List your company

Natural Language Understanding on SuperGLUE (BoolQ, WiC, RTE, CB, ReCoRD, WSC, MultiRC, COPA)

59.36BoolQ AccuracySCAPE (d = 0.1)
51.601653.615855.6357.6442Jul 2, 2026

Evaluation results

MethodLinks
SCAPE (d = 0.1)Protocol=Zero-shot, Model=GPT-345M, d=0.12026.0759.3651.7252.3532.1477.3841.3555.187355.31
SCAPEBackbone=Llama-500M, Evaluation Protocol=Zero-shot, Sparsity (d)=0.012026.0757.925052.7132.1469.6636.5456.486852.93
SCAPESparsity (d)=0.01, Backbone=Llama-500M, Evaluation Protocol=Zero-shot2026.0757.925052.7132.1469.6636.5456.486852.93
ADAMWBackbone=Llama-500M, Evaluation Protocol=Zero-shot, Communication Strategy=DENSE all-reduce2026.0757.6549.6953.4335.7169.8141.3557.036553.71
ADAMWSynchronization=DENSE all-reduce, Backbone=Llama-500M, Evaluation Protocol=Zero-shot2026.0757.6549.6953.4335.7169.8141.3557.036553.71
ADAMSBackbone=Llama-500M, Evaluation Protocol=Zero-shot, Communication Strategy=DENSE all-reduce2026.0757.5550.6353.0741.0769.5434.6249.866853.04
ADAMSSynchronization=DENSE all-reduce, Backbone=Llama-500M, Evaluation Protocol=Zero-shot2026.0757.5550.6353.0741.0769.5434.6249.866853.04
SCAPE (d = 0.01)Protocol=Zero-shot, Model=GPT-345M, d=0.012026.0756.4249.8450.1835.7176.736.5454.836953.65
ADAMS (DENSE all-reduce)Protocol=Zero-shot, Model=GPT-345M, Strategy=DENSE all-reduce2026.0755.725053.435078.0636.5454.027055.97
ADAMW (DENSE all-reduce)Protocol=Zero-shot, Model=GPT-345M, Strategy=DENSE all-reduce2026.0753.1249.5352.7128.5776.2459.6257.247456.38
SCAPEBackbone=Llama-500M, Evaluation Protocol=Zero-shot, Sparsity (d)=0.12026.0751.949.0657.428.577039.4253.116852.18
SCAPESparsity (d)=0.1, Backbone=Llama-500M, Evaluation Protocol=Zero-shot2026.0751.949.0657.428.577039.4253.116852.18