List your company

Zero-shot Natural Language Understanding and Reasoning on MMLU, ARC-c, HellaSwag, BOOLQ, PIQA, WinoGrande

69.72Average Score (Zero-shot)Dense
46.99652.895558.79564.6945Feb 18, 2026

Evaluation results

MethodLinks
DenseModel=Mistral-7B, Sparsity=0%2026.0269.7262.6550.3461.2483.6480.5873.8
DenseModel=LLaMA2-13B, Sparsity=0%2026.0265.9355.1348.4660.0680.6179.0572.22
DenseModel=Qwen1.5-7B, Sparsity=0%2026.0265.7860.5340.6157.882.5178.474.9
DenseModel=LLaMA2-7B, Sparsity=0%2026.0262.7245.8443.4357.1377.7478.0774.11
ActTailModel=LLaMA2-13B, Sparsity=70%2026.0262.1836.8840.2749.5572.5174.163.77
ActTailModel=Mistral-7B, Sparsity=70%2026.0259.4233.234.3946.7867.4973.2364.25
TEALTop-KModel=Mistral-7B, Sparsity=70%2026.0259.1730.3734.347.1767.873.2362.67
ActTailModel=LLaMA2-7B, Sparsity=70%2026.0258.0627.3530.9744.0766.9772.2556.12
TEALTop-KModel=LLaMA2-13B, Sparsity=70%2026.0258.0630.6333.9646.5168.2372.861.4
TEALTop-KModel=LLaMA2-7B, Sparsity=70%2026.0256.0226.5328.541.8465.6670.6750.36
ActTailModel=Qwen1.5-7B, Sparsity=70%2026.0255.8536.53244.6669.6970.7858.25
TEALTop-KModel=Qwen1.5-7B, Sparsity=70%2026.0254.4831.9931.2343.2467.0671.1153.04
TEALTop-KModel=Qwen1.5-7B, Sparsity=80%2026.0253.3225.8419.6230.7260.7364.1550.99
ActTailModel=LLaMA2-13B, Sparsity=80%2026.0252.6126.6125.5136.2364.1663.2252.93
ActTailModel=Qwen1.5-7B, Sparsity=80%2026.0252.4725.3221.7630.7760.461.3750.91
ActTailModel=LLaMA2-7B, Sparsity=80%2026.0249.2725.3221.0830.0956.9461.9251.22
ActTailModel=Mistral-7B, Sparsity=80%2026.0249.1524.721.1632.151.7263.8750.91
TEALTop-KModel=LLaMA2-13B, Sparsity=80%2026.0248.7324.9520.2230.6251.4763.3851.78
TEALTop-KModel=Mistral-7B, Sparsity=80%2026.0248.1224.3718.7731.6444.4663.651.46
TEALTop-KModel=LLaMA2-7B, Sparsity=80%2026.0247.8723.5819.6228.9744.6260.548.78