Misinformation Detection on MISBENCH One-hop based Misinformation 1.0 (test)
91.44Factual Memory Success RateGPT-4o
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| GPT-4oModel Series=Closed-source Models2025.05 | 91.44 | 88.2 | 3.24 | 99.33 | 98.93 | 0.4 | 93.96 | 89.28 | 4.68 | |
| LLaMA3-70BModel Series=LLaMA3 Series2025.05 | 75.12 | 64.67 | 10.45 | 95.02 | 93.26 | 1.76 | 64.07 | 52.83 | 11.24 | |
| Qwen2.5-3BModel Series=Qwen2.5 Series2025.05 | 73.48 | 67.31 | 6.17 | 93.14 | 90.68 | 2.46 | 63.65 | 52.07 | 11.58 | |
| Qwen2.5-14BModel Series=Qwen2.5 Series2025.05 | 68.88 | 58.66 | 10.22 | 99.29 | 99.26 | 0.03 | 71.16 | 56.82 | 14.34 | |
| Claude3.5-haikuModel Series=Closed-source Models2025.05 | 67.15 | 60.33 | 6.82 | 85.04 | 81.24 | 3.8 | 62.96 | 46.29 | 16.67 | |
| Qwen2.5-72BModel Series=Qwen2.5 Series2025.05 | 57.23 | 43.84 | 13.39 | 77.41 | 69.35 | 8.06 | 57.49 | 35.86 | 21.63 | |
| Gemma2-9BModel Series=Gemma2 Series2025.05 | 55.94 | 50.53 | 5.41 | 94.83 | 94.21 | 0.62 | 47.2 | 38.35 | 8.85 | |
| Gemma2-27BModel Series=Gemma2 Series2025.05 | 42.5 | 31.8 | 10.7 | 68.64 | 58.16 | 10.48 | 34.72 | 19.38 | 15.34 | |
| Gemma2-2BModel Series=Gemma2 Series2025.05 | 36.74 | 32.86 | 3.88 | 70.36 | 63.55 | 6.81 | 29.1 | 22.34 | 6.76 | |
| DeepSeek-V2.5Model Series=Closed-source Models2025.05 | 34.56 | 26.42 | 8.14 | 55.61 | 47.8 | 7.81 | 43.78 | 28.93 | 14.85 | |
| LLaMA3-8BModel Series=LLaMA3 Series2025.05 | 19.21 | 16.91 | 2.3 | 36.26 | 33.32 | 2.94 | 13.67 | 9.45 | 4.22 | |
| Qwen2.5-7BModel Series=Qwen2.5 Series2025.05 | 14.22 | 9.47 | 4.75 | 48.32 | 45.71 | 2.61 | 16.13 | 7.83 | 8.3 |