Dialogue Session Performance Analysis on MultiWOZ 1.0 (test)
13.7RAG CallsNo cache (baseline)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| No cache (baseline)Caching=None2026.03 | 13.7 | 18,800 | 0 | 0 | |
| Cosinetau=0.852026.03 | 7.2 | 9,900 | 0 | 143 | |
| CAS onlyMethod Components=Certified Answer Store2026.03 | 2.1 | 2,900 | 0 | 0 | |
| CAS+PABMethod Components=Certified Answer Store + Pre-Answer Blocks, Coverage p=25%2026.03 | 1.89 | 1,240 | 0 | 0 | |
| CAS+PABMethod Components=Certified Answer Store + Pre-Answer Blocks, Coverage p=50%2026.03 | 1.52 | 720 | 0 | 0 | |
| CAS+PABMethod Components=Certified Answer Store + Pre-Answer Blocks, Coverage p=75%2026.03 | 1.38 | 480 | 0 | 0 | |
| CAS+PABMethod Components=Certified Answer Store + Pre-Answer Blocks, Coverage p=100%2026.03 | 1.31 | 340 | 0 | 0 |