A duration-grouped, distribution-transformed multiplicative correction layer removes local watch-time prediction bias on top of frozen rankers, improving offline MAE/XAUC and online time spent.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.IR 2years
2026 2representative citing papers
Dual-Rerank fuses autoregressive and non-autoregressive generative reranking via knowledge distillation and uses list-wise decoupled RL optimization to improve whole-page utility and cut latency in industrial video search.
citing papers explorer
-
DADF: A Distribution-Aware Debiasing Framework for Watch-Time Regression in Recommender Systems
A duration-grouped, distribution-transformed multiplicative correction layer removes local watch-time prediction bias on top of frozen rankers, improving offline MAE/XAUC and online time spent.
-
Dual-Rerank: Fusing Causality and Utility for Industrial Generative Reranking
Dual-Rerank fuses autoregressive and non-autoregressive generative reranking via knowledge distillation and uses list-wise decoupled RL optimization to improve whole-page utility and cut latency in industrial video search.