List your company

3-shot Evaluation on ARC, BoolQ, HSwag, LAMBADA, OBQA, PIQA, and Wino Suite

14.18Lambada PerplexityQK-Normed MLA
14.09614.66315.2315.797Jun 15, 2026

Evaluation results

MethodLinks
QK-Normed MLAshots=3-shot, checkpoints=400M/100B2026.0614.1824.6656.6962.238.9443.9219.670.8953.7546.33
QK-Clipshots=3-shot, checkpoints=400M/100B2026.0616.2822.6153.0759.5138.3242.418.270.6753.244.75