https://arxiv.org/api/mQEW2tfJtG8iabLBrDhhn77n0c8 2026-09-26T16:17:37Z 40 106 0 http://arxiv.org/abs/2607.28019v1 2026-07-30T11:07:54Z

User foundation models have demonstrated strong results in e-commerce and social recommendation, but most industrial deployments assume environments where user identity is stable and persistent. Open-web real-time bidding (RTB) operates on a structurally different data distribution: user identity is fragmented and non-persistent across browsing sessions, and the availability of browsing history depends on user privacy choices. Consequently, a significant portion of traffic carries no historical data, and available records often consist of relatively short, disjointed sessions. As a result, historical signals in this domain are typically represented as aggregated counters and recency buckets, leaving the sequential structure unexploited. To address this limitation, we present a user foundation model that applies self-supervised learning on user browsing histories and show that the learned representation improves multiple downstream production tasks, demonstrating the viability of this approach on the open web. We pre-train a Transformer encoder with masked language modeling and a sequence-level contrastive objective, then fine-tune it on the click prediction task. We optimize the encoder's pre-training pipeline with an LLM-in-the-loop search over a curated catalog of reviewable, code-level edits (lifters), instantiating the LLM-as-optimizer paradigm in an industrial setting. The same encoder representation yields +1.197% RIG on the production bid win-rate model and +1.354% RIG on the production CTR ranker; a 7-day live A/B test confirms +2.13% CTR, -1.13% eCPC (80% CI excluding zero on both metrics).

2026-07-30T11:07:54Z RecSys'26 Solal Vernier Ivan Can Arisoy Merwan Barlier Blaž Škrlj http://arxiv.org/abs/2607.27577v1 2026-07-30T01:41:42Z

Heterogeneous recommendation feeds present complex challenges that extend beyond those found in highly homogeneous environments (e.g., music-only or video-only closed-ecosystem platforms). In Google Discover, a unified feed integrates diverse content sourced from the decentralized open web, including web articles, long-form and short-form videos, user-generated content (UGC), and beyond. Different content types exhibit distinct feature densities and user interaction patterns. Building a unified ranking model that sustains high performance across such heterogeneity, while avoiding negative transfer or majority bias, remains a significant industrial challenge. This paper presents an end-to-end case study on the industrial-scale multi-task ranking of heterogeneous feeds, grounded in real-world deployment. We introduce HA-MoE, a heterogeneity-adaptive multi-gated mixture-of-experts architecture that incorporates explicit heterogeneity context into both gating networks and expert representations. This approach enables effective specialization without significantly increasing operational overhead. To support reliable deployment, we introduce LENS, a lightweight observability framework that provides interpretable diagnostics of expert specialization and tracks this functional heterogeneity across continuous retraining. We evaluate our method using Dual-Level AUC (DL-AUC), a heterogeneity-aware evaluation metric that combines global ranking performance with cross-segment ranking correctness. Offline evaluations on a large-scale industrial dataset demonstrate consistent improvements over baseline models. Furthermore, online A/B testing confirms gains in feed activity and exploration metrics. Together, offline and online results validate the effectiveness of our approach for managing heterogeneity in industrial-scale recommender systems.

2026-07-30T01:41:42Z Accepted to ACM RecSys Industry Track 2026 Proceedings of ACM RecSys 2026 Di Bai Jintao Liu Zhenwei Tang Peifan Wu Nada Al-Thawr Luoshu Wang 10.1145/3773078.3831848 http://arxiv.org/abs/2608.07517v1 2026-07-02T18:51:16Z

Can a multimodal LLM predict which version of a web page will win a real A/B test from screenshots alone? We report the most complete answer we are aware of, from six weeks of pre-registered experiments on real conversion tests: mostly no -- and the exceptions are identifiable in advance. On 330 real A/B tests a Gemini 3 Flash judge reaches Cohen's kappa = 0.14, but on the trustworthy (statistically significant) half of the labels the evidence is inconclusive (kappa = 0.11, CI includes zero). We show that 44% of the "ground-truth" labels in a leading CRO agency's catalog come from non-significant tests, and that the judge agrees more with the unreliable labels than the reliable ones -- a shared prior between labeler and model, not prediction. Every standard improvement lever (a 2.8x more expensive frontier model, prompt redesign, stimulus fidelity, change-type priors) fails its pre-registered gate. The judge's confident calls are different: a vote-margin gate isolates a subset (49% coverage) reaching kappa = 0.31 on significant labels. We measure the mechanism directly -- judges differing in model or prompt agree with each other at kappa = 0.74-0.88 while agreeing with real outcomes at only ~0.2, so a 16-vote panel carries about 2 effective independent votes -- and we reproduce it in humans: 15 CRO experts agree with each other (inter-rater kappa = 0.53) but score at chance against real outcomes (kappa ~ 0). Consensus, human or model, is reproducible, persuasive, and not evidence. We release our pre-registrations, locked gates, negative results, statistical harness, human responses, and a claims ledger in which every number carries an evidence tier.

2026-07-02T18:51:16Z 15 pages. Pre-registered experimental program with a public, tiered claims ledger; includes powered negative results, a label-validity audit, a cross-judge shared-prior measurement (n_eff ~ 2 of 16 votes), and a first-party 15-expert human baseline. Pre-registrations, statistical harness, human responses, and the full experiment ledger are released Tyler Dooskin Squoosh Technical Staff http://arxiv.org/abs/2606.03866v1 2026-06-02T16:39:06Z

Scaling recommender systems via large language models (LLMs) has become a prominent trend in the industry. However, aligning the LLM's semantic space with the recommender's ID space via post-training (e.g., SFT and RL) remains challenging. Existing LLM4Rec paradigms are bottlenecked by two main issues: (1) the difficulty of measuring and improving chain-of-thought (CoT) quality in open-domain recommendation during SFT, and (2) the neglect of the trade-off between LLM semantic rewards and recommendation preference rewards during RL alignment. Inspired by these challenges, we present Taiji, a novel LLM-as-Enhancer framework designed for industrial recommender systems. To overcome the SFT bottleneck, we utilize reverse-engineered reasoning and open-ended rejection sampling to generate high-quality, domain-specific CoT data. To resolve the RL alignment issue, we propose Pareto Optimal Policy Optimization (POPO), which adaptively adjusts cross-domain reward weights. Theoretically, it achieves an optimal trade-off between the semantic world knowledge of LLMs and the collaborative ID features representing online user preferences. Extensive offline evaluations and online A/B tests validate the effectiveness of Taiji. Deployed on Kuaishou's advertising platform since May 2026, Taiji currently serves over 400 million users daily, yielding significant commercial revenue and demonstrating its robust scalability in web-scale environments.

2026-06-02T16:39:06Z 8 pages, 2 figures Yuecheng Li Zeyu Song Jing Yao Chi Lu Peng Jiang Kun Gai http://arxiv.org/abs/2605.13052v1 2026-05-13T06:20:28Z

In commercial web search, aligning content freshness with user intent remains challenging due to the highly varied lifespans of information. Traditional industrial approaches rely on static time-window filtering, resulting in "one-size-fits-all" rankings where content may be chronologically recent but semantically expired. To address the limitation, we present a novel Large Language Models (LLMs)-based Query-Aware Dynamic Content Expiration Prediction Framework deployed in Baidu search, reformulating timeliness as a dynamic validity inference task. Our framework extracts fine-grained temporal contexts from documents and leverages LLMs to deduce a query-specific "validity horizon"-a semantic boundary defining when information becomes obsolete based on user intent. Integrated with robust hallucination mitigation strategies to ensure reliability, our approach has been evaluated through offline and online A/B testing on live production traffic. Results demonstrate significant improvements in search freshness and user experience metrics, validating the effectiveness of LLM-driven reasoning for solving semantic expiration at an industrial scale.

2026-05-13T06:20:28Z Accepted at SIGIR 2026. Final version: https://doi.org/10.1145/3805712.3808457 Tingyu Chen Wenkai Zhang Li Gao Lixin Su Ge Chen Dawei Yin Daiting Shi http://arxiv.org/abs/2604.13468v2 2026-04-19T21:19:46Z

Generative information retrieval (GenIR) formulates the retrieval process as a text-to-text generation task, leveraging the vast knowledge of large language models. However, existing works primarily optimize for relevance while often overlooking document trustworthiness. This is critical in high-stakes domains like healthcare and finance, where relying solely on semantic relevance risks retrieving unreliable information. To address this, we propose an Authority-aware Generative Retriever (AuthGR), the first framework that incorporates authority into GenIR. AuthGR consists of three key components: (i) Multimodal Authority Scoring, which employs a vision-language model to quantify authority from textual and visual cues; (ii) a Three-stage Training Pipeline to progressively instill authority awareness into the retriever; and (iii) a Hybrid Ensemble Pipeline for robust deployment. Offline evaluations demonstrate that AuthGR successfully enhances both authority and accuracy, with our 3B model matching a 14B baseline. Crucially, large-scale online A/B tests and human evaluations conducted on the commercial web search platform confirm significant improvements in real-world user engagement and reliability.

2026-04-15T04:44:45Z ACL 2026 (Industry Track) Sunkyung Lee Jihye Back Donghyeon Jeon Soonhwan Kwon Moonkwon Kim Inho Kang Jongwuk Lee http://arxiv.org/abs/2603.08370v1 2026-03-09T13:32:39Z

Continuous and efficient experimentation is key to the practical success of user-facing applications on the web, both through online A/B-tests and off-policy evaluation. Despite their shared objective -- estimating the incremental value of a treatment -- these domains often operate in isolation, utilising distinct terminologies and statistical toolkits. This paper bridges that divide by establishing a formal equivalence between their canonical variance reduction methods. We prove that the standard online Difference-in-Means estimator is mathematically identical to an off-policy Inverse Propensity Scoring estimator equipped with an optimal (variance-minimising) additive control variate. Extending this unification, we demonstrate that widespread regression adjustment methods (such as CUPED, CUPAC, and ML-RATE) are structurally equivalent to Doubly Robust estimation. This unified view extends our understanding of commonly used approaches, and can guide practitioners and researchers working on either class of problems.

2026-03-09T13:32:39Z Olivier Jeunen http://arxiv.org/abs/2603.00980v1 2026-03-01T08:15:34Z

Generative Recommenders (GRs), exemplified by the Hierarchical Sequential Transduction Unit (HSTU), have emerged as a powerful paradigm for modeling long user interaction sequences. However, we observe that their "flat-sequence" assumption overlooks the rich, intrinsic structure of user behavior. This leads to two key limitations: a failure to capture the temporal hierarchy of session-based engagement, and computational inefficiency, as dense attention introduces significant noise that obscures true preference signals within semantically sparse histories, which deteriorates the quality of the learned representations. To this end, we propose a novel framework named HPGR (Hierarchical and Preference-aware Generative Recommender), built upon a two-stage paradigm that injects these crucial structural priors into the model to handle the drawback. Specifically, HPGR comprises two synergistic stages. First, a structure-aware pre-training stage employs a session-based Masked Item Modeling (MIM) objective to learn a hierarchically-informed and semantically rich item representation space. Second, a preference-aware fine-tuning stage leverages these powerful representations to implement a Preference-Guided Sparse Attention mechanism, which dynamically constrains computation to only the most relevant historical items, enhancing both efficiency and signal-to-noise ratio. Empirical experiments on a large-scale proprietary industrial dataset from APPGallery and an online A/B test verify that HPGR achieves state-of-the-art performance over multiple strong baselines, including HSTU and MTGR.

2026-03-01T08:15:34Z Accepted to the ACM Web Conference 2026 (WWW '26). 9 pages, 9 figures. Zerui Chen and Heng Chang contributed equally to this work Proceedings of the ACM Web Conference 2026 (WWW '26), April 13-17, 2026, Dubai, United Arab Emirates Zerui Chen Heng Chang Tianying Liu Chuantian Zhou Yi Cao Jiandong Ding Ming Liu Bing Qin 10.1145/3774904.3792790 http://arxiv.org/abs/2602.12485v1 2026-02-12T23:59:32Z

The success of businesses depends on their ability to convert consumers into loyal customers. A customer's value proposition is a primary determinant in this process, requiring a balance between affordability and long-term brand equity. Broad marketing campaigns can erode perceived brand value and reduce return on investment, while existing economic algorithms often misidentify highly engaged customers as ideal targets, leading to inefficient engagement and conversion outcomes. This work introduces a two-stage multi-model architecture employing Self-Paced Loss to improve customer categorization. The first stage uses a multi-class neural network to distinguish customers influenced by campaigns, organically engaged customers, and low-engagement customers. The second stage applies a binary label correction model to identify true campaign-driven intent using a missing-label framework, refining customer segmentation during training. By separating prompted engagement from organic behavior, the system enables more precise campaign targeting, reduces exposure costs, and improves conversion efficiency. A/B testing demonstrates over 100 basis points improvement in key success metrics, highlighting the effectiveness of intent-aware segmentation for value-driven marketing strategies.

2026-02-12T23:59:32Z Companion Proceedings of the ACM Web Conference 2025 (WWW Companion 25), ACM, 2025 Keerthi Gopalakrishnan Tianning Dong Chia-Yen Ho Yokila Arora Topojoy Biswas Jason Cho Sushant Kumar Kannan Achan 10.1145/3701716.3715243 http://arxiv.org/abs/2602.07987v1 2026-02-08T14:27:34Z

Modern video recommendation systems aim to optimize user engagement and platform objectives, yet often face structural exposure imbalances caused by behavioral biases. In this work, we focus on the post-ranking stage and present LAFB (Learning to Alleviate Familiarity Bias), a lightweight and model-agnostic framework designed to mitigate familiarity bias in recommendation outputs. LAFB models user-content familiarity using discrete and continuous interaction features, and estimates personalized debiasing factors to adjust user rating prediction scores, thereby reducing the dominance of familiar content in the final ranking. We conduct large-scale offline evaluations and online A/B testing in a real-world recommendation system, under a unified serving stack that also compares LAFB with deployable popularity-oriented remedies. Results show that LAFB increases novel watch-time share and improves exposure for emerging creators and overall content diversity, while maintaining stable overall watch time and short-term satisfaction. LAFB has already been launched in the post-ranking stage of YouTube's recommendation system, demonstrating its effectiveness in real-world applications.

2026-02-08T14:27:34Z Accepted to the Companion Proceedings of the ACM Web Conference 2026 (WWW '26), April 13-17, 2026, Dubai, UAE Zheng Ren Yi Wu Jianan Lu Acar Ary Yiqu Liu Li Wei Lukasz Heldt http://arxiv.org/abs/2601.17836v1 2026-01-25T13:39:26Z

In recent years, the success of large language models (LLMs) has driven the exploration of scaling laws in recommender systems. However, models that demonstrate scaling laws are actually challenging to deploy in industrial settings for modeling long sequences of user behaviors, due to the high computational complexity of the standard self-attention mechanism. Despite various sparse self-attention mechanisms proposed in other fields, they are not fully suited for recommendation scenarios. This is because user behaviors exhibit personalization and temporal characteristics: different users have distinct behavior patterns, and these patterns change over time, with data from these users differing significantly from data in other fields in terms of distribution. To address these challenges, we propose SparseCTR, an efficient and effective model specifically designed for long-term behaviors of users. To be precise, we first segment behavior sequences into chunks in a personalized manner to avoid separating continuous behaviors and enable parallel processing of sequences. Based on these chunks, we propose a three-branch sparse self-attention mechanism to jointly identify users' global interests, interest transitions, and short-term interests. Furthermore, we design a composite relative temporal encoding via learnable, head-specific bias coefficients, better capturing sequential and periodic relationships among user behaviors. Extensive experimental results show that SparseCTR not only improves efficiency but also outperforms state-of-the-art methods. More importantly, it exhibits an obvious scaling law phenomenon, maintaining performance improvements across three orders of magnitude in FLOPs. In online A/B testing, SparseCTR increased CTR by 1.72\% and CPM by 1.41\%. Our source code is available at https://github.com/laiweijiang/SparseCTR.

2026-01-25T13:39:26Z WWW 2026: THE ACM WEB CONFERENCE Weijiang Lai Beihong Jin Di Zhang Siru Chen Jiongyan Zhang Yuhang Gou Jian Dong Xingxing Wang 10.1145/3774904.3792808 http://arxiv.org/abs/2510.26104v3 2026-02-02T06:48:36Z

In recommendation systems, scaling up feature-interaction modules (e.g., Wukong, RankMixer) or user-behavior sequence modules (e.g., LONGER) has achieved notable success. However, these efforts typically proceed on separate tracks, which not only hinders bidirectional information exchange but also prevents unified optimization and scaling. In this paper, we propose OneTrans, a unified Transformer backbone that simultaneously performs user-behavior sequence modeling and feature interaction. OneTrans employs a unified tokenizer to convert both sequential and non-sequential attributes into a single token sequence. The stacked OneTrans blocks share parameters across similar sequential tokens while assigning token-specific parameters to non-sequential tokens. Through causal attention and cross-request KV caching, OneTrans enables precomputation and caching of intermediate representations, significantly reducing computational costs during both training and inference. Experimental results on industrial-scale datasets demonstrate that OneTrans scales efficiently with increasing parameters, consistently outperforms strong baselines, and yields a 5.68% lift in per-user GMV in online A/B tests.

2025-10-30T03:30:12Z Accepted at The Web Conference 2026 (WWW 2026). Camera-ready version forthcoming Zhaoqi Zhang Haolei Pei Jun Guo Tianyu Wang Yufei Feng Hui Sun Shaowei Liu Aixin Sun http://arxiv.org/abs/2510.15299v3 2026-04-01T06:51:34Z

Industrial-scale recommender systems rely on a cascade pipeline in which the retrieval stage must return a high-recall candidate set from billions of items under tight latency. Existing solutions either (i) suffer from limited expressiveness in capturing fine-grained user-item interactions, as seen in decoupled dual-tower architectures that rely on separate encoders, or generative models that lack precise target-aware matching capabilities, or (ii) build structured indices (tree, graph, quantization) whose item-centric topologies struggle to incorporate dynamic user preferences and incur prohibitive construction and maintenance costs. We present GRank, a novel structured-index-free retrieval paradigm that seamlessly unifies target-aware learning with user-centric retrieval. Our key innovations include: (1) A target-aware Generator trained to perform personalized candidate generation via GPU-accelerated MIPS, eliminating semantic drift and maintenance costs of structured indexing; (2) A lightweight but powerful Ranker that performs fine-grained, candidate-specific inference on small subsets; (3) An end-to-end multi-task learning framework that ensures semantic consistency between generation and ranking objectives. Extensive experiments on two public benchmarks and a billion-item production corpus demonstrate that GRank improves Recall@500 by over 30% and 1.7$\times$ the P99 QPS of state-of-the-art tree- and graph-based retrievers. GRank has been fully deployed in production in our recommendation platform since Q2 2025, serving 400 million active users with 99.95% service availability. Online A/B tests confirm significant improvements in core engagement metrics, with Total App Usage Time increasing by 0.160% in the main app and 0.165% in the Lite version.

2025-10-17T04:15:09Z Proceedings of the ACM Web Conference 2026 (WWW '26), April 13--17, 2026, Dubai, United Arab Emirates Yijia Sun Shanshan Huang Zhiyuan Guan Qiang Luo Ruiming Tang Kun Gai Guorui Zhou 10.1145/3774904.3792810 http://arxiv.org/abs/2510.09801v3 2026-06-09T14:05:10Z

While benchmarks measure the accuracy of LLM-powered agents, they mostly assume full automation, failing to represent the collaborative nature of real-world use cases. In this paper, we make two major steps towards the rigorous assessment of human-agent interactions. First, we propose PULSE, a framework for more efficient human-centric evaluation of agent designs, which comprises collecting user feedback, training an ML model to predict user satisfaction, and computing results by combining human satisfaction ratings with model-generated pseudo-labels. Second, we deploy PULSE in software engineering -- one of the highest-impact, real-world domains for LLM agents -- via a large-scale web platform built around the open-source agent OpenHands. Across 15k users, we evaluate how three agent design decisions impact developer satisfaction rates. We also show how PULSE can lead to more robust conclusions about agent design, reducing confidence intervals by 40\% compared to a standard A/B test. Finally, we find substantial discrepancies between in-the-wild results with benchmark performance (e.g., the anti-correlation between claude-sonnet-4 and gpt-5), underscoring the limitations of benchmark-driven evaluation. Our framework PULSE provides guidance for future evaluations, and our findings identify opportunities for better software agent designs.

2025-10-10T19:04:28Z ICML 2026 Valerie Chen Rohit Malhotra Xingyao Wang Juan Michelini Xuhui Zhou Aditya Bharat Soni Hoang H. Tran Calvin Smith Ameet Talwalkar Graham Neubig http://arxiv.org/abs/2510.09136v2 2026-05-22T09:10:51Z

Personalized news recommendations have become a standard feature of large news aggregation services, optimizing user engagement through automated content selection. In contrast, legacy news media often approach personalization cautiously, striving to balance technological innovation with core editorial values. As a result, online platforms of traditional news outlets typically combine editorially curated content with algorithmically selected articles - a strategy we term controlled personalization. In this industry article, we evaluate the effectiveness of controlled personalization through an A/B test conducted on the website of a major Norwegian legacy news organization. Our findings indicate that even a modest level of personalization yields substantial benefits. Specifically, we observe that users exposed to personalized content demonstrate higher click-through-rates and reduced navigation effort, suggesting improved discovery of relevant content. Moreover, our analysis reveals that controlled personalization contributes to greater content diversity and catalog coverage and in addition reduces popularity bias. Overall, our results suggest that controlled personalization can successfully align user needs with editorial goals, offering a viable path for legacy media to adopt personalization technologies while upholding journalistic values.

2025-10-10T08:37:13Z Marlene Holzleitner Stephan Leitner Hanna Lind Jorgensen Christoph Schmitz Jacob Welander Dietmar Jannach 10.1145/3814614 http://arxiv.org/abs/2509.15576v1 2025-09-19T04:22:31Z

Online controlled experiments, also known as A/B testing, are the digital equivalent of randomized controlled trials for estimating the impact of marketing campaigns on website visitors. Stratified sampling is a traditional technique for variance reduction to improve the sensitivity (or statistical power) of controlled experiments; this technique first divides the population into strata (homogeneous subgroups) based on stratification variables and then draws samples from each stratum to avoid sampling bias. To enhance the estimation accuracy of stratified sampling, we focus on the problem of selecting a subset of stratification variables that are effective in variance reduction. We design an efficient algorithm that selects stratification variables one by one by simulating a series of stratified sampling processes. We also estimate the computational complexity of our subset selection algorithm. Computational experiments using synthetic and real-world datasets demonstrate that our method can outperform other variance reduction techniques especially when multiple variables have a certain correlation with the outcome variable. Our subset selection method for stratified sampling can improve the sensitivity of online controlled experiments, thus enabling more reliable marketing decisions.

2025-09-19T04:22:31Z 14 pages, 15 figures, The 22nd Pacific Rim International Conference on Artificial Intelligence 2025 (PRICAI 2025) Haru Momozu Yuki Uehara Naoki Nishimura Koya Ohashi Deddy Jobson Yilin Li Phuong Dinh Noriyoshi Sukegawa Yuichi Takano http://arxiv.org/abs/2508.15281v1 2025-08-21T06:15:49Z

Recommender systems traditionally represent items using unique identifiers (ItemIDs), but this approach struggles with large, dynamic item corpora and sparse long-tail data, limiting scalability and generalization. Semantic IDs, derived from multimodal content such as text and images, offer a promising alternative by mapping items into a shared semantic space, enabling knowledge transfer and improving recommendations for new or rare items. However, existing methods face two key challenges: (1) balancing cross-modal synergy with modality-specific uniqueness, and (2) bridging the semantic-behavioral gap, where semantic representations may misalign with actual user preferences. To address these challenges, we propose Multimodal Mixture-of-Quantization (MMQ), a two-stage framework that trains a novel multimodal tokenizer. First, a shared-specific tokenizer leverages a multi-expert architecture with modality-specific and modality-shared experts, using orthogonal regularization to capture comprehensive multimodal information. Second, behavior-aware fine-tuning dynamically adapts semantic IDs to downstream recommendation objectives while preserving modality information through a multimodal reconstruction loss. Extensive offline experiments and online A/B tests demonstrate that MMQ effectively unifies multimodal synergy, specificity, and behavioral adaptation, providing a scalable and versatile solution for both generative retrieval and discriminative ranking tasks.

2025-08-21T06:15:49Z Proceedings of the Nineteenth ACM International Conference on Web Search and Data Mining, 2026 Yi Xu Moyu Zhang Chenxuan Li Zhihao Liao Haibo Xing Hao Deng Jinxin Hu Yu Zhang Xiaoyi Zeng Jing Zhang 10.1145/3773966.3777923 http://arxiv.org/abs/2509.09690v1 2025-08-19T21:35:43Z

Query understanding is essential in modern relevance systems, where user queries are often short, ambiguous, and highly context-dependent. Traditional approaches often rely on multiple task-specific Named Entity Recognition models to extract structured facets as seen in job search applications. However, this fragmented architecture is brittle, expensive to maintain, and slow to adapt to evolving taxonomies and language patterns. In this paper, we introduce a unified query understanding framework powered by a Large Language Model (LLM), designed to address these limitations. Our approach jointly models the user query and contextual signals such as profile attributes to generate structured interpretations that drive more accurate and personalized recommendations. The framework improves relevance quality in online A/B testing while significantly reducing system complexity and operational overhead. The results demonstrate that our solution provides a scalable and adaptable foundation for query understanding in dynamic web applications.

2025-08-19T21:35:43Z CIKM2025 Ping Liu Jianqiang Shen Qianqi Shen Chunnan Yao Kevin Kao Dan Xu Rajat Arora Baofen Zheng Caleb Johnson Liangjie Hong Jingwei Wu Wenjing Zhang 10.1145/3746252.3760913 http://arxiv.org/abs/2508.08077v1 2025-08-11T15:21:59Z

Bayesian A/B testing investigates metric changes using the joint posterior distribution of two (or more) experimentally-derived datasets. The construction of said joint posterior is often a time-consuming process requiring specialized knowledge and domain expertise. In businesses that perform tens to hundreds of A/B tests per month it is important to have a robust analysis pipeline that can handle the variety of experiments performed on a modern web platform; requiring a domain expert to select appropriate prior and likelihood distributions for each experiment simply does not scale. In this work, we highlight a solution to this problem using a generalized approximation of the true joint posterior using a Dirichlet-Categorical model. While a manually-constructed, expert-tuned model for every dataset is preferable, the Dirichlet-Categorical approximation performs sufficiently well in both simulations and real-world scenarios to be internally used as the standard analysis method.

2025-08-11T15:21:59Z 30 pages, 20 figures, companion paper to accepted talk at the Royal Statistical Society International Conference 2025 Dustin Hayden Thomas Armitage http://arxiv.org/abs/2508.00201v1 2025-07-31T23:01:14Z

Existing web-scale recommendation systems commonly use supervised learning methods that prioritize immediate user feedback. Although reinforcement learning (RL) offers a solution to optimize longer-term goals, such as in-session engagement, applying it at web scale is challenging due to the extremely large action space and engineering complexity. In this paper, we introduce RecoMind, a simulator-based RL framework designed for the effective optimization of session-based goals at web-scale. RecoMind leverages existing recommendation models to establish a simulation environment and to bootstrap the RL policy to optimize immediate user interactions from the outset. This method integrates well with existing industry pipelines, simplifying the training and deployment of RL policies. Additionally, RecoMind introduces a custom exploration strategy to efficiently explore web-scale action spaces with hundreds of millions of items. We evaluated RecoMind through extensive offline simulations and online A/B testing on a video streaming platform. Both methods showed that the RL policy trained using RecoMind significantly outperforms traditional supervised learning recommendation approaches in in-session user satisfaction. In online A/B tests, the RL policy increased videos watched for more than 10 seconds by 15.81\% and improved session depth by 4.71\% for sessions with at least 10 interactions. As a result, RecoMind presents a systematic and scalable approach for embedding RL into web-scale recommendation systems, showing great promise for optimizing session-based user satisfaction.

2025-07-31T23:01:14Z Mehdi Ben Ayed Fei Feng Jay Adams Vishwakarma Singh Kritarth Anand Jiajing Xu http://arxiv.org/abs/2507.15113v1 2025-07-20T20:25:20Z

User journeys in e-commerce routinely violate the one-to-one assumption that a clicked item on an advertising platform is the same item later purchased on the merchant's website/app. For a significant number of converting sessions on our platform, users click product A but buy product B -- the Click A, Buy B (CABB) phenomenon. Training recommendation models on raw click-conversion pairs therefore rewards items that merely correlate with purchases, leading to biased learning and sub-optimal conversion rates. We reframe conversion prediction as a multi-task problem with separate heads for Click A Buy A (CABA) and Click A Buy B (CABB). To isolate informative CABB conversions from unrelated CABB conversions, we introduce a taxonomy-aware collaborative filtering weighting scheme where each product is first mapped to a leaf node in a product taxonomy, and a category-to-category similarity matrix is learned from large-scale co-engagement logs. This weighting amplifies pairs that reflect genuine substitutable or complementary relations while down-weighting coincidental cross-category purchases. Offline evaluation on e-commerce sessions reduces normalized entropy by 13.9% versus a last-click attribution baseline. An online A/B test on live traffic shows +0.25% gains in the primary business metric.

2025-07-20T20:25:20Z Xiangyu Zeng Amit Jaspal Bin Liu Goutham Panneeru Kevin Huang Nicolas Bievre Mohit Jaggi Prathap Maniraju Ankur Jain http://arxiv.org/abs/2507.12118v1 2025-07-16T10:37:29Z

In recent years, attention has increasingly focused on enhancing user satisfaction with user interfaces, spanning both mobile applications and websites. One fundamental aspect of human-machine interaction is the concept of web usability. In order to assess web usability, the A/B testing technique enables the comparison of data between two designs. Expanding the scope of tests to include the designs being evaluated, in conjunction with the involvement of both real and fictional users, presents a challenge for which few online tools offer support. We propose a methodology for web usability evaluation based on user-centered approaches such as design thinking and linguistic decision-making, named Linguistic Decision-Making for Web Usability Evaluation. This engages people in role-playing scenarios and conducts a number of usability tests, including the widely recognized System Usability Scale. We incorporate the methodology into a decision support system based on A/B testing. We use real users in a case study to assess three Moodle platforms at the University of Guadalajara, Mexico.

2025-07-16T10:37:29Z Noe Zermeño Cristina Zuheros Lucas Daniel Del Rosso Calache Francisco Herrera Rosana Montes http://arxiv.org/abs/2506.17863v1 2025-06-22T00:28:35Z

Offsite marketing is essential in e-commerce, enabling businesses to reach customers through external platforms and drive traffic to retail websites. However, most current offsite marketing content is overly generic, template-based, and poorly aligned with landing pages, limiting its effectiveness. To address these limitations, we propose MarketingFM, a retrieval-augmented system that integrates multiple data sources to generate keyword-specific ad copy with minimal human intervention. We validate MarketingFM via offline human and automated evaluations and large-scale online A/B tests. In one experiment, keyword-focused ad copy outperformed templates, achieving up to 9% higher CTR, 12% more impressions, and 0.38% lower CPC, demonstrating gains in ad ranking and cost efficiency. Despite these gains, human review of generated ads remains costly. To address this, we propose AutoEval-Main, an automated evaluation system that combines rule-based metrics with LLM-as-a-Judge techniques to ensure alignment with marketing principles. In experiments with large-scale human annotations, AutoEval-Main achieved 89.57% agreement with human reviewers. Building on this, we propose AutoEval-Update, a cost-efficient LLM-human collaborative framework to dynamically refine evaluation prompts and adapt to shifting criteria with minimal human input. By selectively sampling representative ads for human review and using a critic LLM to generate alignment reports, AutoEval-Update improves evaluation consistency while reducing manual effort. Experiments show the critic LLM suggests meaningful refinements, improving LLM-human agreement. Nonetheless, human oversight remains essential for setting thresholds and validating refinements before deployment.

2025-06-22T00:28:35Z KDD LLM4ECommerce Workshop 2025 Haoran Liu Amir Tahmasbi Ehtesham Sam Haque Purak Jain http://arxiv.org/abs/2506.21598v1 2025-06-19T00:33:05Z

Search Engine marketing teams in the e-commerce industry manage global search engine traffic to their websites with the aim to optimize long-term profitability by delivering the best possible customer experience on Search Engine Results Pages (SERPs). In order to do so, they need to run continuous and rapid Search Marketing A/B tests to continuously evolve and improve their products. However, unlike typical e-commerce A/B tests that can randomize based on customer identification, their tests face the challenge of anonymized users on search engines. On the other hand, simply randomizing on products violates Stable Unit Treatment Value Assumption for most treatments of interest. In this work, we propose leveraging censored observational data to construct bipartite (Search Query to Product Ad or Text Ad) SERP interference networks. Using a novel weighting function, we create weighted projections to form unipartite graphs which can then be use to create clusters to randomized on. We demonstrate this experimental design's application in evaluating a new bidding algorithm for Paid Search. Additionally, we provide a blueprint of a novel system architecture utilizing SageMaker which enables polyglot programming to implement each component of the experimental framework.

2025-06-19T00:33:05Z This is an extended version of our paper published at the AdKDD 2024 workshop, co-located with ACM KDD. CEUR-WS proceedings: https://ceur-ws.org/Vol-3837/paper_12_ceur_paper.pdf Purak Jain Sandeep Appala http://arxiv.org/abs/2506.09084v2 2026-05-23T00:31:27Z

Whole-page optimization (WPO) decides how search and recommendation results are surfaced to users, and large language models (LLMs) open a new route to it by treating page generation as sequence generation. Adapting LLMs to web-scale WPO, however, remains bottlenecked by the need for costly human annotations and by the mismatched granularity between page-level coherence and item-level placement. In this work we show that these two challenges are coupled: implicit user feedback alone suffices for alignment, provided the reward signal is decoupled into two complementary granularities. We propose PageLLM, a reward-based fine-tuning framework that (i) turns implicit feedback into four contrastive preference-pair families covering relevance, ranking, diversity, and redundancy, (ii) learns a coarse page-level reward and a fine item-level reward that captures engagement-sensitive position swaps, and (iii) combines both rewards in PPO-based RLHF over a pre-trained LLM. Extensive experiments on seven Amazon categories against eleven baselines show that neither reward alone is sufficient -- dropping the page-level or item-level signal reduces NDCG@100 by 17.8% and 15.2% respectively, whereas the joint reward improves NDCG@100 by up to 46.8%. Deployed in a 10M-user online A/B test, PageLLM raises GMV by 0.44% and click-through rate by 0.14%, confirming that multi-grained rewards from implicit feedback scale to production WPO. Code and data are available at an anonymized repository.

2025-06-10T08:05:42Z Xinyuan Wang Liang Wu Dongjie Wang Yanjie Fu http://arxiv.org/abs/2505.19643v3 2026-09-17T23:42:49Z

Online A/B tests are the standard tool for data-driven decision-making at scale. Among the design choices with the largest impact on statistical power is the triggering mechanism: how many users to expose and for how long. This often requires forecasting user engagement, i.e., whether enough users will trigger, and when a target participation level will be reached, from limited pilot data. We introduce a Bayesian nonparametric model for predicting both new-user counts and total triggers, accommodating the heavy-tailed engagement patterns typical of web experiments. All predictive quantities can be computed without intensive numerical procedures such as Markov chain Monte Carlo (MCMC) or variational inference. We evaluate on three public datasets (over 450 public benchmark evaluations) and a proprietary benchmark drawn from 759 production A/B tests comprising 1,774 arms. Across the benchmark analyses, our models are competitive and frequently improve accuracy in forecasting new users, total triggers, and time to reach a target sample size compared with state-of-the-art competitors, especially when only a few pilot days are observed.

2025-05-26T07:59:51Z Accepted for publication in the Annals of Applied Statistics. This paper supersedes the technical reports arXiv:2401.14722 and arXiv:2402.03231 by the same authors Mario Beraha Lorenzo Masoero Stefano Favaro Thomas S. Richardson http://arxiv.org/abs/2504.09723v4 2026-03-10T20:51:15Z

A/B testing experiment is a widely adopted method for evaluating UI/UX design decisions in modern web applications. Yet, traditional A/B testing remains constrained by its dependence on the large-scale and live traffic of human participants, and the long time of waiting for the testing result. Through formative interviews with six experienced industry practitioners, we identified critical bottlenecks in current A/B testing workflows. In response, we present AgentA/B, a novel system that leverages Large Language Model-based autonomous agents (LLM Agents) to automatically simulate user interaction behaviors with real webpages. AgentA/B enables scalable deployment of LLM agents with diverse personas, each capable of navigating the dynamic webpage and interactively executing multi-step interactions like search, clicking, filtering, and purchasing. In a demonstrative controlled experiment, we employ AgentA/B to simulate a between-subject A/B testing with 1,000 LLM agents Amazon.com, and compare agent behaviors with real human shopping behaviors at a scale. Our findings suggest AgentA/B can emulate human-like behavior patterns.

2025-04-13T21:10:56Z Yuxuan Lu Ting-Yao Hsu Hansu Gu Limeng Cui Yaochen Xie William Headden Bingsheng Yao Akash Veeragouni Jiapeng Liu Sreyashi Nag Jessie Wang Dakuo Wang http://arxiv.org/abs/2502.04793v2 2025-02-24T13:31:14Z

A/B-tests are a cornerstone of experimental design on the web, with wide-ranging applications and use-cases. The statistical $t$-test comparing differences in means is the most commonly used method for assessing treatment effects, often justified through the Central Limit Theorem (CLT). The CLT ascertains that, as the sample size grows, the sampling distribution of the Average Treatment Effect converges to normality, making the $t$-test valid for sufficiently large sample sizes. When outcome measures are skewed or non-normal, quantifying what "sufficiently large" entails is not straightforward. To ensure that confidence intervals maintain proper coverage and that $p$-values accurately reflect the false positive rate, it is critical to validate this normality assumption. We propose a practical method to test this, by analysing repeatedly resampled A/A-tests. When the normality assumption holds, the resulting $p$-value distribution should be uniform, and this property can be tested using the Kolmogorov-Smirnov test. This provides an efficient and effective way to empirically assess whether the $t$-test's assumptions are met, and the A/B-test is valid. We demonstrate our methodology and highlight how it helps to identify scenarios prone to inflated Type-I errors. Our approach provides a practical framework to ensure and improve the reliability and robustness of A/B-testing practices.

2025-02-07T09:55:24Z Olivier Jeunen http://arxiv.org/abs/2412.01837v1 2024-11-17T10:57:31Z

How to leverage large language model's superior capability in e-commerce recommendation has been a hot topic. In this paper, we propose LLM-PKG, an efficient approach that distills the knowledge of LLMs into product knowledge graph (PKG) and then applies PKG to provide explainable recommendations. Specifically, we first build PKG by feeding curated prompts to LLM, and then map LLM response to real enterprise products. To mitigate the risks associated with LLM hallucination, we employ rigorous evaluation and pruning methods to ensure the reliability and availability of the KG. Through an A/B test conducted on an e-commerce website, we demonstrate the effectiveness of LLM-PKG in driving user engagements and transactions significantly.

2024-11-17T10:57:31Z This paper was accepted by The First International OpenKG Workshop Large Knowledge-Enhanced Models @IJCAI 2024 Menghan Wang Yuchen Guo Duanfeng Zhang Jianian Jin Minnie Li Dan Schonfeld Shawn Zhou http://arxiv.org/abs/2411.01354v1 2024-11-02T20:05:31Z

Recommender systems are widely used AI applications designed to help users efficiently discover relevant items. The effectiveness of such systems is tied to the satisfaction of both users and providers. However, user satisfaction is complex and cannot be easily framed mathematically using information retrieval and accuracy metrics. While many studies evaluate accuracy through offline tests, a growing number of researchers argue that online evaluation methods such as A/B testing are better suited for this purpose. We have employed a variety of algorithms on different types of datasets divergent in size and subject, producing recommendations in various platforms, including media streaming services, digital publishing websites, e-commerce systems, and news broadcasting networks. Notably, our target websites and datasets are in Persian (Farsi) language. This study provides a comparative analysis of a large-scale recommender system that has been operating for the past year across about 70 websites in Iran, processing roughly 300 requests per second collectively. The system employs user-based and item-based recommendations using content-based, collaborative filtering, trend-based methods, and hybrid approaches. Through both offline and online evaluations, we aim to identify where these algorithms perform most efficiently and determine the best method for our specific needs, considering the dataset and system scale. Our methods of evaluation include manual evaluation, offline tests including accuracy and ranking metrics like hit-rate@k and nDCG, and online tests consisting of click-through rate (CTR). Additionally we analyzed and proposed methods to address cold-start and popularity bias.

2024-11-02T20:05:31Z 9 pages, 9 figures Ali Elahi Armin Zirak http://arxiv.org/abs/2408.08913v1 2024-08-14T05:53:02Z

Click-through rate (CTR) prediction is one of the fundamental tasks in the industry, especially in e-commerce, social media, and streaming media. It directly impacts website revenues, user satisfaction, and user retention. However, real-world production platforms often encompass various domains to cater for diverse customer needs. Traditional CTR prediction models struggle in multi-domain recommendation scenarios, facing challenges of data sparsity and disparate data distributions across domains. Existing multi-domain recommendation approaches introduce specific-domain modules for each domain, which partially address these issues but often significantly increase model parameters and lead to insufficient training. In this paper, we propose a Multi-domain Low-Rank Adaptive network (MLoRA) for CTR prediction, where we introduce a specialized LoRA module for each domain. This approach enhances the model's performance in multi-domain CTR prediction tasks and is able to be applied to various deep-learning models. We evaluate the proposed method on several multi-domain datasets. Experimental results demonstrate our MLoRA approach achieves a significant improvement compared with state-of-the-art baselines. Furthermore, we deploy it in the production environment of the Alibaba.COM. The online A/B testing results indicate the superiority and flexibility in real-world production environments. The code of our MLoRA is publicly available.

2024-08-14T05:53:02Z 11 pages. Accepted by RecSys'2024, full paper Zhiming Yang Haining Gao Dehong Gao Luwei Yang Libin Yang Xiaoyan Cai Wei Ning Guannan Zhang http://arxiv.org/abs/2407.14346v2 2024-10-18T13:59:54Z

Accurately retrieving relevant bid keywords for user queries is critical in Sponsored Search but remains challenging, particularly for short, ambiguous queries. Existing dense and generative retrieval models often fail to capture nuanced user intent in these cases. To address this, we propose an approach to enhance query understanding by augmenting queries with rich contextual signals derived from web search results and large language models, stored in an online cache. Specifically, we use web search titles and snippets to ground queries in real-world information and utilize GPT-4 to generate query rewrites and explanations that clarify user intent. These signals are efficiently integrated through a Fusion-in-Decoder based Unity architecture, enabling both dense and generative retrieval with serving costs on par with traditional context-free models. To address scenarios where context is unavailable in the cache, we introduce context glancing, a curriculum learning strategy that improves model robustness and performance even without contextual signals during inference. Extensive offline experiments demonstrate that our context-aware approach substantially outperforms context-free models. Furthermore, online A/B testing on a prominent search engine across 160+ countries shows significant improvements in user engagement and revenue.

2024-07-19T14:28:53Z Accepted to EMNLP 2024 Industry Track. 10 pages, 10 tables, 1 figure Akash Kumar Mohankumar Gururaj K Gagan Madan Amit Singh http://arxiv.org/abs/2406.16629v1 2024-06-24T13:23:00Z

A/B testing is widexly used in the industry to optimize customer facing websites. Many companies employ experimentation specialists to facilitate and improve the process of A/B testing. Here, we present the application of A/B testing to this improvement effort itself, by running experiments on the experimentation process, which we call 'meta-experiments'. We discuss the challenges of this approach using the example of one of our meta-experiments, which helped experimenters to run more sufficiently powered A/B tests. We also point out the benefits of 'dog fooding' for the experimentation specialists when running their own experiments.

2024-06-24T13:23:00Z 6 pages, 2 figures, 1 table Melanie J. I. Müller http://arxiv.org/abs/2405.20565v1 2024-05-31T01:07:37Z

Incorporating Knowledge Graphs into Recommendation has attracted growing attention in industry, due to the great potential of KG in providing abundant supplementary information and interpretability for the underlying models. However, simply integrating KG into recommendation usually brings in negative feedback in industry, due to the ignorance of the following two factors: i) users' multiple intents, which involve diverse nodes in KG. For example, in e-commerce scenarios, users may exhibit preferences for specific styles, brands, or colors. ii) knowledge noise, which is a prevalent issue in Knowledge Enhanced Recommendation (KGR) and even more severe in industry scenarios. The irrelevant knowledge properties of items may result in inferior model performance compared to approaches that do not incorporate knowledge. To tackle these challenges, we propose a novel approach named Knowledge Enhanced Multi-intent Transformer Network for Recommendation (KGTN), comprising two primary modules: Global Intents Modeling with Graph Transformer, and Knowledge Contrastive Denoising under Intents. Specifically, Global Intents with Graph Transformer focuses on capturing learnable user intents, by incorporating global signals from user-item-relation-entity interactions with a graph transformer, meanwhile learning intent-aware user/item representations. Knowledge Contrastive Denoising under Intents is dedicated to learning precise and robust representations. It leverages intent-aware representations to sample relevant knowledge, and proposes a local-global contrastive mechanism to enhance noise-irrelevant representation learning. Extensive experiments conducted on benchmark datasets show the superior performance of our proposed method over the state-of-the-arts. And online A/B testing results on Alibaba large-scale industrial recommendation platform also indicate the real-scenario effectiveness of KGTN.

2024-05-31T01:07:37Z Accept By The Web Conf 2024 (WWW 2024) Industry Track. arXiv admin note: text overlap with arXiv:2204.08807 Ding Zou Wei Wei Feida Zhu Chuanyu Xu Tao Zhang Chengfu Huo 10.1145/3589335.3648296 http://arxiv.org/abs/2403.10049v1 2024-03-15T06:42:23Z

Click-through rate (CTR) prediction is a core task in recommender systems. Existing methods (IDRec for short) rely on unique identities to represent distinct users and items that have prevailed for decades. On one hand, IDRec often faces significant performance degradation on cold-start problem; on the other hand, IDRec cannot use longer training data due to constraints imposed by iteration efficiency. Most prior studies alleviate the above problems by introducing pre-trained knowledge(e.g. pre-trained user model or multi-modal embeddings). However, the explosive growth of online latency can be attributed to the huge parameters in the pre-trained model. Therefore, most of them cannot employ the unified model of end-to-end training with IDRec in industrial recommender systems, thus limiting the potential of the pre-trained model. To this end, we propose a $\textbf{P}$re-trained $\textbf{P}$lug-in CTR $\textbf{M}$odel, namely PPM. PPM employs multi-modal features as input and utilizes large-scale data for pre-training. Then, PPM is plugged in IDRec model to enhance unified model's performance and iteration efficiency. Upon incorporating IDRec model, certain intermediate results within the network are cached, with only a subset of the parameters participating in training and serving. Hence, our approach can successfully deploy an end-to-end model without causing huge latency increases. Comprehensive offline experiments and online A/B testing at JD E-commerce demonstrate the efficiency and effectiveness of PPM.

2024-03-15T06:42:23Z Accepted by ACM Web Conference 2024 (WWW'24) Yuanbo Gao Peng Lin Dongyue Wang Feng Mei Xiwei Zhao Sulong Xu Jinghe Hu 10.1145/3589335.3648329 http://arxiv.org/abs/2403.00133v2 2025-01-28T16:38:57Z

Making ideal decisions as a product leader in a web-facing company is extremely difficult. In addition to navigating the ambiguity of customer satisfaction and achieving business goals, one must also pave a path forward for ones' products and services to remain relevant, desirable, and profitable. Data and experimentation to test product hypotheses are key to informing product decisions. Online controlled experiments by A/B testing may provide the best data to support such decisions with high confidence, but can be time-consuming and expensive, especially when one wants to understand impact to key business metrics such as retention or long-term value. Offline experimentation allows one to rapidly iterate and test, but often cannot provide the same level of confidence, and cannot easily shine a light on impact on business metrics. We introduce a novel, lightweight, and flexible approach to investigating hypotheses, called scenario analysis, that aims to support product leaders' decisions using data about users and estimates of business metrics. Its strengths are that it can provide guidance on trade-offs that are incurred by growing or shifting consumption, estimate trends in long-term outcomes like retention and other important business metrics, and can generate hypotheses about relationships between metrics at scale.

2024-02-29T21:31:45Z Georges Dupret Konstantin Sozinov Carmen Barcena Gonzalez Ziggy Zacks Amber Yuan Benjamin Carterette Manuel Mai Shubham Bansal Gwo Liang Leo Lien Andrey Gatash Roberto Sanchis Ojeda Mounia Lalmas http://arxiv.org/abs/2401.14939v2 2024-05-08T15:25:31Z

Predicting Click-Through Rate (CTR) in billion-scale recommender systems poses a long-standing challenge for Graph Neural Networks (GNNs) due to the overwhelming computational complexity involved in aggregating billions of neighbors. To tackle this, GNN-based CTR models usually sample hundreds of neighbors out of the billions to facilitate efficient online recommendations. However, sampling only a small portion of neighbors results in a severe sampling bias and the failure to encompass the full spectrum of user or item behavioral patterns. To address this challenge, we name the conventional user-item recommendation graph as "micro recommendation graph" and introduce a more suitable MAcro Recommendation Graph (MAG) for billion-scale recommendations. MAG resolves the computational complexity problems in the infrastructure by reducing the node count from billions to hundreds. Specifically, MAG groups micro nodes (users and items) with similar behavior patterns to form macro nodes. Subsequently, we introduce tailored Macro Graph Neural Networks (MacGNN) to aggregate information on a macro level and revise the embeddings of macro nodes. MacGNN has already served Taobao's homepage feed for two months, providing recommendations for over one billion users. Extensive offline experiments on three public benchmark datasets and an industrial dataset present that MacGNN significantly outperforms twelve CTR baselines while remaining computationally efficient. Besides, online A/B tests confirm MacGNN's superiority in billion-scale recommender systems.

2024-01-26T15:23:25Z 11 pages, 7 figures, accepted by The Web Conference 2024 Hao Chen Yuanchen Bei Qijie Shen Yue Xu Sheng Zhou Wenbing Huang Feiran Huang Senzhang Wang Xiao Huang http://arxiv.org/abs/2401.07769v3 2024-08-04T14:46:14Z

The recommendation has been playing a key role in many industries, e.g., e-commerce, streaming media, social media, etc. Recently, a new recommendation scenario, called Trigger-Induced Recommendation (TIR), where users are able to explicitly express their instant interests via trigger items, is emerging as an essential role in many e-commerce platforms, e.g., Alibaba.com and Amazon. Without explicitly modeling the user's instant interest, traditional recommendation methods usually obtain sub-optimal results in TIR. Even though there are a few methods considering the trigger and target items simultaneously to solve this problem, they still haven't taken into account temporal information of user behaviors, the dynamic change of user instant interest when the user scrolls down and the interactions between the trigger and target items. To tackle these problems, we propose a novel method -- Deep Evolutional Instant Interest Network (DEI2N), for click-through rate prediction in TIR scenarios. Specifically, we design a User Instant Interest Modeling Layer to predict the dynamic change of the intensity of instant interest when the user scrolls down. Temporal information is utilized in user behavior modeling. Moreover, an Interaction Layer is introduced to learn better interactions between the trigger and target items. We evaluate our method on several offline and real-world industrial datasets. Experimental results show that our proposed DEI2N outperforms state-of-the-art baselines. In addition, online A/B testing demonstrates the superiority over the existing baseline in real-world production environments.

2024-01-15T15:27:24Z 7 pages, 6 figures, accepted by the 17th ACM International Conference on Web Search and Data Mining(WSDM'2024) Zhibo Xiao Luwei Yang Tao Zhang Wen Jiang Wei Ning Yujiu Yang 10.1145/3616855.3635829 http://arxiv.org/abs/2312.00372v2 2023-12-04T11:42:35Z

Information retrieval in real-time search presents unique challenges distinct from those encountered in classical web search. These challenges are particularly pronounced due to the rapid change of user search intent, which is influenced by the occurrence and evolution of breaking news events, such as earthquakes, elections, and wars. Previous dense retrieval methods, which primarily focused on static semantic representation, lack the capacity to capture immediate search intent, leading to inferior performance in retrieving the most recent event-related documents in time-sensitive scenarios. To address this issue, this paper expands the query with event information that represents real-time search intent. The Event information is then integrated with the query through a cross-attention mechanism, resulting in a time-context query representation. We further enhance the model's capacity for event representation through multi-task training. Since publicly available datasets such as MS-MARCO do not contain any event information on the query side and have few time-sensitive queries, we design an automatic data collection and annotation pipeline to address this issue, which includes ModelZoo-based Coarse Annotation and LLM-driven Fine Annotation processes. In addition, we share the training tricks such as two-stage training and hard negative sampling. Finally, we conduct a set of offline experiments on a million-scale production dataset to evaluate our approach and deploy an A/B testing in a real online system to verify the performance. Extensive experimental results demonstrate that our proposed approach significantly outperforms existing state-of-the-art baseline methods.

2023-12-01T06:30:31Z Nan Yang Shusen Zhang Yannan Zhang Xiaoling Bai Hualong Deng Tianhua Zhou Jin Ma http://arxiv.org/abs/2311.07870v2 2023-11-16T17:21:15Z

Web-scale ranking systems at Meta serving billions of users is complex. Improving ranking models is essential but engineering heavy. Automated Machine Learning (AutoML) can release engineers from labor intensive work of tuning ranking models; however, it is unknown if AutoML is efficient enough to meet tight production timeline in real-world and, at the same time, bring additional improvements to the strong baselines. Moreover, to achieve higher ranking performance, there is an ever-increasing demand to scale up ranking models to even larger capacity, which imposes more challenges on the efficiency. The large scale of models and tight production schedule requires AutoML to outperform human baselines by only using a small number of model evaluation trials (around 100). We presents a sampling-based AutoML method, focusing on neural architecture search and hyperparameter optimization, addressing these challenges in Meta-scale production when building large capacity models. Our approach efficiently handles large-scale data demands. It leverages a lightweight predictor-based searcher and reinforcement learning to explore vast search spaces, significantly reducing the number of model evaluations. Through experiments in large capacity modeling for CTR and CVR applications, we show that our method achieves outstanding Return on Investment (ROI) versus human tuned baselines, with up to 0.09% Normalized Entropy (NE) loss reduction or $25\%$ Query per Second (QPS) increase by only sampling one hundred models on average from a curated search space. The proposed AutoML method has already made real-world impact where a discovered Instagram CTR model with up to -0.36% NE gain (over existing production baseline) was selected for large-scale online A/B test and show statistically significant gain. These production results proved AutoML efficacy and accelerated its adoption in ranking systems at Meta.

2023-11-14T03:00:50Z Hang Yin and Kuang-Hung Liu contribute equally Hang Yin Kuang-Hung Liu Mengying Sun Yuxin Chen Buyun Zhang Jiang Liu Vivek Sehgal Rudresh Rajnikant Panchal Eugen Hotaj Xi Liu Daifeng Guo Jamey Zhang Zhou Wang Shali Jiang Huayu Li Zhengxing Chen Wen-Yen Chen Jiyan Yang Wei Wen