General Reasoning on MMLU-P, ARC-c, and GPQA
52.1MMLU-P ScorePOPO
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| POPOBackbone=DSR-7B, Finetuning=true, Runtime=34h2026.05 | 52.1 | 76.3 | 25.2 | 51.2 | |
| DAPOBackbone=DSR-7B, Finetuning=true, Runtime=55h2026.05 | 52 | 76.9 | 23.4 | 50.8 | |
| ARPOBackbone=DSR-7B, Finetuning=true, Runtime=31h2026.05 | 51.6 | 75.7 | 24.4 | 50.6 | |
| MoPPSBackbone=DSR-7B, Finetuning=true, Runtime=32h2026.05 | 51.6 | 76.7 | 25.4 | 51.2 | |
| GRPOBackbone=DSR-7B, Finetuning=true, Runtime=31h2026.05 | 50.9 | 75.6 | 23.8 | 50.1 | |
| DSR-7BBackbone=DSR-7B, Finetuning=false2026.05 | 50.5 | 76.1 | 19.2 | 48.6 | |
| GRPO + Meta-HintsBackbone=Qwen2.5-7B2025.10 | 50.2 | 88.8 | 30.4 | 56.5 | |
| HINTBackbone=Qwen2.5-7B2025.10 | 50.2 | 89 | 32.8 | 57.3 | |
| BREADBackbone=Qwen2.5-7B2025.10 | 49.3 | 88.2 | 30.2 | 55.9 | |
| LUFFYBackbone=Qwen2.5-7B2025.10 | 48.8 | 86 | 26.8 | 53.9 | |
| GRPOBackbone=Qwen2.5-7B2025.10 | 48 | 88 | 29.4 | 55.1 | |
| SFTBackbone=Qwen2.5-7B2025.10 | 46.2 | 85.1 | 25.6 | 52.3 | |
| VanillaBackbone=Qwen2.5-7B2025.10 | 46 | 85.3 | 25.6 | 52.3 | |
| HINTBackbone=LLaMa3.1-8B2025.10 | 39 | 75.3 | 30.4 | 48.2 | |
| GRPO + Meta-HintsBackbone=LLaMa3.1-8B2025.10 | 36.4 | 74.8 | 28 | 46.4 | |
| BREADBackbone=Qwen2.5-3B2025.10 | 36.3 | 72 | 18.2 | 42.2 | |
| GRPOBackbone=Qwen2.5-3B2025.10 | 36.2 | 74.4 | 16 | 42.2 | |
| HINTBackbone=Qwen2.5-3B2025.10 | 35.5 | 78.8 | 20.4 | 44.9 | |
| GRPO + Meta-HintsBackbone=Qwen2.5-3B2025.10 | 35 | 77.6 | 18 | 43.5 | |
| LUFFYBackbone=Qwen2.5-3B2025.10 | 34.2 | 70.2 | 15.2 | 39.9 | |
| BREADBackbone=LLaMa3.1-8B2025.10 | 33.9 | 70.4 | 27.2 | 43.8 | |
| LUFFYBackbone=LLaMa3.1-8B2025.10 | 33.3 | 66.8 | 25.5 | 41.9 | |
| GRPOBackbone=LLaMa3.1-8B2025.10 | 33 | 70 | 26.4 | 43.1 | |
| SFTBackbone=Qwen2.5-3B2025.10 | 32 | 46.4 | 11 | 29.8 | |
| VanillaBackbone=Qwen2.5-3B2025.10 | 28.8 | 44.8 | 11.4 | 28.3 | |
| SFTBackbone=LLaMa3.1-8B2025.10 | 26.5 | 52.4 | 18.3 | 32.4 | |
| GRESOBackbone=DSR-1.5B, Finetuning=true, Runtime=27h2026.05 | 24.7 | 46.7 | 26.4 | 32.6 | |
| DAPOBackbone=DSR-1.5B, Finetuning=true, Runtime=30h2026.05 | 24.5 | 46.7 | 26.8 | 32.7 | |
| POPOBackbone=DSR-1.5B, Finetuning=true, Runtime=16h2026.05 | 22.6 | 45.9 | 30.3 | 32.9 | |
| ReMixBackbone=DSR-1.5B, Finetuning=true, Runtime=16h2026.05 | 22.5 | 45.3 | 27.7 | 31.8 | |
| GRPOBackbone=DSR-1.5B, Finetuning=true, Runtime=16h2026.05 | 22.3 | 45.4 | 27.5 | 31.7 | |
| ARPOBackbone=DSR-1.5B, Finetuning=true, Runtime=16h2026.05 | 22.3 | 45.3 | 26.3 | 31.3 | |
| REPOBackbone=DSR-1.5B, Finetuning=true, Runtime=19h2026.05 | 21.9 | 44.9 | 27.4 | 31.4 | |
| MoPPSBackbone=DSR-1.5B, Finetuning=true, Runtime=17h2026.05 | 21.4 | 44.6 | 27.5 | 31.2 | |
| RRBackbone=DSR-1.5B, Finetuning=true, Runtime=14h2026.05 | 21.2 | 45.4 | 27.1 | 31.2 | |
| DSR-1.5BBackbone=DSR-1.5B, Finetuning=false2026.05 | 21.1 | 42.2 | 23 | 28.8 | |
| VanillaBackbone=LLaMa3.1-8B2025.10 | 0 | 0 | 0 | 0 |