Control-level compliance prediction on OrgB
84.91AccuracyPROPARAG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PROPARAGModel=Sonnet, Type=Closed2026.05 | 84.91 | 80.81 | 84.84 | 82.31 | |
| PROPARAGModel=Haiku, Type=Closed2026.05 | 80.04 | 75.45 | 80.03 | 76.69 | |
| PROPARAGModel=GPT-4o, Type=Closed2026.05 | 79.64 | 74.89 | 79.18 | 76.1 | |
| PROPARAGModel=GPT-4o-mini, Type=Closed2026.05 | 75.57 | 71.4 | 76.25 | 72.57 | |
| PROPARAGModel=Mistral, Type=Open2026.05 | 70.61 | 66.04 | 69.75 | 66.58 | |
| PROPARAGModel=Qwen, Type=Open2026.05 | 68.42 | 64.24 | 68.07 | 64.67 | |
| PROPARAGModel=Llama, Type=Open2026.05 | 54.02 | 51.38 | 53.08 | 50.4 | |
| PROPARAGModel=DeepSeek, Type=Open2026.05 | 40.81 | 39.96 | 41.53 | 38.62 |