Scientific Reasoning on GPQA (Wall-clock time (s))
36.98Wall-clock Time (s)DSC
Evaluation Results
| Method | Links | |
|---|---|---|
| DSCBudget (B)=16, LLM batch size=12025.11 | 36.98 | |
| CGES-DeepConf (tail)Budget (B)=16, LLM batch size=1, DeepConf setting=tail2025.11 | 63.51 | |
| CGES-LNS (arith)Budget (B)=16, LLM batch size=12025.11 | 65.83 | |
| ASC-βBudget (B)=16, LLM batch size=12025.11 | 68.21 | |
| CGES-DeepConf (B10)Budget (B)=16, LLM batch size=1, DeepConf setting=B102025.11 | 68.38 | |
| ESCBudget (B)=16, LLM batch size=1, window size w=42025.11 | 71.03 | |
| ASC-DirBudget (B)=16, LLM batch size=12025.11 | 72.54 | |
| CGES-LNS (geom)Budget (B)=16, LLM batch size=12025.11 | 75.37 | |
| ESCBudget (B)=16, LLM batch size=1, window size w=82025.11 | 83.85 | |
| CGES-RMBudget (B)=16, LLM batch size=12025.11 | 86 | |
| SCBudget (B)=16, LLM batch size=12025.11 | 88.25 | |
| CGES-MARSBudget (B)=16, LLM batch size=12025.11 | 89.7 |