Control-level coverage prediction on OrgA policy corpus
88.54F1 ScorePROPARAG
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PROPARAG2026.05 | 88.54 | 90.67 | 86.43 | 91.85 | — | |
| PROPARAG2026.05 | 88.54 | — | — | — | 87.6 | |
| B4: Dense RAG + Labelretrieval=Dense RAG, mode=Label2026.05 | 76.88 | 78.46 | 77.21 | 78.46 | — | |
| Dense RAG + Label2026.05 | 76.88 | — | — | — | 75 | |
| B2: Retrieval + Labelmode=Retrieval + Label2026.05 | 68.93 | 70.21 | 69.44 | 70.21 | — | |
| Retrieval + Label2026.05 | 68.93 | — | — | — | 66.9 | |
| B3: BM25 + Labelretrieval=BM25, mode=Label2026.05 | 58.12 | 60.47 | 59.36 | 60.47 | — | |
| BM25 + Label2026.05 | 58.12 | — | — | — | 56 | |
| B5: Document-Levelgranularity=Document-Level2026.05 | 52.61 | 56.28 | 54.11 | 56.28 | — | |
| Document-Level2026.05 | 52.61 | — | — | — | 50.4 | |
| B1: Single-Shotmode=Single-Shot2026.05 | 50.87 | 54.61 | 52.74 | 54.61 | — | |
| Single-Shot LLM2026.05 | 50.87 | — | — | — | 48.6 |