2026-11-09 ~ 11-15 · Personalization
Generative · Personalization · AI Search · Infra · Embodied AI
- 1. Request-Only Optimization for Recommendation Systems
Recommendation training samples are usually cut per impression. One request sends the user several items, each impression becomes a sample, and the same user's features and long behavior sequence are copied into every…
Appeared in: 11-13
- 2. STCRank: Spatio-temporal Collaborative Ranking for Interactive Recommender System at Kuaishou E-shop
In a full-screen e-commerce feed where each swipe down brings the next item, the two ranking objectives of whether the user buys and whether they keep swiping fight each other…
Appeared in: 11-10
- 3. SepLRM: A Foundation Model Paradigm for Recommendation · 📖 Deep dive
On recommendation platforms, large models for different stages and business lines each encode the same long user history, and this repeated cost gets harder to justify as models and sequences multiply…
Appeared in: 11-11
- 4. Exploiting Feature-invariant Scaling Laws for Efficient A/B Experimentation in Recommendation Systems · 📖 Deep dive
Add a feature or change the architecture of a recommendation model, and a small-scale experiment may say it does nothing, while full-scale training shows it helps. The Meta authors hit exactly this on an LLM…
Appeared in: 11-12
- 5. Proactive Dimension-Adaptive Denoising for Graph Collaborative Filtering via Representation Coherence · 📖 Deep dive
Implicit feedback mixes in accidental clicks and impulse purchases, and existing denoising methods almost all use training loss to find noise, but loss only grows after the model has been misled and is also disturbed…
Appeared in: 11-11
- 6. SeqFold: Timestamp-Addressable Compressed States for User Action Sequence Backbones · 📖 Deep dive
Downstream search and recommendation models all want to reuse the output of a frozen long-sequence user model, but the full multi-layer hidden state for one user at one point in time averages several million bytes…
Appeared in: 11-12
- 7. OptiCover: A Scalable and Multi-Surface Framework for Cold-Start Visual Optimization under Semantic Constraints · 📖 Deep dive
Airbnb wants to choose cover photos for hosts, and there are two difficulties. Over 90% of listing photos have never been a cover and so have no click data to learn from, and the most clickable photo may show a nearby…
Appeared in: 11-11
- 8. SAGE-PAIR: Source-Pretrained Routed Residual Adaptation for Sparse-Feedback Recommendation · 📖 Deep dive
When a recommender moves into a new serving window, the target window often has too little feedback to retrain the live model. This paper freezes the original model and adds only a residual: the item features are click…
Appeared in: 11-10
- 9. Query- and Candidate-Aware Long User Sequence Modeling for Sponsored Product Ranking in Search · 📖 Deep dive
When sponsored search ranking uses long-term behavior, the usual approach searches history for actions similar to the candidate product, which can miss history highly relevant to the current query but only loosely tied…
Appeared in: 11-11
- 10. INTENT: Out-of-Sample, Out-of-Time, and Cold-Start Evaluation on User Event Sequences · 📖 Deep dive
Once deployed, sequence models in banking and recommendation face a period after the training cutoff, users never seen in training, and newly appearing items, yet the usual evaluation just holds out each user's last…
Appeared in: 11-12