Language Modeling on MLGym Language Modeling
3.5LossHuman Best
Evaluation results
| Method | Links | |
|---|---|---|
| Human BestMethod Type=Human Baseline2026.06 | 3.5 | |
| ARTS*Base Model=Qwen 4B, Test-time trained=true2026.06 | 3.518 | |
| ARTSBase Model=o32026.06 | 3.827 | |
| LinearMethod Type=Prior Works2026.06 | 3.986 | |
| MLEvolveMethod Type=Prior Works2026.06 | 4.015 | |
| ARTSBase Model=Qwen 4B2026.06 | 4.34 | |
| AIRAMethod Type=Prior Works2026.06 | 4.673 |