| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Dialogue Generation | UltraChat | ASR Accuracy98.7 | 32 | |
| Fingerprint Spoofing | UltraChat (8:2) | MET ASR100 | 21 | |
| Memorization Reduction | Ultrachat | Memorization Reduction95.7 | 20 | |
| Unlearning Detection | UltraChat | Accuracy99.86 | 12 | |
| Natural Language Generation | UltraChat | BLEU36.8 | 8 | |
| Language Model Training Efficiency | UltraChat | Throughput3,008.3 | 6 | |
| Conversational Instruction Following | UltraChat | Overall Score9.1 | 6 | |
| Personalized Interaction | ultrachat Synthetic | Personalization Score7.26 | 6 | |
| Toxic Example Retrieval | UltraChat 8K XSTest-response-Hom variant | AUPRC0.7562 | 5 | |
| Toxic Example Retrieval | UltraChat 8K XSTest-response-Het variant | AUPRC70.75 | 5 | |
| Unlearning detection (distinguishing original vs. RMU-unlearned) | UltraChat (test) | Accuracy0.8746 | 4 | |
| Instruction Following | UltraChat | RM Score67.8 | 2 |