Knowledge & Instruction Following on GPQA Diamond (Accuracy)
80.3AccuracyQwen3.5
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3.5Params=4B, Sampling mode=AR2026.06 | 80.3 | |
| FLAREParams=4B, Sampling mode=Diffusion-Trust2026.06 | 64.65 | |
| FLAREParams=4B, Sampling mode=AR-Trust2026.06 | 63.64 | |
| Qwen3.5Params=2B, Sampling mode=AR2026.06 | 62.12 | |
| SDARParams=8B, Sampling mode=Diffusion2026.06 | 40.2 | |
| FLAREParams=2B, Sampling mode=AR-Trust2026.06 | 37.37 | |
| FLAREParams=2B, Sampling mode=Diffusion-Trust2026.06 | 35.35 | |
| SDARParams=4B, Sampling mode=Diffusion2026.06 | 33 | |
| SDARParams=1.7B, Sampling mode=Diffusion2026.06 | 29.8 |