| Benchmark | Model | Rank | Results |
|---|---|---|---|
| common-sense-reasoning-on-arc-easy | DeepSeek-MoE-16B (INT4 w/o MW ECO + RTN) | – | Accuracy: 71.59 |
| question-answering-on-piqa | DeepSeek-MoE-16B (INT4 w/o MW ECO + SR) | – | Accuracy: 81.5 |
| sentence-completion-on-hellaswag | DeepSeek-MoE-16B (INT4 w/o MW ECO + RTN) | – | Accuracy: 78.88 |