Multi-task Language Understanding on MMLU (Source Lib., Unseen Lib., Delta)
75.68MMLU Source Lib. AccuracySIGMA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SIGMABase Model=GPT-4o-mini2026.06 | 75.68 | 75.03 | 0.65 | |
| CARDBase Model=GPT-4o-mini2026.06 | 75.11 | 73.68 | 1.43 | |
| G-DesignerBase Model=GPT-4o-mini2026.06 | 73.32 | 70.05 | 3.27 | |
| GPTSwarmBase Model=GPT-4o-mini2026.06 | 71.37 | 64.84 | 6.53 | |
| LLM-DebateBase Model=GPT-4o-mini2026.06 | 68.96 | 63.73 | 5.23 |