List your company

Commonsense Reasoning on HellaSwag (200-sample subset)

67AccuracyLlama
45.1650.8356.562.17Jun 11, 2026

Evaluation results

MethodLinks
LlamaSetting=CPU2026.0667
LlamaSetting=NPU2026.0660
LLaDASetting=CPU2026.0651
LLaDA (Ablation)Setting=CPU2026.0651
LLaDA (Ablation)Setting=NPU2026.0650
LLaDA llada.cppSetting=llada.cpp2026.0649.5
LLaDA (Ablation)Setting=NPU + MBSD + DPPR2026.0649.5
LLaDA (Ablation)Setting=NPU + MBSD2026.0646