List your company

General Reasoning and Generation on GSM8K, MATH500, HumanEval, and MBPP

51.34Mean AccuracyVoidExpansion
38.100841.537944.97548.4121Jun 16, 2026

Evaluation results

MethodLinks
VoidExpansionFine-tuned=Yes, Expansion Signal=[VOID]2026.0651.34
DAEDALFine-tuned=No, Expansion Signal=[EOS]2026.0650.77
ρ-[EOS]Fine-tuned=No, Expansion Signal=[EOS]2026.0645.5
LLaDA fixed-64Fine-tuned=No, Expansion Signal=None2026.0639.38
VOID fixed-64Fine-tuned=Yes, Expansion Signal=None2026.0638.61