Complexity Prediction on MMLU+MMLU-PRO+GSM8K
89.1ROC-AUCIntroLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| IntroLMBackbone Model=Qwen3-8B2026.01 | 89.1 | 63.4 | |
| DeBERTa-v3-LargeNumber of Parameters=435M, Backbone Model=DeBERTa-v3-Large2026.01 | 75.8 | 45.5 | |
| DeBERTa-v3-BaseNumber of Parameters=184M, Backbone Model=DeBERTa-v3-Base2026.01 | 74.3 | 44.3 |