Multimodal Understanding on Evaluation Suite Combined (held-out)
48.4AccuracySKILLRATER
Evaluation results
| Method | Links | |
|---|---|---|
| SKILLRATERTraining Steps=100k, Model Scale=2B, Training Recipe=Isaac 0.2, Rater Parameters=150M, Rater Configuration=capability-specific2026.02 | 48.4 | |
| SKILLRATER CombinedTraining Steps=100k, Model Scale=2B, Training Recipe=Isaac 0.2, Rater Parameters=150M, Rater Configuration=pooled-validation2026.02 | 47.01 | |
| DataRaterTraining Steps=100k, Model Scale=2B, Training Recipe=Isaac 0.2, Rater Parameters=150M, Rater Configuration=monolithic2026.02 | 45.89 | |
| Mammoth (Baseline)Training Steps=100k, Model Scale=2B, Training Recipe=Isaac 0.2, Training Dataset=Mammoth2026.02 | 44.68 |