Research map / Label-efficient and robust learning / Recognition and long-tail
REACTION: Parameter-Efficient Learning for Recommendation
Song-Li Wu, Zhaocheng Du, Qinglin Jia, Zhenhua Dong
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 dynamic tower fusion (DTF) (ablation) | REACTION baseline towers · Criteo | AUC: 0.8118 -> 0.8154 | Improved |
| add dynamic tower fusion (DTF) (ablation) | REACTION baseline towers · UGC | AUC: 0.8022 -> 0.8040 | Improved |
| add dynamic tower fusion (DTF) (ablation) | REACTION baseline towers · MovieLens | AUC: 0.9543 -> 0.9579 | Improved |
| add dynamic tower fusion (DTF) (ablation) | REACTION baseline towers · Avazu | AUC: 0.7833 -> 0.7872 | Improved |
| add entropy regularization instead of DTF (ablation) | REACTION baseline towers · Criteo | AUC: 0.8118 -> 0.8122 | No clear effect |
| add knowledge distillation instead of DTF (ablation) | REACTION baseline towers · UGC | AUC: 0.8022 -> 0.8024 | No clear effect |
| add adaptive feature extraction (AFE) instead of keeping all fields (ablation) | REACTION baseline towers · UGC | AUC: 0.8008 -> 0.8030 | Improved |
| use PCA to drop feature fields (ablation) | REACTION baseline towers · UGC | AUC: 0.8008 -> 0.7988 | Hurt |
| use IDARTS to drop feature fields (ablation) | REACTION baseline towers · UGC | AUC: 0.8008 -> 0.7962 | Hurt |
| use TayFCS instead of all fields (ablation) | REACTION baseline towers · UGC | AUC: 0.8008 -> 0.8023 | Improved |
| increase tower cohort size from 1 to 8 (ablation) | REACTION baseline towers · Criteo | AUC: 0.8154 -> 0.8212 | Improved |
| increase tower cohort size from 1 to 8 (ablation) | REACTION baseline towers · UGC | AUC: 0.8057 -> 0.8115 | 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.
- FinalMLP — FinalMLP: An Enhanced Two-Stream MLP Model for CTR Prediction (AAAI 2023) · better on 4 of 4 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
- Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations (ICML 2024 · similarity 0.89)
- Infinite Recommendation Networks: A Data-Centric Approach (NeurIPS 2022 · similarity 0.89)
- On Softmax Direct Preference Optimization for Recommendation (NeurIPS 2024 · similarity 0.87)
- ContextGNN: Beyond Two-Tower Recommendation Systems (ICLR 2025 · similarity 0.86)
- Future Sight and Tough Fights: Revolutionizing Sequential Recommendation with FENRec (AAAI 2025 · similarity 0.86)
- CURE4Rec: A Benchmark for Recommendation Unlearning with Deeper Influence (NeurIPS 2024 · similarity 0.86)
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.
