List your company

Physical Commonsense Reasoning on PIQA (Score and Speedup)

86.4ScoreLSP
80.6882.16583.6585.135Mar 5, 2026

Evaluation results

MethodLinks
LSPModel=Dream-7B, Inference Strategy=LSP2026.0386.42.31
FullModel=Dream-7B, Inference Strategy=Full2026.0386.1—
LSPModel=LLaDA-8B, Inference Strategy=LSP2026.0382.12.02
FullModel=LLaDA-8B, Inference Strategy=Full2026.0380.9—