STEM & Knowledge on MathVision, MMMU, MMMU-Pro (held-out)
23.61AccuracySKILLRATER
Evaluation Results
| Method | Links | |
|---|---|---|
| SKILLRATERTraining Steps=100k, Model Scale=2B, Training Recipe=Isaac 0.2, Rater Parameters=150M, Rater Configuration=capability-specific2026.02 | 23.61 | |
| SKILLRATER CombinedTraining Steps=100k, Model Scale=2B, Training Recipe=Isaac 0.2, Rater Parameters=150M, Rater Configuration=pooled-validation2026.02 | 22.39 | |
| DataRaterTraining Steps=100k, Model Scale=2B, Training Recipe=Isaac 0.2, Rater Parameters=150M, Rater Configuration=monolithic2026.02 | 20.88 | |
| Mammoth (Baseline)Training Steps=100k, Model Scale=2B, Training Recipe=Isaac 0.2, Training Dataset=Mammoth2026.02 | 20.08 |