Personalized Response Generation on Real-world failure cases from large-scale commercial PA
73.4Macro AccuracyRP-Reasoner
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RP-ReasonerReasoning=Rational Personalization Reasoning2026.01 | 73.4 | 83.4 | 1.07 | |
| VanillaPrompting=Vanilla2026.01 | 48.2 | 54.8 | 1.899 | |
| CoTPrompting=Chain-of-Thought2026.01 | 34.2 | 49.7 | 2.216 | |
| ReminderPrompting=Reminder2026.01 | 12.6 | 24.1 | 3.271 |