Coding Accuracy (avg@32) on LiveCodeBench
74.2Acc (avg@32)EqLen-GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| EqLen-GRPOBackbone=Qwen3-Next2026.04 | 74.2 | |
| LUSPOBackbone=Qwen3-Next2026.04 | 70.8 | |
| DAPOBackbone=Qwen3-Next2026.04 | 70.4 | |
| Dr. GRPOBackbone=Qwen3-Next2026.04 | 70 | |
| GSPOBackbone=Qwen3-Next2026.04 | 69.6 | |
| GRPOBackbone=Qwen3-Next2026.04 | 69.2 | |
| BaseBackbone=Qwen3-Next2026.04 | 68.7 | |
| EqLen-GRPOBackbone=Qwen3-32B2026.04 | 68.5 | |
| LUSPOBackbone=Qwen3-32B2026.04 | 63.4 | |
| GRPOParams=4B2026.05 | 63.13 | |
| DAPOBackbone=Qwen3-32B2026.04 | 63.1 | |
| Dr. GRPOBackbone=Qwen3-32B2026.04 | 62.5 | |
| GSPOBackbone=Qwen3-32B2026.04 | 61.8 | |
| GRPOBackbone=Qwen3-32B2026.04 | 61.2 | |
| BaseBackbone=Qwen3-32B2026.04 | 60.6 | |
| SODParams=1.7B2026.05 | 40.63 | |
| OPDParams=1.7B2026.05 | 32.45 | |
| SODParams=0.6B2026.05 | 27.72 | |
| SFTParams=1.7B2026.05 | 24.63 | |
| OPSDhintParams=1.7B2026.05 | 23.12 | |
| VanillaParams=1.7B2026.05 | 22.73 | |
| OPSDgtParams=1.7B2026.05 | 22.73 | |
| OPDParams=0.6B2026.05 | 22.65 | |
| GRPOParams=1.7B2026.05 | 20.7 | |
| OPSDgtParams=0.6B2026.05 | 16.73 | |
| GRPOParams=0.6B2026.05 | 15.95 | |
| VanillaParams=0.6B2026.05 | 14.89 | |
| OPSDhintParams=0.6B2026.05 | 12.65 | |
| SFTParams=0.6B2026.05 | 9.61 |