Research map / Architectures and efficiency / Attention and transformers
Length-Adaptive Interest Network for Balancing Long and Short Sequence Modeling in CTR Prediction
Zhicheng Zhang, Zhaocheng Du, Jieming Zhu, Jiwei Tang, Fengyuan Lu, Wang Jiaheng, Song-Li Wu, Qianhui Zhu, Jingyu Li, Hai-Tao Zheng, 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 LAIN to backbones | DIN/DIEN/SIM/SDIM/TWIN · EBNeRD-small | AUC: e.g. TWIN 0.6993 -> 0.7074 | Improved |
| add LAIN to backbones | DIN · MicroVideo1.7M | GAUC: 0.7023 -> 0.7019 (AUC 0.7093 -> 0.7101) | No clear effect |
| add LAIN to DIEN | DIEN · EBNeRD-small | GAUC: 0.7090 -> 0.7078 | Hurt |
| add LAIN to TWIN, short sequences | TWIN · MicroVideo1.7M (0-100 length) | AUC: 0.7108 -> 0.7184 | Improved |
| add LAIN to TWIN, long sequences | TWIN · MicroVideo1.7M (200+ length) | AUC: 0.7125 -> 0.7238 | Improved |
| add length-modulated attention (ablation) | LAIN on TWIN · MicroVideo1.7M | AUC: 0.7189 -> 0.7233 | Improved |
| add query-key conditioning (ablation) | LAIN on TWIN · MicroVideo1.7M | AUC: 0.7195 -> 0.7233 | Improved |
| add temperature scaling (ablation) | LAIN on TWIN · MicroVideo1.7M | AUC: 0.7212 -> 0.7233 | Improved |
| add length-conditioned prompting (ablation) | LAIN on TWIN · MicroVideo1.7M | AUC: 0.7228 -> 0.7233 | Improved |
| apply LAIN to short-term branch (ablation) | LAIN on TWIN · MicroVideo1.7M | logloss: 0.4137 -> 0.4097 | Improved |
| add LAIN lowers attention Gini | baseline CTR models · EBNeRD-small | Gini coefficient: 0.346 -> 0.318 (short, L=1000 baseline) | Improved |
| vary prompt/Fourier/hidden dims (12 configs) (ablation) | LAIN on TWIN · MicroVideo1.7M | AUC: 0.7216 to 0.7248 vs baseline 0.7158 | No clear effect |
Closest papers by meaning
- Iterative Sparse Attention for Long-sequence Recommendation (AAAI 2025 · similarity 0.88)
- Long-Sequence Recommendation Models Need Decoupled Embeddings (ICLR 2025 · similarity 0.86)
- UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence (NeurIPS 2026 · similarity 0.86)
- MARS: Multi-resolution Adaptive Routing for Sequential Recommendation (NeurIPS 2026 · similarity 0.86)
- LS-TGNN: Long and Short-Term Temporal Graph Neural Network for Session-Based Recommendation (AAAI 2025 · similarity 0.86)
- Understanding and Improving Length Generalization in Recurrent Models (ICML 2025 · 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.
