Question Answering on SimpleQA (pass@128)
53.97pass@128WRAP++
Evaluation results
| Method | Links | |
|---|---|---|
| WRAP++Model=OLMo-3-32B, Training=1-epoch continued training, Data Tokens=~82.7B2026.04 | 53.97 | |
| WRAP++Model=OLMo-3-7B, Training=1-epoch continued training, Data Tokens=~82.7B2026.04 | 49.13 | |
| Extended WRAPModel=OLMo-3-32B, Training=1-epoch continued training, Data Tokens=~17.4B2026.04 | 47.91 | |
| WRAPModel=OLMo-3-32B, Training=1-epoch continued training, Data Tokens=~5.4B2026.04 | 44.43 | |
| Extended WRAPModel=OLMo-3-7B, Training=1-epoch continued training, Data Tokens=~17.4B2026.04 | 43.69 | |
| Pretrained BaseModel=OLMo-3-32B, Training=1-epoch continued training2026.04 | 42.35 | |
| WRAPModel=OLMo-3-7B, Training=1-epoch continued training, Data Tokens=~5.4B2026.04 | 39.55 | |
| + WRAP++ Mix (100B)Setting=+ WRAP++ Mix (100B)2026.04 | 37.58 | |
| Pretrained BaseModel=OLMo-3-7B, Training=1-epoch continued training2026.04 | 34.76 | |
| Pretrained BaseSetting=Pretrained Base2026.04 | 34.76 | |
| + Midtrain (100B)Setting=+ Midtrain (100B)2026.04 | 34.74 |