Math & Science on GPQA Diamond
28.79Accuracy (Pass@1)AdaBlock
Evaluation Results
| Method | Links | |
|---|---|---|
| AdaBlockToken budget=64, Number of shots=02026.04 | 28.79 | |
| VSBToken budget=64, Number of shots=0, Backbone=LLaDA-8B2026.04 | 28.79 | |
| DAEDALToken budget=64, Number of shots=02026.04 | 28.28 | |
| VSBToken budget=64, Number of shots=0, Backbone=LLaDA-1.52026.04 | 28.28 | |
| LLaDA-1.5Token budget=64, Number of shots=02026.04 | 27.78 | |
| LLaDA-8BToken budget=64, Number of shots=02026.04 | 26.52 | |
| Blockwise-SFTToken budget=64, Number of shots=02026.04 | 26.26 | |
| dLLM-VarToken budget=64, Number of shots=02026.04 | 23.73 | |
| D2FToken budget=64, Number of shots=02026.04 | 20.2 |