Research map / Label-efficient and robust learning / Knowledge distillation
Bidirectional Counterfactual Distillation for Review-Based Recommendation
Sheng Sang, Shujie Li, Shuaiyang Li, Kang Liu, Teng Li, Wei Jia, Dan Guo, Feng Xue
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add counterfactual distillation (subtract rating-only counterfactual features) (ablation) | BiCoD · Art Crafts | MSE: 0.6306 -> 0.5963 | Improved |
| add counterfactual distillation (ablation) | BiCoD · Digital Music | MSE: 0.2620 -> 0.2403 | Improved |
| add adversarial learning of gamma (ablation) | BiCoD · Art Crafts | MSE: 0.6254 -> 0.5963 | Improved |
| learn gamma adversarially instead of fixing it at 0.5 (ablation) | BiCoD · Software | MSE: 0.9967 -> 0.9798 | Improved |
| apply counterfactual and adversarial parts in both FD and RD rather than one direction (ablation) | BiCoD · Yelp | MSE: 1.1144 -> 1.1193 | No clear effect |
| add stage-aware dynamic scheduling of distillation (ablation) | BiCoD · four datasets | MSE | Improved |
| add forward distillation (ablation) | BiCoD · Digital Music, Luxury Beauty, Software, Yelp | MSE | Improved |
| add reverse distillation (ablation) | BiCoD · Digital Music, Luxury Beauty, Software, Yelp | MSE | Improved |
| add online distillation (ablation) | BiCoD · four datasets | MSE | Improved |
| cosine-scheduled distillation strength instead of fixed phi (ablation) | BiCoD · Art Crafts, Digital Music | MSE | Improved |
| BiCoD over best baseline | BiCoD · Yelp | MSE: 1.1596 -> 1.1193 | Improved |
| BiCoD over best baseline | BiCoD · Digital Music | MSE: 0.2753 -> 0.2403 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- EMKD — EM-KD: Distilling Efficient Multimodal Large Language Model with Unbalanced Vision Tokens (AAAI 2026) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Alleviating Dual Biases in Recommendation (Student Abstract) (AAAI 2025 · similarity 0.86)
- Social Recommendation via Graph-Level Counterfactual Augmentation (AAAI 2025 · similarity 0.85)
- Disentangled Modeling of Preferences and Social Influence for Group Recommendation (AAAI 2025 · similarity 0.85)
- Distributional Preference Learning: Understanding and Accounting for Hidden Context in RLHF (ICLR 2024 · similarity 0.85)
- Tree of Preferences for Diversified Recommendation (NeurIPS 2025 · similarity 0.85)
- Multiple Purchase Chains with Negative Transfer Elimination for Multi-Behavior Recommendation (AAAI 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
