Adversarial Claim Rewriting Fact Verification on FEVER
2.63ASRGEM
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| GEMGenerator=GPT-42026.04 | 2.63 | 2.25 | 10 | 37.86 | 0 | 0 | 13.83 | 78.36 | 132.9 | |
| DeceptionGenerator=GPT-42026.04 | 6.84 | 5.75 | 3.85 | 62.11 | 22.73 | 26.08 | 33.61 | 79.13 | 119.6 | |
| OmissionGenerator=GPT-42026.04 | 7.11 | 4.75 | 7.41 | 57.84 | 34.91 | 31.13 | 31.07 | 78.88 | 101.8 | |
| Fact-SaboteursGenerator=GPT-42026.04 | 8.16 | 6.25 | 3.23 | 61.4 | 27.4 | 29.21 | 35.42 | 77.04 | 108.5 | |
| AdvAddGenerator=GPT-42026.04 | 8.16 | 6.75 | 0 | 56.17 | 34.91 | 31.09 | 44.04 | 78.97 | 77.7 | |
| AdvAddGenerator=Qwen2.5-32B2026.04 | 13.5 | 10.75 | 14.81 | 36.78 | 40.88 | 26.65 | 44.59 | 76.95 | 84.88 | |
| GEMGenerator=Mistral-7B2026.04 | 16.32 | 11.75 | 17.74 | 42.82 | 75 | 33 | 23.98 | 74.41 | 154.4 | |
| AdvAddGenerator=Mistral-7B2026.04 | 17.76 | 13.11 | 13.85 | 51.48 | 27.38 | 30.75 | 41.09 | 67.79 | 129.1 | |
| GEMGenerator=Qwen2.5-32B2026.04 | 18.94 | 12.5 | 15.62 | 41.73 | 63.28 | 28.54 | 21.36 | 70.82 | 128.6 | |
| Fact-SaboteursGenerator=Qwen2.5-32B2026.04 | 33.75 | 21.75 | 23.7 | 46.32 | 58.83 | 29.61 | 28.13 | 72.47 | 139.39 | |
| OmissionGenerator=Mistral-7B2026.04 | 36.68 | 20.6 | 7.38 | 43.84 | 100 | 51.88 | 13.44 | 71.34 | 121.5 | |
| DeceptionGenerator=Qwen2.5-32B2026.04 | 45.89 | 23.19 | 10.33 | 62.64 | 74.35 | 39.92 | 23.91 | 54.35 | 113.97 | |
| DeceptionGenerator=Mistral-7B2026.04 | 51.84 | 24 | 5.88 | 68.55 | 60.35 | 35.26 | 26.83 | 55.55 | 113.97 | |
| OmissionGenerator=Qwen2.5-32B2026.04 | 55.5 | 52 | 5.86 | 44.11 | 83.5 | 49.09 | 9.37 | 77.62 | 84.88 | |
| Fact-SaboteursGenerator=Mistral-7B2026.04 | 55.79 | 33.5 | 16.04 | 51.25 | 59.61 | 27.11 | 37.17 | 64.3 | 116.2 |