Truthfulness Detection on NoVo benchmark
71.97TQA AccuracyTruthV(argmin)
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| TruthV(argmin)Model=Llama-2-7B-Chat2025.09 | 71.97 | 53.88 | 54.43 | 74.92 | 70.85 | 65.25 | 75.68 | 66.23 | 50.34 | 73.49 | 65.7 | |
| TruthV(argmin)Model=Gemma-2-2B-it2025.09 | 71.48 | 55.37 | 73.54 | 71.15 | 70.79 | 69.81 | 80.14 | 71.42 | 60.72 | 78.05 | 70.25 | |
| TruthV(argmax)Model=Llama-2-7B-Chat2025.09 | 70.13 | 53.68 | 57.02 | 75.43 | 69.67 | 66.38 | 77.04 | 62.35 | 53.07 | 74.98 | 65.98 | |
| NoVoModel=Llama-2-7B-Chat2025.09 | 70.01 | 55.41 | 43.63 | 68.67 | 59.8 | 60.03 | 69.75 | 52.19 | 35.91 | 62.79 | 57.82 | |
| TruthV(argmax)Model=Gemma-2-2B-it2025.09 | 69.4 | 56.55 | 73.97 | 73.4 | 69.85 | 69.86 | 79.92 | 71.66 | 60.78 | 77.87 | 70.33 | |
| NoVoModel=Qwen3-4B2025.09 | 68.67 | — | 66.74 | 68.3 | 59.69 | 65.61 | 73.07 | 59.6 | 43.14 | 70.31 | 63.26 | |
| TruthV(argmin)Model=Qwen3-4B2025.09 | 67.81 | 60.72 | 75.7 | 73.96 | 70.28 | 68.78 | 80.36 | 63.92 | 59.44 | 76.87 | 69.78 | |
| TruthV(argmax)Model=Qwen3-4B2025.09 | 67.32 | 60.57 | 76.24 | 74.42 | 71.34 | 69.04 | 80.9 | 65.96 | 59.4 | 77.62 | 70.28 | |
| NoVoModel=Gemma-2-2B-it2025.09 | 63.16 | 56.2 | 51.73 | 59.63 | 59.22 | 67.76 | 73.94 | 65.19 | 50.44 | 69.39 | 61.67 | |
| TruthV(argmax)Model=Llama-3.2-3B-Instruct2025.09 | 60.71 | 59.74 | 50.11 | 59.35 | 68.15 | 61.31 | 72.63 | 53.23 | 37.2 | 57.84 | 58.03 | |
| NoVoModel=Llama-3.2-3B-Instruct2025.09 | 58.02 | 59.78 | 40.71 | 61.75 | 61.18 | 55.17 | 66.65 | 44.62 | 27.8 | 47.04 | 52.27 | |
| TruthV(argmin)Model=Llama-3.2-3B-Instruct2025.09 | 55.45 | 59.62 | 39.63 | 58.7 | 66.24 | 60.7 | 69.48 | 52.86 | 37.01 | 55.1 | 55.48 | |
| Log-LikelihoodModel=Gemma-2-2B-it2025.09 | 47 | 58.84 | 60.26 | 53.28 | 48.21 | 49.13 | 76.66 | 29.48 | 32.9 | 36.78 | 49.25 | |
| Log-LikelihoodModel=Llama-3.2-3B-Instruct2025.09 | 39.29 | 58.76 | 53.35 | 49.84 | 48.3 | 49.28 | 74.37 | 30.12 | 29.12 | 33.82 | 46.62 | |
| Log-LikelihoodModel=Qwen3-4B2025.09 | 35.74 | 64.94 | 44.92 | 44.47 | 36.13 | 39.76 | 67.79 | 22.61 | 21.08 | 32.36 | 40.98 | |
| Log-LikelihoodModel=Llama-2-7B-Chat2025.09 | 34.88 | 55.73 | 17.49 | 48.67 | 47.57 | 46.37 | 75.08 | 25.43 | 31.43 | 34.28 | 41.69 |