List your company

Commonsense Question Answering on CSQA (Score, Gate Density)

0.676ScoreSelf-Pruned KV
0.674960.675230.67550.67577May 13, 2026

Evaluation results

MethodLinks
Self-Pruned KVAttention Mechanism=Self-Pruned KV, tau=0.5, Model Size=8.1B, Training Context=32k2026.050.6760.128
VanillaAttention Mechanism=Full Attention, Model Size=8.1B, Training Context=32k2026.050.675——