End-to-end dialogue generation on Controlled setting pilot dialogues
83.3CoverageFull system
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Full system2026.03 | 83.3 | 81.4 | 0.69 | 15.8 | 80 | 5.8 | |
| Baseline C: Rule-template interaction2026.03 | 77.8 | 78.6 | 0.61 | 28.4 | 70 | 7.4 | |
| Baseline B: Chunk-only RAG2026.03 | 69.4 | 71.4 | 0.54 | 24.8 | 60 | 6.8 | |
| Baseline A: Direct generationtype=non-interactive2026.03 | 55.6 | 59.3 | 0.41 | — | 30 | — |