Question Answering on GPQA Think (test)
61.1AccuracyQwen3-8B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-8BTraining=RL, Mode=Mid-Think2026.01 | 61.1 | 8,257 | 55,272 | |
| Qwen3-8BTraining=No, Mode=Untrained2026.01 | 60.9 | 9,008 | 71,817 | |
| Qwen3-8BTraining=RL, Mode=No-Think2026.01 | 59.6 | 9,047 | 59,419 | |
| Qwen3-8BTraining=RL, Mode=Think2026.01 | 58.5 | 9,413 | 52,228 | |
| Qwen3-4BTraining=RL, Mode=Mid-Think2026.01 | 55.3 | 8,696 | 60,790 | |
| Qwen3-4BTraining=No, Mode=Untrained2026.01 | 53.2 | 8,729 | 72,509 | |
| Qwen3-4BTraining=RL, Mode=No-Think2026.01 | 48.7 | 10,774 | 82,568 | |
| Qwen3-4BTraining=RL, Mode=Think2026.01 | 47.3 | 10,481 | 58,476 |