General Performance Evaluation on Aggregate (IFEval, GPQA, LCB, Arena-Hard, CW, MT-Bench, WildBench)
63.69Average ScoreSPARD
Evaluation Results
| Method | Links | |
|---|---|---|
| SPARDBackbone=Qwen3-8B2026.04 | 63.69 | |
| + GRPOavgBackbone=Qwen3-8B2026.04 | 62.17 | |
| + GRPOimpBackbone=Qwen3-8B2026.04 | 61.74 | |
| + DPOBackbone=Qwen3-8B2026.04 | 61.2 | |
| + GRPOrmBackbone=Qwen3-8B2026.04 | 61.15 | |
| BaseBackbone=Qwen3-8B2026.04 | 60.56 | |
| SPARDBackbone=Qwen2.5-7B-Instruct2026.04 | 50.03 | |
| + SFTBackbone=Qwen3-8B2026.04 | 49.6 | |
| + GRPOavgBackbone=Qwen2.5-7B-Instruct2026.04 | 48.46 | |
| + DPOBackbone=Qwen2.5-7B-Instruct2026.04 | 47.55 | |
| + GRPOrmBackbone=Qwen2.5-7B-Instruct2026.04 | 47.06 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.04 | 46.12 | |
| + GRPOimpBackbone=Qwen2.5-7B-Instruct2026.04 | 45.59 | |
| + SFTBackbone=Qwen2.5-7B-Instruct2026.04 | 40.5 |