Mathematical Reasoning on DAPO-Math In-Distribution (test)
84.1Avg@16FwdKL
Evaluation Results
| Method | Links | |
|---|---|---|
| FwdKLBackbone=Qwen3-8B2026.07 | 84.1 | |
| GEOSDBackbone=Qwen3-8B2026.07 | 82.7 | |
| SkewKLBackbone=Qwen3-8B2026.07 | 82 | |
| RevKLBackbone=Qwen3-8B2026.07 | 81.8 | |
| TrOPDBackbone=Qwen3-8B2026.07 | 81.6 | |
| GRPOBackbone=Qwen3-8B2026.07 | 80.9 | |
| SFTBackbone=Qwen3-8B2026.07 | 80.8 | |
| JSDBackbone=Qwen3-8B2026.07 | 80.7 | |
| RevKLBackbone=Olmo-3-7B-Think2026.07 | 77.2 | |
| TIPBackbone=Qwen3-8B2026.07 | 76.9 | |
| GEOSDBackbone=Olmo-3-7B-Think2026.07 | 75.5 | |
| FwdKLBackbone=Olmo-3-7B-Think2026.07 | 75.1 | |
| TrOPDBackbone=Olmo-3-7B-Think2026.07 | 74.9 | |
| GRPOBackbone=Olmo-3-7B-Think2026.07 | 74.7 | |
| SFTBackbone=Olmo-3-7B-Think2026.07 | 73.8 | |
| TIPBackbone=Olmo-3-7B-Think2026.07 | 73.6 | |
| JSDBackbone=Olmo-3-7B-Think2026.07 | 73.1 | |
| SkewKLBackbone=Olmo-3-7B-Think2026.07 | 70.6 | |
| FwdKLBackbone=DS-R1-Llama-8B2026.07 | 70.1 | |
| GEOSDBackbone=DS-R1-Llama-8B2026.07 | 69.4 | |
| RevKLBackbone=DS-R1-Llama-8B2026.07 | 68.3 | |
| JSDBackbone=DS-R1-Llama-8B2026.07 | 66.7 | |
| TIPBackbone=DS-R1-Llama-8B2026.07 | 66 | |
| TrOPDBackbone=DS-R1-Llama-8B2026.07 | 65.2 | |
| SkewKLBackbone=DS-R1-Llama-8B2026.07 | 65.1 | |
| GRPOBackbone=DS-R1-Llama-8B2026.07 | 62.5 | |
| SFTBackbone=DS-R1-Llama-8B2026.07 | 61.8 | |
| π0Backbone=Qwen3-8B2026.07 | 61.4 | |
| π0Backbone=Olmo-3-7B-Think2026.07 | 53.3 | |
| π0Backbone=DS-R1-Llama-8B2026.07 | 49.6 |