Bidirectional Counterfactual Distillation for Review-Based Recommendation
Sheng Sang, Shujie Li, Shuaiyang Li, Kang Liu, Teng Li, Wei Jia, Dan Guo, Feng Xue
Abstract
Review-based recommendation methods typically integrate multiple behaviors, including interactions, reviews, and ratings, to model user preferences. To effectively extract preference signals from diverse behaviors, some studies train multiple student models to capture distinct behavioral patterns, and leverage online distillation to facilitate collaborative learning among them. However, we argue that these techniques suffer from bias contamination from rating distributions and feature homogenization during cross-behavior knowledge transfer: (1) Rating distribution bias, arising from non-uniform historical ratings, propagates across behaviors through distillation, contaminating the true preference representations of other behaviors. (2) Static distillation strategies often lead to homogenized behavioral features, hindering the learning of behavior-specific preferences. To address these issues, we propose a novel Bidirectional Counterfactual Distillation (BiCoD) framework for review-based recommendation. In BiCoD, we first design an adversarial counterfactual distillation module to suppress the impact of non-uniform rating distributions on distillation, thereby preventing it from contaminating the user
BibTeX
@inproceedings{aaai2026_bidirectionalcou,
title = {Bidirectional Counterfactual Distillation for Review-Based Recommendation},
author = {Sheng Sang and Shujie Li and Shuaiyang Li and Kang Liu and Teng Li and Wei Jia and Dan Guo and Feng Xue},
booktitle = {AAAI 2026},
year = {2026}
}