List your company

Mathematical Reasoning on GSM8K (Score & Speedup)

77.6AccuracyLSP
-1.887218.748939.38560.0211May 26, 2025Jul 12, 2025Aug 28, 2025Oct 14, 2025Nov 30, 2025Jan 16, 2026Mar 5, 2026

Evaluation results

MethodLinks
LSPModel=LLaDA-8B, Inference Strategy=LSP2026.0377.61.51
FullModel=LLaDA-8B, Inference Strategy=Full2026.0377.1—
LSPModel=Dream-7B, Inference Strategy=LSP2026.0375.41.69
FullModel=Dream-7B, Inference Strategy=Full2026.0375.3—
Llama3.2-1BForgetting Task=none2025.0519.71—
LWFForgetting Task=dental2025.0510.4—
LWFForgetting Task=mixed2025.056.95—
LWFForgetting Task=qasc2025.055.38—
LWFForgetting Task=sst52025.052.67—
LWFForgetting Task=psychol2025.051.17—