Out-of-Distribution Generalization on MMLU-Redux and GPQA-Diamond
76.7Average AccuracyQwen3-30B-A3B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-30B-A3BBase Model=Qwen3-30B-A3B, Max Generation Length=32k, GPQA Evaluation=avg@8, MMLU Evaluation=avg@12026.05 | 76.7 | 0 | 90.1 | 63.3 | |
| ZEDABase Model=Qwen3-30B-A3B, Max Generation Length=32k, GPQA Evaluation=avg@8, MMLU Evaluation=avg@12026.05 | 76.2 | 47.2 | 89.2 | 63.2 | |
| GLM-4.7-FlashBase Model=GLM-4.7-Flash, Max Generation Length=32k, GPQA Evaluation=avg@8, MMLU Evaluation=avg@12026.05 | 76.1 | 0 | 89.8 | 62.4 | |
| ZEDABase Model=GLM-4.7-Flash, Max Generation Length=32k, GPQA Evaluation=avg@8, MMLU Evaluation=avg@12026.05 | 72.9 | 50 | 89 | 56.8 |