High-Entropy SOP Evaluation on MultiWOZ 2.2
94.2Vanilla APAVanilla
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| VanillaModel Pair (Architect / Executive)=GPT 5.4 / 5.4 mini2026.04 | 94.2 | — | — | — | |
| VanillaModel Pair (Architect / Executive)=Claude Sonnet 4.6 / Haiku 4.52026.04 | 53.8 | — | — | — | |
| VanillaModel Pair (Architect / Executive)=Gemini 3.1 Pro / 2.5 Flash2026.04 | 41.3 | — | — | — | |
| VanillaModel Pair (Architect / Executive)=DeepSeek Reasoner / V3.22026.04 | 31 | — | — | — | |
| SSRPModel Pair (Architect / Executive)=Gemini 3.1 Pro / 2.5 Flash2026.04 | — | 50.6 | 22.52 | 1.8 | |
| SSRPModel Pair (Architect / Executive)=Claude Sonnet 4.6 / Haiku 4.52026.04 | — | 92.5 | 71.93 | 1.7 | |
| SSRPModel Pair (Architect / Executive)=DeepSeek Reasoner / V3.22026.04 | — | 49 | 58.06 | 9.8 | |
| SSRPModel Pair (Architect / Executive)=GPT 5.4 / 5.4 mini2026.04 | — | 95 | 0.85 | 4.52 |