List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM8K (Acc@1)

83.04Accuracy@1dOPSD
56.852863.651470.4577.2486Jul 5, 2026
Updated 4mo ago

Evaluation results

MethodLinks
dOPSDBase Model=Dream-7B-Instruct2026.0783.04
BaseBase Model=Dream-7B-Instruct2026.0781.41
GRPOBase Model=Dream-7B-Instruct2026.0781.28
SFTBase Model=Dream-7B-Instruct2026.0780.44
OPSD (answer-only)Base Model=Dream-7B-Instruct2026.0780.34
dOPSDBase Model=LLaDA-8B-Instruct2026.0772.87
OPSD (answer-only)Base Model=LLaDA-8B-Instruct2026.0771.47
BaseBase Model=LLaDA-8B-Instruct2026.0771.23
GRPOBase Model=LLaDA-8B-Instruct2026.0770.45
SFTBase Model=LLaDA-8B-Instruct2026.0769.45
OPSD (full-solution)Base Model=Dream-7B-Instruct2026.0767.25
OPSD (full-solution)Base Model=LLaDA-8B-Instruct2026.0757.86