List your company
ResearchFormerly WizWand

Code Generation on HEval

68.29Pass@1Temporal Guidance
18.827631.668844.5157.3512Jan 29, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Temporal GuidanceBackbone=Qwen3-8B, Alpha=0.12026.0168.29———
Temporal GuidanceBackbone=Qwen3-8B, Alpha=0.22026.0167.07———
DoLaBackbone=Qwen3-8B, Penalty=1.02026.0166.46———
DoLaBackbone=Qwen3-8B, Penalty=1.22026.0166.46———
Temporal GuidanceBackbone=Qwen3-8B, Alpha=0.42026.0165.24———
Temporal GuidanceBackbone=Qwen3-8B, Alpha=0.32026.0164.63———
Temporal GuidanceBackbone=Qwen3-8B, Alpha=0.52026.0164.63———
Baseline (Greedy)Backbone=Qwen3-8B2026.0163.41———
Contrastive DecodingBackbone=Qwen3-8B, Amateur Model=Qwen3-1.7B2026.0161.58———
Contrastive DecodingBackbone=Qwen3-8B, Amateur Model=Qwen3-0.6B2026.0153.65———
Temporal GuidanceBackbone=Qwen3-1.7B, Alpha=0.22026.0142.07———
Temporal GuidanceBackbone=Qwen3-1.7B, Alpha=0.12026.0141.46———
Baseline (Greedy)Backbone=Qwen3-1.7B2026.0140.24———
Contrastive DecodingBackbone=Qwen3-1.7B, Amateur Model=Qwen3-0.6B2026.0139.02———
Temporal GuidanceBackbone=Llama-3.2-3B, Alpha=0.12026.0132.32———
Temporal GuidanceBackbone=Llama-3.2-3B, Alpha=0.22026.0131.71———
DoLaBackbone=Llama-3.2-3B, Penalty=1.22026.0129.27———
DoLaBackbone=Qwen3-1.7B, Penalty=1.22026.0128.05———
Baseline (Greedy)Backbone=Llama-3.2-3B2026.0127.44———
DoLaBackbone=Llama-3.2-3B, Penalty=1.02026.0126.22———
DoLaBackbone=Qwen3-1.7B, Penalty=1.02026.0120.73———
DAEDALFine-tuned=No, Expansion Signal=[EOS], Initial canvas=short2026.06—44.51254.661
LLaDA fixed-64Fine-tuned=No, Expansion Signal=None, Initial canvas=short2026.06—29.27——
VOID fixed-64Fine-tuned=Yes, Expansion Signal=None, Initial canvas=short2026.06—36.59——
VoidExpansionFine-tuned=Yes, Expansion Signal=[VOID], Initial canvas=short2026.06—42.0773.065.59
ρ-[EOS]Fine-tuned=No, Expansion Signal=[EOS], Initial canvas=short2026.06—40.24191.521.34