Arithmetic Reasoning on MultiArith (Source/Unseen Library Accuracy)
99.64Source Library AccuracySIGMA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SIGMABase Model=GPT-4o-mini2026.06 | 99.64 | 99.46 | 0.18 | |
| CARDBase Model=GPT-4o-mini2026.06 | 96.94 | 95.95 | 0.99 | |
| G-DesignerBase Model=GPT-4o-mini2026.06 | 96.75 | 95.5 | 1.25 | |
| GPTSwarmBase Model=GPT-4o-mini2026.06 | 96.07 | 93.93 | 2.14 | |
| LLM-DebateBase Model=GPT-4o-mini2026.06 | 94.11 | 91.96 | 2.15 |