List your company

Code Generation on HumanEval+ (Accuracy)

81.5AccuracyDiffMAS
31.37244.38657.470.414Apr 23, 2026

Evaluation results

MethodLinks
DiffMASModel=Qwen3-8B, Communication Category=Trained Latent Communication2026.0481.5
TextMASModel=Qwen3-8B, Communication Category=Text Communication2026.0480.7
LatentMASModel=Qwen3-8B, Communication Category=Training-free Latent Communication2026.0480.7
DiffMASModel=Qwen3-4B, Communication Category=Trained Latent Communication2026.0479.8
LatentMASModel=Qwen3-4B, Communication Category=Training-free Latent Communication2026.0478.9
TextMASModel=Qwen3-4B, Communication Category=Text Communication2026.0476.3
SingleModel=Qwen3-4B, Communication Category=Text Communication2026.0474.5
SingleModel=Qwen3-8B, Communication Category=Text Communication2026.0474.5
C2CModel=Qwen3-8B, Communication Category=Trained Latent Communication2026.0458.8
C2CModel=Qwen3-4B, Communication Category=Trained Latent Communication2026.0433.3