3-70B, Feedback Type=AttnLRP2026.01 | 99.2 | 87.9 | 34.58 | |
| iFlipBackbone LLM=OLMo2-7B, Feedback Type=Natural Language2026.01 | 98.4 | 82.2 | 36.47 | |
| iFlipBackbone LLM=Qwen3-32B, Feedback Type=Natural Language2026.01 | 98 | 87.8 | 32.99 | |
| iFlipBackbone LLM=OLMo2-7B, Feedback Type=AttnLRP2026.01 | 97.8 | 80.3 | 42.61 | |
| iFlipBackbone LLM=OLMo2-7B, Feedback Type=Grad×Input2026.01 | 97.6 | 80.1 | 43.3 | |
| iFlipBackbone LLM=OLMo2-7B, Feedback Type=SHAP2026.01 | 96.8 | 80.5 | 41.87 | |
| iFlipBackbone LLM=OLMo2-7B, Feedback Type=LIME2026.01 | 96.8 | 80.5 | 52.82 | |
| iFlipBackbone LLM=OLMo2-7B, Feedback Type=Confidence2026.01 | 96.4 | 80.9 | 41.19 | |
| FIZLEExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B2026.01 | 95.6 | 86.8 | 34.88 | |
| CGGExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B2026.01 | 93 | 89.2 | 54.57 | |
| FIZLEBackbone LLM=LLaMA3.3-70B2026.01 | 92.4 | 86.8 | 34.88 | |
| Causal What-IfsBackbone LLM=LLaMA3.3-70B2026.01 | 92.2 | 78.8 | 31.93 | |
| Causal What-IfsExplained Model=RoBERTa-base, Generator Model=LLaMA 3.3-70B2026.01 | 90.8 | 78.8 | 32.78 | |
| Causal What-IfsBackbone LLM=OLMo2-7B2026.01 | 89 | 80.1 | 37.97 | |
| CGGBackbone LLM=LLaMA3.3-70B2026.01 | 88.7 | 86.9 | 52.81 | |
| CGGBackbone LLM=OLMo2-7B2026.01 | 88.5 | 83.7 | 44.67 | |
| CGGBackbone LLM=Qwen3-32B2026.01 | 86.3 | 87.7 | 62.47 | |
| Causal What-IfsBackbone LLM=Qwen3-32B2026.01 | 84.8 | 82.1 | 42.03 | |
| FIZLEBackbone LLM=Qwen3-32B2026.01 | 68.4 | 90 | 36 | |
| BAEExplained Model=RoBERTa-base2026.01 | 63.7 | 97.7 | 62.26 | |
| FIZLEBackbone LLM=OLMo2-7B2026.01 | 58.5 | 58.4 | 38.71 | |
| BAE2026.01 | 55.3 | 97.6 | 63.15 | |
| Polyjuice2026.01 | 27.5 | 73.6 | 60.81 | |
| PolyjuiceExplained Model=RoBERTa-base2026.01 | 27.4 | 73.6 | 60.81 |