跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.09294 · v1

Orthogonal Norm Growth in Two-Tower Retrieval: A Popularity Mechanism for Unified Discovery

首次提交 2026-11-06 18:47
版本更新 2026-11-06 18:47
雷达观测 2026-11-09 11:19
arXiv ↗PDF ↗

摘要

Unified search and recommendation systems may update the same item vectors from both interaction streams. We study how repeated updates can increase an item vector's norm and affect retrieval under an unnormalized dot product. A cosine-only loss makes the output gradient perpendicular to the current item vector - tangent to its constant-norm sphere. Training changes encoder parameters, however, so the realized item-vector movement need not remain tangent. We derive sufficient conditions under which it does; every nonzero update then increases squared norm by its squared length. When the tasks share an item encoder, sequential updates from both can accumulate in the same norm, which raw-inner-product serving can expose for either task. This conditional result is not a theorem about joint SGD. Controlled ablations test these conditions. On MovieLens 32M and Amazon Books, accumulated tangent movement accompanies frequency-norm association and additional head-item exposure under raw-dot rather than normalized-item ranking. In a nonlinear tower combining a per-item ID vector with a shared content encoding, active item movement remains near-tangent. We separate its norm change into the ID-update contribution and other output changes, then use a matched training intervention that removes immediate norm growth caused by the ID-row update. The intervention reduces norm-related exposure but also recommendation utility, so it is diagnostic rather than a proposed mitigation. Our logs are recommendation-only: the formulation supplies a shared-representation audit, while prevalence in joint traffic remains open.

作者备注:6 pages, 1 figure. Accepted at USRW 2026 (non-archival), co-located with ACM RecSys 2026. Supplementary material included as an ancillary file. Workshop: https://usrw-workshop.github.io/2026/accepted-papers.html . Code: https://gitlab.com/wrapper228/usrw_pb

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-06 18:47 · 最新观测版本

Orthogonal Norm Growth in Two-Tower Retrieval: A Popularity Mechanism for Unified Discovery

Unified search and recommendation systems may update the same item vectors from both interaction streams. We study how repeated updates can increase an item vector's norm and affect retrieval under an unnormalized dot product. A cosine-only loss makes the output gradient perpendicular to the current item vector - tangent to its constant-norm sphere. Training changes encoder parameters, however, so the realized item-vector movement need not remain tangent. We derive sufficient conditions under which it does; every nonzero update then increases squared norm by its squared length. When the tasks share an item encoder, sequential updates from both can accumulate in the same norm, which raw-inner-product serving can expose for either task. This conditional result is not a theorem about joint SGD. Controlled ablations test these conditions. On MovieLens 32M and Amazon Books, accumulated tangent movement accompanies frequency-norm association and additional head-item exposure under raw-dot rather than normalized-item ranking. In a nonlinear tower combining a per-item ID vector with a shared content encoding, active item movement remains near-tangent. We separate its norm change into the ID-update contribution and other output changes, then use a matched training intervention that removes immediate norm growth caused by the ID-row update. The intervention reduces norm-related exposure but also recommendation utility, so it is diagnostic rather than a proposed mitigation. Our logs are recommendation-only: the formulation supplies a shared-representation audit, while prevalence in joint traffic remains open.

查看此版本解读 →查看此版本原文 ↗