Scientific Reasoning on GPQA Diamond (Pass@1, Token Count, Saved %)
2,004.7Token CountTRACES
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TRACESthreshold=0.9, Backbone=DS-Qwen14B, Seed=422026.04 | 2,004.7 | 55.8 | 71.46 | |
| TRACESthreshold=0.8, Backbone=DS-Qwen14B, Seed=422026.04 | 2,874.6 | 58.3 | 59.07 | |
| Basel. P^{(1)}_{system}Backbone=DS-Qwen14B, Seed=422026.04 | 3,736.2 | 53.9 | 46.81 | |
| TRACESthreshold=0.7, Backbone=DS-Qwen14B, Seed=422026.04 | 3,866.4 | 62.2 | 44.95 | |
| Basel. P^{(3)}_{system}Backbone=DS-Qwen14B, Seed=422026.04 | 4,530.1 | 56.4 | 35.51 | |
| TRACESthreshold=0.6, Backbone=DS-Qwen14B, Seed=422026.04 | 4,570.1 | 60.3 | 34.94 | |
| Basel. P^{(0)}_{system}Backbone=DS-Qwen14B, Seed=422026.04 | 4,768.3 | 57.1 | 32.12 | |
| Basel. P^{(0)}_{user}Backbone=DS-Qwen14B, Seed=422026.04 | 4,801 | 57.7 | 31.66 | |
| TRACESthreshold=0.5, Backbone=DS-Qwen14B, Seed=422026.04 | 5,297.7 | 58.3 | 24.58 | |
| TRACESthreshold=0.4, Backbone=DS-Qwen14B, Seed=422026.04 | 5,683.2 | 58.9 | 19.1 | |
| StandardBackbone=DS-Qwen14B, Seed=422026.04 | 7,024.7 | 60.3 | — |