Metraipolis › archive › dc_cfe107da4a79
Knowledge-Geometry Decoupling: Refreshable Pretrained Transfer for Streaming Recommendation — arXiv:2608.02738
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.02738
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T12:57:49.374Z
- SHA-256 of the archived text
56ef18caef18f0dd168a641fd2a63331b222fd03b995a426159ead882c29512a- Stable URL of this page
https://metraipolis.com/documento/dc_cfe107da4a79
How to cite
Knowledge-Geometry Decoupling: Refreshable Pretrained Transfer for Streaming Recommendation — arXiv:2608.02738 — arXiv (en), https://arxiv.org/abs/2608.02738 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cfe107da4a79 (Metraipolis archive dc_cfe107da4a79, sha256 56ef18caef18f0dd168a641fd2a63331b222fd03b995a426159ead882c29512a)
BibTeX
@misc{dccfe107da4a79,
title = {Knowledge-Geometry Decoupling: Refreshable Pretrained Transfer for Streaming Recommendation — arXiv:2608.02738},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.02738},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cfe107da4a79; sha256 56ef18caef18f0dd168a641fd2a63331b222fd03b995a426159ead882c29512a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.02738. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Industrial recommenders increasingly adopt the pretrain-then-transfer paradigm, yet behavioral distribution drift raises two questions: what to learn from behavior sequences, and how to transfer the learned knowledge while the pretrained model is continually refreshed. To resolve them, we propose Knowledge-Geometry Decoupling (KGD). For what to learn, conventional next-token prediction treats adjacency as dependency and may encode spurious transitions across unrelated sessions. We introduce Behavioral Multi-Token Prediction (BMTP) to retain only collaboratively or semantically related future items as supervision, yielding cleaner and more transferable behavioral knowledge. For how to transfer, pretrained knowledge and task-specific geometry impose conflicting optimization demands on shared parameters. To handle it, KGD assigns them to separate parameter sets: a refreshable encoder owns behavioral knowledge, while a task learner reads contextualized encoder states through read-only cross-attention and writes task-specific geometry through Anchored Calibration Residual (ACR) orthogonal to the pretrained embedding. The decoupled ownership enables continual knowledge refresh without task-gradient interference or invalidating downstream adaptation. KGD improves over strong pretrain-transfer baselines by 4-12% on eight public benchmarks and sustains its advantage over a 90-day production stream where baselines show no gains. KGD has been fully deployed in Shopee. In a live A/B test on Shopee Homepage Search, it increases GMV per user by 1.75% and advertising revenue by 1.53%, demonstrating its high practical value. We provide the core implementation of KGD at this https URL . Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-08-05 in cs.IR). Categorie arXiv: Information Retrieval (cs.IR). Pagina del lavoro: https://arxiv.org/abs/2608.02738 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-31T12:45:13.291Z dalla pagina https://arxiv.org/catchup/cs.IR/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).