Mend the Measurement Gap: Latent User Preference Modeling for Short-Form Video Recommendation

Shuo Chang; Yueqi Wang; Zihuan Diao; Ali Montazer; Jiangguo Zhang; Joyneel Misra; Dapeng Hong; Tomer Margolin

cs.IRSep 26, 2026

Mend the Measurement Gap: Latent User Preference Modeling for Short-Form Video Recommendation

Authors: Shuo Chang, Yueqi Wang, Zihuan Diao, Ali Montazer, Jiangguo Zhang, Joyneel Misra, Dapeng Hong, Tomer Margolin,

Organizations: Google Mountain View, CA, USA · Google New York City, NY, USA · YouTube Google Mountain View, CA, USA

Abstract

Recommender systems rely heavily on heterogeneous behavioral feedback to infer user preference. Although abundant, these signals are imperfect measurements: the same observed behavior can arise from different underlying states, such as genuine enjoyment, passive consumption, or inattention. The challenge is especially acute in short-form video, where watch-based signals are strongly affected by measurement confounders such as video duration - the same watch time can imply different levels of preference for videos of different lengths, while ratio-based metrics can systematically favor short videos. As a result, optimizing raw engagement can amplify measurement artifacts rather than improving user value. We propose a Factorized Latent Value Model (FLVM) for measuring user preference from heterogeneous behavioral feedback. The model treats observed behaviors as noisy measurements of a low-dimensional, factorized latent value state and uses structured output heads to model heterogeneous feedback signals. A restricted baseline path captures predictable variation from measurement-confounding features such as video duration, user propensity, and session context, while a routed latent path estimates preference-relevant value advantage. The resulting latent value score can be integrated into an existing recommender system as a ranking feature or ranking score. On YouTube Shorts, a major short-form video platform, this model improves offline metrics and lifts a primary viewer enjoyment metric by 2.67% in online A/B tests.

Figures & tables

Figure 1. Factorized Latent Value Model. The encoder maps full impression features

x

to a factorized latent value state

z=[z_{p},z_{a},z_{s}]

. For each feedback reconstruction task

k

, a task-specific hard sparse mask

m_{k}

restricts the latent input to decoder

h_{k}

, which produces a logit-scale advantage

\delta_{k}

. In parallel, a restricted baseline

b_{k}

uses only measurement-confounding features

x_{b}\subseteq x

to produce

\ell_{k}^{b}

. The final logit is

\ell_{k}=\mathrm{sg}(\ell_{k}^{b})+\delta_{k}

, and predictions are

\hat{y}_{k}=o_{k}(\ell_{k})

. The exact mask configuration is abstracted; the latent value score is scalarized from

z

for serving-time ranking. Architecture diagram of the Factorized Latent Value Model. Full impression features feed a latent value encoder, which produces three latent factors for consumption, engagement, and valence. A hard sparse routing module sends the latent state to several task-specific feedback reconstruction heads, while a restricted set of measurement-confounding features feeds matching baseline heads. Each reconstruction row combines a routed latent advantage with the corresponding stopped baseline logit to produce a final logit and prediction. A separate latent value score is derived from the factorized latent state for serving-time ranking.

Panel of Figure 2 — Figure 2. Empirical validation of the two main design choices. (a) The baseline path absorbs duration confounding: predicted item watch time increases with video duration, while the served consumption score $\mathrm{softplus}(z_{p})$ stays roughly flat across duration buckets. (b) The three latent factors separate semantically: liked and disliked impressions are most separated along the valence factor $z_{s}$ , less so along the active-engagement factor $z_{a}$ , and approximately invariant on the consumption factor $z_{p}$ . Two side-by-side charts. (a) Line plot showing predicted item watch time increasing approximately linearly across six video length buckets from 0-15 seconds to over 120 seconds, while softplus(z_p) remains roughly flat at around 3.5 across the same buckets. (b) Bar chart of three latent factors (consumption z_p, engagement z_a, valence z_s), each with two bars showing the relative shift of liked versus disliked impressions from the average user. The valence factor z_s shows the largest separation (positive for liked, negative for disliked); engagement z_a shows moderate separation in the same direction; consumption z_p is approximately invariant.

Signal	Metric	Production	FLVM
Item watch time	RMSE $\downarrow$	24.47	23.35
Completion	PR-AUC $\uparrow$	0.6898	0.6722
Like	PR-AUC $\uparrow$	0.2522	0.3190
Dislike	PR-AUC $\uparrow$	0.0038	0.0054

Table 1. Offline predictive performance versus the production multi-task baseline. PR-AUC (higher is better) for binary signals; RMSE (lower is better) for item watch time.

Signal	Baseline	Latent	Full
Completion	0.3065	0.6617	0.6722
Like	0.1860	0.0715	0.3190
Skip	0.1840	0.5016	0.5100
Dislike	0.0035	0.0035	0.0054

Table 2. Decomposition: PR-AUC of each path in FLVM. Baseline-only

=o_{k}(\ell_{k}^{b})

; Latent-only

=o_{k}(\delta_{k})

; Full

=o_{k}(\ell_{k})

. The two paths together exceed either alone.

Recommender Systems

Recommender systems predict and rank items for people using interaction histories, content, context, and feedback.

See all 516 papers

Action-Aware Generative Sequence Modeling for Short Video Recommendation
Apr 28, 2026Wenhao Li, Zihan Lin, Zhengxiao Guo +7Sequential Recommendation User Preference Modeling
Kuaishou Inc. · Beijing, China · Unaffiliated +2
Beyond Item IDs: Scaling Short-Form-Video Recommendation via Semantic-Native Long Sequence Modeling
May 4, 2026Ruixiao Sun, Diego Uribe Mora, Zhimeng Jiang +7Semantic ID Recommendation Sequential Recommendation
Google Mountain View, USA
A Computational Model of Message Sensation Value in Short Video Multimodal Features that Predicts Sensory and Behavioral Engagement
Apr 21, 2026Haoning Xue, Jingwen Zhang, Xiaohui Wang +2Social Media Analysis
Department of Communication, University of Utah · Department of Communication, University of California, Davis · Department of Media and Communication, City University of Hong Kong +1

Apr 28, 2026J/K move · Enter open · S save

Action-Aware Generative Sequence Modeling for Short Video Recommendation

Wenhao Li, Zihan Lin, Zhengxiao Guo, Jie Zhou+6

Kuaishou Inc. · Beijing, China · Unaffiliated+2

Sequential Recommendation User Preference Modeling Sequence Modeling Generative Recommendation

Open paper AskPDF arXiv

With the rapid development of the Internet, users have increasingly higher expectations for the recommendation accuracy of online content consumption platforms. However, short videos often contain diverse segments, and users may not hold the same attitude toward all of them. Traditional binary-classification recommendation models, which treat a video as a single holistic entity, face limitations in accurately capturing such nuanced preferences. Considering that user consumption is a temporal process, this paper demonstrates that the timing of user actions can represent diverse intentions through statistical analysis and examination of action patterns. Based on this insight, we propose a novel modeling paradigm: Action-Aware Generative Sequence Network (A2Gen), which refines user actions along the temporal dimension and chains them into sequences for unified processing and prediction. First, we introduce the Context-aware Attention Module (CAM) to model action sequences enriched with item-specific contextual features. Building upon this, we develop the Hierarchical Sequence Encoder (HSE) to learn temporal action patterns from users' historical actions. Finally, through leveraging CAM, we design a module for action sequence generation: the Action-seq Autoregressive Generator (AAG). Extensive offline experiments on the Kuaishou's dataset and the Tmall public dataset demonstrate the superiority of our proposed model. Furthermore, through large-scale online A/B testing deployed on Kuaishou's platform, our model achieves significant improvements over baseline methods in multi-task prediction by leveraging sequential information. Specifically, it yields increases of 0.34% in user watch time, 8.1% in interaction rate, and 0.162% in overall user retention (LifeTime-7), leading to successful deployment across all traffic, serving over 400 million users every day.

Abstract

Recommender Systems

Explore similar work

Action-Aware Generative Sequence Modeling for Short Video Recommendation

Beyond Item IDs: Scaling Short-Form-Video Recommendation via Semantic-Native Long Sequence Modeling

A Computational Model of Message Sensation Value in Short Video Multimodal Features that Predicts Sensory and Behavioral Engagement