Question Answering on Sahih al-Bukhari and Sahih Muslim Hadith QA dataset (val)
6.2Training Time (hours)DistilBERT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DistilBERTTrainable Parameters=82M2026.02 | 6.2 | 95 | 420 | |
| CGRATrainable Parameters=1.2M (Gating + LoRA)2026.02 | 8.5 | 142 | 768 | |
| BERTTrainable Parameters=110M2026.02 | 9.8 | 118 | 580 | |
| RoBERTaTrainable Parameters=125M2026.02 | 10.1 | 125 | 612 | |
| DeBERTaTrainable Parameters=184M, Fine-tuning mode=Full FT2026.02 | 12.3 | 131 | 760 |