Dialogue Alignment on Relational Failure Mode Benchmark (test)
0.4116Mean UtilityReranked-best
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Reranked-beststrategy=utility-based reranking2026.04 | 0.4116 | 0.2498 | |
| Manual-DPO2026.04 | 0.3422 | 0.1804 | |
| SFT-baseline2026.04 | 0.1618 | — | |
| CCN-candidatemode=candidate generation2026.04 | -0.1498 | -0.3116 |