Language Understanding and Reasoning on PIQA, ARC-e, HellaSwag, GPQA, Lambada, MMLU, and BBH Suite
50.67PIQADAG-MoE-l
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DAG-MoE-lFine-tuning=Instruction-tuned, HellaSwag shots=10-shot, MMLU shots=5-shot, Other shots=0-shot2026.05 | 50.67 | 25.57 | 25.73 | 27.78 | 11.57 | 24.03 | 17.55 | 26.13 | |
| MoE-lFine-tuning=Instruction-tuned, HellaSwag shots=10-shot, MMLU shots=5-shot, Other shots=0-shot2026.05 | 47.52 | 24.34 | 25.9 | 21.72 | 8.11 | 24.17 | 16.65 | 24.06 |