Scientific Q&A on GPQA Diamond (test)
65.22AccuracyPIR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PIRBackbone=DeepSeek-R1-32B2026.02 | 65.22 | 3,024.81 | |
| s1-32BBackbone=DeepSeek-R1-32B2026.02 | 60.72 | 3,402.53 | |
| SEALBackbone=DeepSeek-R1-32B2026.02 | 56.13 | 3,468.96 | |
| DEERBackbone=DeepSeek-R1-32B2026.02 | 54.2 | 3,402.51 | |
| Specutive ThinkingBackbone=DeepSeek-R1-32B2026.02 | 51.37 | 5,031.91 | |
| SpecReasonBackbone=DeepSeek-R1-32B2026.02 | 44.43 | 4,885.98 | |
| QwQ-32BBackbone=QwQ-32B2026.02 | 37.42 | 5,585.19 | |
| DeepSeek-R1-32BBackbone=DeepSeek-R1-32B2026.02 | 36.3 | 5,028.59 | |
| prompt to conciseBackbone=DeepSeek-R1-32B2026.02 | 31.98 | 4,120.47 | |
| L1-maxBackbone=DeepSeek-R1-32B2026.02 | 25.81 | 3,769.06 | |
| DeepSeek-R1-14BBackbone=DeepSeek-R1-14B2026.02 | 24.66 | 4,411.01 | |
| DeepSeek-R1-7BBackbone=DeepSeek-R1-7B2026.02 | 22.8 | 5,020.86 |