KuaiFormer is a Transformer-based retrieval model with multi-interest query tokens and adaptive sequence compression, deployed at Kuaishou with reported online watch-time gains of 0.126% to 0.411%.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
KuaiFormer: Transformer-Based Retrieval at Kuaishou
KuaiFormer is a Transformer-based retrieval model with multi-interest query tokens and adaptive sequence compression, deployed at Kuaishou with reported online watch-time gains of 0.126% to 0.411%.