Multi-agent debate on Closed regime
87.7AccuracyMAD
Evaluation results
| Method | Links | |
|---|---|---|
| MADModel=gpt-oss-120b2026.06 | 87.7 | |
| FlatModel=gpt-oss-120b2026.06 | 87 | |
| R0-voteModel=gpt-oss-120b2026.06 | 86.3 | |
| Situational leadership controllerModel=gpt-oss-120b2026.06 | 85.7 | |
| Transformational leadership controllerModel=gpt-oss-120b2026.06 | 85.7 | |
| MADModel=gemma-4-31B-it2026.06 | 80.7 | |
| Transformational leadership controllerModel=llama-4-scout2026.06 | 80.3 | |
| FlatModel=llama-4-scout2026.06 | 79.7 | |
| FlatModel=gemma-4-31B-it2026.06 | 79.7 | |
| R0-voteModel=llama-4-scout2026.06 | 79 | |
| Situational leadership controllerModel=llama-4-scout2026.06 | 79 | |
| MADModel=llama-4-scout2026.06 | 78.3 | |
| R0-voteModel=gemma-4-31B-it2026.06 | 77.7 | |
| Situational leadership controllerModel=gemma-4-31B-it2026.06 | 77.7 | |
| Transformational leadership controllerModel=gemma-4-31B-it2026.06 | 77.7 |