Multi-task Evaluation on Mixed Benchmark Average
40.86Average Accuracyphi-balancing
Evaluation results
| Method | Links | |
|---|---|---|
| phi-balancingModel=DeepSeek-V2-Lite2026.05 | 40.86 | |
| ST-MoEModel=DeepSeek-V2-Lite2026.05 | 40.45 | |
| phi-balancingModel=DeepSeek-MoE-Chat2026.05 | 36.12 | |
| ST-MoEModel=DeepSeek-MoE-Chat2026.05 | 35.92 | |
| Frozen checkpointModel=DeepSeek-V2-Lite2026.05 | 35.46 | |
| Frozen checkpointModel=DeepSeek-MoE-Chat2026.05 | 33.7 |