Science Reasoning on GPQA 5-shot
33.3AccuracyBaseline
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| BaselineBase Model=Dream-7B-Instruct2026.05 | 33.3 | 0.5 | 256 | 1 | |
| KLASSBase Model=Dream-7B-Instruct2026.05 | 32.6 | 11.6 | 8 | 24.8 | |
| LEAPBase Model=Dream-7B-Instruct2026.05 | 32.4 | 22 | 4 | 62 | |
| LEAPBase Model=LLaDA-8B-Instruct2026.05 | 32.1 | 12.1 | 17 | 13.3 | |
| Conf-BasedBase Model=Dream-7B-Instruct2026.05 | 32.1 | 22.3 | 5 | 49.6 | |
| Conf-BasedBase Model=LLaDA-8B-Instruct2026.05 | 31.9 | 8.3 | 28 | 9.1 | |
| BaselineBase Model=LLaDA-8B-Instruct2026.05 | 30.6 | 4 | 256 | 1 | |
| LoPABase Model=Dream-7B-Instruct2026.05 | 30.6 | 7.6 | 6 | 20.7 | |
| KLASSBase Model=LLaDA-8B-Instruct2026.05 | 29 | 10.1 | 124 | 2.1 |