Validity Certification on HH-RLHF (test)
100FTV@k=1TPA
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| TPAModel=OLMo-1B2026.02 | 100 | 100 | 99.1 | 82.57 | 54.78 | 20.78 | 19.67 | 19.34 | 14.69 | 9.21 | |
| TPA+MSCModel=OLMo-1B2026.02 | 100 | 100 | 100 | 100 | 99 | 26.13 | 25.61 | 24.96 | 24.23 | 22.85 | |
| TPAModel=Gemma2-2B2026.02 | 100 | 100 | 98.5 | 65.63 | 26.19 | 20.49 | 19.09 | 18.45 | 13.12 | 8.19 | |
| TPA+MSCModel=Gemma2-2B2026.02 | 100 | 100 | 100 | 99.8 | 95.5 | 25.56 | 24.93 | 24.25 | 23.23 | 21.52 | |
| TPAModel=Qwen1.5-4B2026.02 | 100 | 98.99 | 95.59 | 72.57 | 38.61 | 25.43 | 23.87 | 21.84 | 14.72 | 8.56 | |
| TPA+MSCModel=Qwen1.5-4B2026.02 | 100 | 100 | 100 | 99.2 | 92.59 | 26.64 | 25.85 | 24.91 | 23.57 | 21.16 |