List your company

Zero-shot Language Evaluation on WikiText2, BoolQ, PIQA, HellaSwag, WinoGrande, ARC, OBQA, and MTQA (test)

7.43WikiText2 PerplexitySimDiff (MSSD)
-428.91842,516.43335,461.7858,407.1367Apr 21, 2026

Evaluation results

MethodLinks
SimDiff (MSSD)Model=LLaMA2-13B, Layers=82026.047.4362.2374.2152.2367.7262.8838.3127.226.651.42
SimDiff (MASD)Model=LLaMA2-13B, Layers=82026.047.4562.4274.9252.4466.7760.7337.827.827.2451.26
ShortGPTModel=LLaMA2-13B, Layers=82026.048.362.1773.8850.9965.4361.2436.3526.626.1650.35
SimDiffModel=LLaMA2-13B, Layers=102026.048.8762.2672.0350.1867.0158.2136.1822.625.3349.22
SimDiff (MASD)Model=LLaMA2-13B, Layers=122026.0410.4462.4571.8247.366.6154.8833.9620.824.3247.77
SimDiff (MSSD)Model=LLaMA2-13B, Layers=122026.0411.562.4569.8646.7565.1952.9532.172124.8246.9
ShortGPTModel=Qwen2.5-3B, Layers=82026.0413.5862.7574.1644.2653.3570.7534.1329.629.9849.87
SimDiffModel=Qwen2.5-3B, Layers=82026.0413.5862.7574.1644.2653.3570.7534.1329.629.9849.87
SimDiffModel=Qwen2.5-1.5B, Layers=82026.0418.4262.2370.1337.3653.5962.1626.8825.424.6245.3
ShortGPTModel=LLaMA2-13B, Layers=102026.0420.0362.6371.2747.2962.9853.4534.323.824.0547.47
SimDiffModel=Qwen2.5-3B, Layers=122026.0423.1446.1868.6635.9452.1756.2724.9122.424.8641.42
SimDiffModel=Qwen2.5-1.5B, Layers=102026.0426.3459.9767.8534.7753.6760.2325.4323.225.7643.86
ShortGPTModel=Qwen2.5-3B, Layers=102026.0426.3459.9767.8534.7753.6760.2325.4323.225.7643.86
SimDiffModel=Qwen2.5-3B, Layers=102026.0426.3459.9767.8534.7753.6760.2325.4323.225.7643.86
ShortGPTModel=LLaMA2-13B, Layers=122026.0434.7162.3968.4445.4962.2748.0231.9120.224.3245.38
ShortGPTModel=Qwen2.5-1.5B, Layers=82026.0437.761.6564.9135.6250.5953.2827.9922.826.242.88
SimDiffModel=Qwen2.5-0.5B, Layers=82026.043843.1863.1730.2650.4348.8220.4819.622.4537.3
SimDiffModel=Qwen2.5-1.5B, Layers=122026.0444.3439.1764.0430.8952.2548.7422.118.222.7137.26
ShortGPTModel=Qwen2.5-3B, Layers=122026.0444.3439.1764.0430.8952.2548.7422.118.222.7137.26
ShortGPTModel=Qwen2.5-1.5B, Layers=102026.0464.4555.5764.1532.953.4347.5222.352323.4840.3
SimDiffModel=Qwen2.5-0.5B, Layers=102026.0487.741.1960.9428.9751.3844.6118.521722.5835.65
ShortGPTModel=Qwen2.5-0.5B, Layers=82026.04110.2347.1659.930.0751.741.0820.6518.422.3536.41
SimDiff (MSSD)Model=Qwen2.5-0.5B, Layers=122026.04208.2338.4757.7828.2650.2839.9420.2215.420.7733.89
SimDiff (MASD)Model=Qwen2.5-0.5B, Layers=122026.04399.3940.2458.1627.549.5737.2519.81720.1333.71
ShortGPTModel=Qwen2.5-1.5B, Layers=122026.04634.9446.6159.2529.3450.2836.2423.2918.220.835.5
ShortGPTModel=Qwen2.5-0.5B, Layers=102026.04858.5242.9758.2728.2849.836.9521.251621.9434.43
ShortGPTModel=Qwen2.5-0.5B, Layers=122026.0410,916.1437.855.5526.2549.1729.8821.6717.619.5632.19