User embeddings are quantized into hierarchical codes, and listwise ranking losses are applied within each code-defined user group to create harder negatives without retrieval infrastructure.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Hierarchical Group-wise Ranking Framework for Recommendation Models
User embeddings are quantized into hierarchical codes, and listwise ranking losses are applied within each code-defined user group to create harder negatives without retrieval infrastructure.