Science Reasoning on GPQA diamond (Acc., Tokens)
72.5AccuracySC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SCBackbone=DeepSeek-R1-0528-Qwen3-8B, Voting size budget=5122026.05 | 72.5 | 99.2 | |
| ACBackbone=DeepSeek-R1-0528-Qwen3-8B, Voting size budget=5122026.05 | 72.4 | 38 | |
| D-HBackbone=DeepSeek-R1-0528-Qwen3-8B, Voting size budget=5122026.05 | 72.4 | 69.1 | |
| DDCBackbone=Qwen3-32B, Voting size budget=5122026.05 | 72.2 | 14.1 | |
| D-LBackbone=DeepSeek-R1-0528-Qwen3-8B, Voting size budget=5122026.05 | 71.7 | 34.6 | |
| SCBackbone=Qwen3-32B, Voting size budget=5122026.05 | 71.1 | 74.4 | |
| D-LBackbone=Qwen3-32B, Voting size budget=5122026.05 | 70.9 | 32.1 | |
| ACBackbone=Qwen3-32B, Voting size budget=5122026.05 | 70.1 | 23.7 | |
| D-HBackbone=Qwen3-32B, Voting size budget=5122026.05 | 70 | 41.6 | |
| DDCBackbone=DeepSeek-R1-0528-Qwen3-8B, Voting size budget=5122026.05 | 69.6 | 13.9 | |
| D-LBackbone=Qwen3-8B, Voting size budget=5122026.05 | 65.2 | 33.1 | |
| D-HBackbone=Qwen3-8B, Voting size budget=5122026.05 | 63.8 | 49.4 | |
| SCBackbone=Qwen3-8B, Voting size budget=5122026.05 | 63.7 | 74.7 | |
| DDCBackbone=Qwen3-8B, Voting size budget=5122026.05 | 63.6 | 13.7 | |
| ACBackbone=Qwen3-8B, Voting size budget=5122026.05 | 63.3 | 24.8 | |
| DDCBackbone=Qwen3-4B, Voting size budget=5122026.05 | 59.1 | 13.7 | |
| D-LBackbone=Qwen3-4B, Voting size budget=5122026.05 | 59 | 48.2 | |
| D-HBackbone=Qwen3-4B, Voting size budget=5122026.05 | 48.7 | 50.6 | |
| SCBackbone=Qwen3-4B, Voting size budget=5122026.05 | 48.5 | 80.4 | |
| ACBackbone=Qwen3-4B, Voting size budget=5122026.05 | 48.5 | 23.7 | |
| DDCBackbone=Qwen3-1.7B, Voting size budget=5122026.05 | 42.9 | 20.6 | |
| D-LBackbone=Qwen3-1.7B, Voting size budget=5122026.05 | 40.4 | 126.6 | |
| D-HBackbone=Qwen3-1.7B, Voting size budget=5122026.05 | 40.3 | 166.5 | |
| SCBackbone=Qwen3-1.7B, Voting size budget=5122026.05 | 40.2 | 186.4 | |
| ACBackbone=Qwen3-1.7B, Voting size budget=5122026.05 | 40.1 | 69.7 |