Natural Language Inference on MLGym MNLI
92.5AccuracyHuman Best
Evaluation Results
| Method | Links | |
|---|---|---|
| Human BestMethod Type=Human Baseline2026.06 | 92.5 | |
| ARTSBase Model=o32026.06 | 84.71 | |
| LinearMethod Type=Prior Works2026.06 | 84.42 | |
| MLEvolveMethod Type=Prior Works2026.06 | 84.26 | |
| ARTS*Base Model=Qwen 4B, Test-time trained=true2026.06 | 84.01 | |
| AIRAMethod Type=Prior Works2026.06 | 83.77 | |
| ARTSBase Model=Qwen 4B2026.06 | 83.33 |