List your company

Paged Decode Attention Serving on Bimodal B8 trace

1.299GPU RatioPersistentKV
1.234051.2665251.2991.331475Jun 25, 2026

Evaluation results

MethodLinks
PersistentKVSelected route=PersistentKV workqueue, split 20, Page size=16, Hq=32, Hkv=8, d=128, Precision=FP162026.061.2991.2656.1044.201