A model-agnostic causal plug-in improves multi-behavior recommenders via backdoor adjustment on user/item bias proxies, MoE aggregation of auxiliaries, and bias-aware contrastive alignment.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MCLMR: A Model-Agnostic Causal Learning Framework for Multi-Behavior Recommendation
A model-agnostic causal plug-in improves multi-behavior recommenders via backdoor adjustment on user/item bias proxies, MoE aggregation of auxiliaries, and bias-aware contrastive alignment.