Control-level compliance prediction on OrgA
90.67AccuracyPROPARAG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PROPARAGModel=Sonnet, Type=Closed2026.05 | 90.67 | 86.43 | 91.85 | 88.54 | |
| PROPARAGModel=GPT-4o, Type=Closed2026.05 | 87.09 | 81.7 | 87.33 | 83.67 | |
| PROPARAGModel=Haiku, Type=Closed2026.05 | 84.91 | 79.45 | 85.53 | 81.46 | |
| PROPARAGModel=GPT-4o-mini, Type=Closed2026.05 | 77.06 | 70.44 | 75.11 | 71.49 | |
| PROPARAGModel=Qwen, Type=Open2026.05 | 51.94 | 50.19 | 53.8 | 48.65 | |
| PROPARAGModel=Mistral, Type=Open2026.05 | 51.24 | 48.54 | 50.51 | 47.06 | |
| PROPARAGModel=Llama, Type=Open2026.05 | 41.31 | 39.4 | 39.83 | 37.61 | |
| PROPARAGModel=DeepSeek, Type=Open2026.05 | 35.35 | 34.59 | 34.72 | 32.49 |