Grade School Math Reasoning on GSM8K (Accuracy, #Calls)
94.9AccuracyCGES-RM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CGES-RMBudget (B)=162025.11 | 94.9 | 1.25 | |
| SCBudget (B)=162025.11 | 94.2 | 16 | |
| ESCBudget (B)=16, Window size (w)=42025.11 | 94.2 | 5.16 | |
| ESCBudget (B)=16, Window size (w)=82025.11 | 94.2 | 9.41 | |
| ASC-βBudget (B)=162025.11 | 94.2 | 5 | |
| ASC-DirBudget (B)=162025.11 | 94.2 | 5.17 | |
| CGES-MARSBudget (B)=162025.11 | 94.2 | 3.11 | |
| CGES-LNS (arith)Budget (B)=16, Aggregation function=arith2025.11 | 94.1 | 2.41 | |
| CGES-LNS (geom)Budget (B)=16, Aggregation function=geom2025.11 | 94.1 | 3.04 | |
| CGES-DeepConf (B10)Budget (B)=16, DeepConf variant=B102025.11 | 94.1 | 2.41 | |
| CGES-DeepConf (tail)Budget (B)=16, DeepConf variant=tail2025.11 | 94.1 | 2.41 | |
| DSCBudget (B)=162025.11 | 93.9 | 2.85 |