Single-Needle-In-A-Haystack retrieval on RULER S-NIAH-1
100Accuracy (512 tokens)DN
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DNNumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 100 | 100 | |
| PDNNumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 99.8 | 99.8 | |
| DNNumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[-1, 1]2026.04 | 100 | 100 | 100 | 100 | 100 | |
| GDNNumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 99.6 | 95.2 | |
| PGDNNumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 99.8 | 100 | 98.8 | |
| GDNNumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[-1, 1]2026.04 | 100 | 100 | 100 | 100 | 99.8 | |
| KDANumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 99.8 | 96.8 | |
| PKDANumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 100 | 98.4 | |
| KDANumber of Parameters=340M, Training Tokens=15B, Recurrence Parameter Range (beta_t)=[-1, 1]2026.04 | 100 | 100 | 100 | 100 | 99.6 | |
| GDNNumber of Parameters=1B, Training Tokens=50B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 100 | 100 | |
| PGDNNumber of Parameters=1B, Training Tokens=50B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 100 | 100 | |
| KDANumber of Parameters=1B, Training Tokens=50B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 100 | 100 | |
| PKDANumber of Parameters=1B, Training Tokens=50B, Recurrence Parameter Range (beta_t)=[0, 1]2026.04 | 100 | 100 | 100 | 100 | 100 |