Scientific Reasoning on GPQA (Acc, Tok, CR)
66.2AccuracyDTSR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DTSRBackbone=Qwen3-14B2026.04 | 66.2 | 5,916 | — | |
| VanillaBackbone=Qwen3-14B2026.04 | 65.7 | 7,513 | — | |
| TALEBackbone=Qwen3-14B2026.04 | 65.6 | 7,138 | — | |
| VanillaBackbone=Ministral-3-14B-Reasoning-25122026.03 | 62.7 | 9,539 | 100 | |
| VanillaBackbone=Ministral-3-8B-Reasoning-25122026.03 | 62.3 | 11,765 | 100 | |
| DEERBackbone=Ministral-3-8B-Reasoning-25122026.03 | 61.1 | 11,312 | 94.1 | |
| TERMINATORBackbone=Ministral-3-14B-Reasoning-25122026.03 | 60.6 | 7,727 | 76.5 | |
| VanillaBackbone=Qwen3-14B2026.03 | 59.9 | 7,628 | 100 | |
| DEERBackbone=Ministral-3-14B-Reasoning-25122026.03 | 58.5 | 9,185 | 95 | |
| TERMINATORBackbone=Ministral-3-8B-Reasoning-25122026.03 | 58.2 | 9,588 | 77.4 | |
| VanillaBackbone=Qwen3-8B2026.03 | 57.6 | 8,594 | 100 | |
| Thought-CalibBackbone=Qwen3-14B2026.03 | 54.6 | 5,757 | 81.9 | |
| TERMINATORBackbone=Qwen3-14B2026.03 | 53.9 | 6,798 | 87.1 | |
| Thought-CalibBackbone=Qwen3-8B2026.03 | 52.6 | 6,240 | 78.9 | |
| TERMINATORBackbone=Qwen3-8B2026.03 | 52.1 | 7,543 | 85.7 | |
| DynasorBackbone=Ministral-3-8B-Reasoning-25122026.03 | 51.8 | 9,766 | 83.7 | |
| DynasorBackbone=Ministral-3-14B-Reasoning-25122026.03 | 51.3 | 7,917 | 85.2 | |
| CoDBackbone=Qwen3-14B2026.04 | 51 | 1,177 | — | |
| DEERBackbone=Qwen3-8B2026.03 | 50.5 | 8,553 | 99.6 | |
| Thought-CalibBackbone=Ministral-3-14B-Reasoning-25122026.03 | 49.5 | 6,028 | 73.8 | |
| DEERBackbone=Qwen3-14B2026.03 | 49 | 7,451 | 97.4 | |
| Thought-CalibBackbone=Ministral-3-8B-Reasoning-25122026.03 | 47.3 | 7,406 | 71.8 | |
| DynasorBackbone=Qwen3-14B2026.03 | 45.6 | 2,101 | 29.1 | |
| DynasorBackbone=Qwen3-8B2026.03 | 43.2 | 2,455 | 28.4 | |
| NoThinkingBackbone=Ministral-3-14B-Reasoning-25122026.03 | 41 | 570 | 6.6 | |
| NoThinkingBackbone=Ministral-3-8B-Reasoning-25122026.03 | 39.7 | 2,106 | 16 | |
| NoThinkingBackbone=Qwen3-14B2026.03 | 32.1 | 1,265 | 18.8 | |
| NoThinkingBackbone=Qwen3-8B2026.03 | 30.7 | 1,204 | 15.8 |