OpenThoughts Reasoning Suite (GPQA, AIME 25, AMC, Math500) (test)
59.5GPQA ScoreQwQ
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| QwQSFT training data rows=10k, Training=SFT and GRPO (RL)2025.12 | 59.5 | 15.3 | 36.5 | 58.6 | 42.5 | |
| SkillFactorySFT training data rows=10k, Training=SFT and GRPO (RL)2025.12 | 57.9 | 7.3 | 35.2 | 61.9 | 40.6 | |
| SkillFactorySFT training data rows=1k, Training=SFT and GRPO (RL)2025.12 | 56.7 | 9.7 | 37.5 | 64.6 | 42.1 | |
| RL OnlySFT training data rows=0, Training=SFT and GRPO (RL)2025.12 | 53.8 | 5.4 | 33.5 | 59.1 | 38 | |
| QwQSFT training data rows=1k, Training=SFT and GRPO (RL)2025.12 | 48.5 | 10.6 | 19.9 | 55.2 | 33.5 |