2026-10-19 ~ 10-25 · Personalization
- 1. Progressive Learning from Content to Relevance to Personalization for CTR Prediction · 📖 Deep dive
Taobao search CTR models mainly remember historical clicks through item IDs, so the same item with a different main image is almost indistinguishable to the model. FineSID turns each item image into 30 discrete codes…
Appeared in: 10-23
- 2. ECHO: Efficient Cascade with Homogeneous Operators for Distillation-Guided 100K-Sequence Modeling · 📖 Deep dive
ECHO makes 100K-event histories affordable by retaining reusable original records before candidate-aware retrieval. A full-history teacher supervises selection, while retrieval and aggregation share relevance…
Appeared in: 10-22
- 3. A Framework for Value-Aligned Modeling and Robust Experimentation · 📖 Deep dive
In the Google Discover ads team's model iteration, offline and online results did not line up. The correlation between the change in the conversion model's offline loss and online advertiser value was only 0.1…
Appeared in: 10-24
- 4. KuaFu: Compressing Long User Behavior into Understanding at Billion Scale · 📖 Deep dive
KuaFu avoids repeatedly encoding the same content across users and profiling tasks. It caches compact item representations, assembles them chronologically and uses staged training to preserve evidence…
Appeared in: 10-22
- 5. FLMNet: Self-Ensemble Distillation from Masks to Seeds for Single-Tower CTR Prediction · 📖 Deep dive
CTR self-distillation can be replaced with a computable disagreement penalty. FLMNet keeps gates on clean inputs and distills independently trained towers into one serving model…
Appeared in: 10-21
- 6. LSNR: Learned Sparse News Recommendations with Attentive User Modeling · 📖 Deep dive
Encode articles separately, then learn which ones matter. LSNR combines weighted sparse article vectors with a user-level semantic expansion, improving MIND NDCG@10 by about 11% over STAR.
Appeared in: 10-20
- 7. Predictable Scaling: Efficient Model Exploration via Scaling-Law Prediction · 📖 Deep dive
Predictable Scaling evaluates ranking-model changes with reduced-scale proxies and estimates their full-scale effects using historical learning curves. Training compute falls by 88% and 79% on two industrial ranking…
Appeared in: 10-20
- 8. DuoLM: Fast and Effective Personalized Recommendation with Language Models · 📖 Deep dive
DuoLM precomputes one vector per item and uses a personalized ranker with user history and feedback. Shared history computation and a specialized runtime reduce online work…
Appeared in: 10-21
- 9. EASformer: Environment-Action-State Alignment for Industrial Generative Ranking
EASformer addresses a mismatch between historical tokens containing observed actions and candidates with no observed action. It separates environment, action and state before matching them…
Appeared in: 10-22
- 10. SPAR: Stepwise Personalized Autoregressive Recommendation with Prefix-Balanced Semantic IDs
SPAR reduces early branch errors in semantic-ID recommendation by retrieving user evidence at every decoding step and balancing traffic across sibling branches. A Youku online experiment reports a 3.95% increase in…
Appeared in: 10-22