General Reasoning on GPQA (Accuracy)
41.4GPQA AccuracySDAR-8B-Chat
Evaluation Results
| Method | Links | |
|---|---|---|
| SDAR-8B-Chatgeneration length=optimal, block length=optimal, denoising steps=optimal2026.04 | 41.4 | |
| SDAR-30B-A3Bgeneration length=optimal, block length=optimal, denoising steps=optimal2026.04 | 36.36 | |
| LLaDA2.0-minigeneration length=optimal, block length=optimal, denoising steps=optimal2026.04 | 34.34 | |
| LLaDA2.1-minigeneration length=optimal, block length=optimal, denoising steps=optimal2026.04 | 34.34 | |
| LLaDA-8B-Instructgeneration length=optimal, block length=optimal, denoising steps=optimal2026.04 | 31.82 | |
| Dream-7B-Instructgeneration length=optimal, block length=optimal, denoising steps=optimal2026.04 | 26.77 |