List your company

Zero-shot Reasoning and Knowledge Evaluation on Reasoning and Common-Knowledge Tasks

80.02HellaSWAGSDM
60.5265.582570.64575.7075Jul 8, 2026

Evaluation results

MethodLinks
SDMScale=8B (L13), Local layer=SWA, Global layer=SDM2026.0780.0275.378.5253.0580.4744.662.0544.9770.667.1359.1124.3925.9357.8128.6656.84
FullAttnScale=8B (L13), Local layer=SWA, Global layer=FullAttn2026.0779.3373.6478.150.8279.9843.864.8347.9468.8871.8355.2324.3922.9458.7329.3456.65
GDNScale=8B (L13), Local layer=SWA, Global layer=GDN2026.0779.173.2479.1552.2781.014360.1743.466.8374.9855.3618.2922.657.2428.8155.7
SDMScale=1.4B (L08), Local layer=SWA, Global layer=SDM2026.0762.0462.126334.7674.273649.6537.7620.7262.7826.619.768.2327.242.7338.51
FullAttnScale=1.4B (L08), Local layer=SWA, Global layer=FullAttn2026.0761.4760.5461.5633.5673.8335.853.5539.6219.5762.4224.248.547.724.542.8137.98
Mamba2Scale=1.4B (L08), Local layer=SWA, Global layer=Mamba22026.0761.3162.6762.8834.0873.6736.450.2836.2520.3162.7523.489.768.0925.033.0338
GDNScale=1.4B (L08), Local layer=SWA, Global layer=GDN2026.0761.2761.5662.4934.5973.8336.850.4236.6819.3363.0323.688.547.7626.022.9637.93