Industrial generative recommendation systems operating over discrete Semantic IDs (SIDs) are largely behavior-driven, and thus struggle to proactively activate latent demand before explicit user signals emerge, leading to intent cold-start. To address this, we propose RIGER (Reasoning-grounded Intent injection for GE nerative Recommendation), a deployable two-stage framework that integrates offline large language model (LLM) reasoning into an online generative recommender under strict latency constraints. Offline, to ensure scalable deployment, we distill the latent-intent inference capability of a strong LLM into a lightweight forecasting model using an automated data curation pipeline---leveraging judge-guided prompt calibration and future-query-guided rejection filtering. Online, to bridge the representation mismatch between free-form textual intents and the discrete SID token space, predicted intents are converted into SID-native tokens through a behavior-grounded mapping and injected into the deployed decoder-only retrieval backbone. We further fine-tune the model with beam-aware GRPO, introducing a hierarchical intent-alignment exploration reward in SID space while preserving exploitation behavior through KL regularization. Offline evaluations demonstrate a substantial increase in intent-aligned density and diversity with only a marginal reduction in hindsight recall, indicating that RIGER effectively enhances proactive intent exploration while preserving its capability to exploit historical behaviors. In a large-scale e-commerce display advertising system, RIGER improves clicks by 1.6% and advertiser spend by 1.3%.

Click-through rate (CTR) prediction is a pivotal component in large-scale industrial systems. Historically, CTR prediction paradigms have been confined to monolithic architectures governed by a single-policy optimization process. However, such isolated learning paths lack the intrinsic evolutionary mechanisms necessary for optimal convergence. Without policy diversity and internal competition, models tend to get trapped in local optima as performance reaches saturation, hindering further breakthroughs in modeling capacity. In this paper, we propose DPEO (Dynamic Preference Evolution Optimization), a co-evolutionary framework that transforms CTR modeling into a dynamic policy contention task. DPEO decouples the monolithic architecture into dual sub-learners to induce policy diversity, constructing an internal preference landscape without external rewards. A performance-driven Role Arbiter then dynamically designates the superior sub-learner as the Reference Policy and the other sub-learner as the Target Policy per batch, driving continuous model evolution. Through an asymmetric gradient flow, the target policy is optimized to surpass the reference policy in both probability and logit spaces. This process drives a co-evolution, enabling the sub-learners to serve as alternating evolutionary benchmarks and 'self-evolve' toward the global optimum. Extensive experiments on public benchmarks and a massive industrial dataset with over 10 billion samples demonstrate that DPEO significantly outperforms state-of-the-art models.

Generative Recommendation (GenRec) has emerged as a significant evolutionary direction in the field of recommendation systems in recent years. However, during the reinforcement learning (RL) alignment stage of end-to-end GenRec, a core challenge is that the low signal-to-noise ratio (SNR) of feedback signals triggers reward hacking, which in turn leads to severe distributional drift. Massive click noise and sparse rewards result in highly unstable policy gradients, making it difficult to balance reward optimization and generative stability. To address this, we propose AIPO (Adaptive Anchored Intent-aware Policy Optimization) framework. AIPO synergistically enhances both optimization stability and performance from both data and prior anchoring perspectives. First, it introduces an intent-aware asymmetric resampling mechanism to purify high-confidence conversion intent data, thereby amplifying the gradient contribution of high-value paths. Second, it introduces a prior anchoring mechanism, which dynamically regulates its intensity through a barrier mapping based on the degree of policy distributional drift, mitigating distributional drift and reward hacking caused by low-SNR data. Offline experiments on both public benchmarks and JD's industrial datasets validated the superior performance of AIPO.

The ranking stage serves as the central optimization and allocation hub in advertising systems, governing economic value distribution through eCPM and orchestrating the user-centric blending of organic and advertising content. Prevailing ranking models often rely on fragmented modules and hand-crafted features, limiting their ability to interpret complex user intent. This challenge is further amplified in location-based services such as food delivery, where user decisions are shaped by dynamic spatial, temporal, and individual contexts. To address these limitations, we propose a novel generative framework that reframes ranking as a context comprehension task, modeling heterogeneous signals in a unified architecture. Our architecture consists of two core components: the Generative Contextual Encoder (GCE) and the Generative Contextual Fusion (GCF). The GCE comprises three specialized modules: a Personalized Context Enhancer (PCE) for user-specific modeling, a Collective Context Enhancer (CCE) for group-level patterns, and a Dynamic Context Enhancer (DCE) for real-time situational adaptation. The GCF module then seamlessly integrates these contextual representations through low-rank adaptation. Extensive experiments confirm that our method achieves significant gains in critical business metrics, including click-through rate and platform revenue. We have successfully deployed our method on a large-scale food delivery advertising platform, demonstrating its substantial practical impact. This work pioneers a new perspective on generative recommendation and highlights its practical potential in industrial advertising systems.

Trigger-Induced Recommendation (TIR), which aims to predict user interest based on a trigger item, has gained considerable traction on e-commerce platforms. Current TIR methods typically analyze user intent by integrating explicit interest in the trigger item and implicit interest derived from user historical behaviors. However, these methods often overlook the contextual information and occurring scenarios related to the trigger, resulting in an undue emphasis on isolated trigger items and a consequently restrictive understanding of users' short-term intentions. To address these challenges, we propose a novel scenario-aware multi-trigger induction method featuring three key enhancements: (1) The Context Modeling Network learns contextual information associated with the trigger during the request, improving the understanding of users' real intentions regarding the trigger item; (2) The Multi-Trigger Learning Network introduces user latent triggers from various scenarios to uncover users' potential external preferences; (3) The Scenario Induction Network captures the characteristics of the scenarios in which triggers occur and performs induction to yield scenario-aware user intentions prediction. We validate our approach through experiments on multiple industrial datasets, demonstrating the model's effectiveness. Furthermore, we have integrated the model into an online advertising system, achieving a 5.46% improvement in Click-Through Rate (CTR).

E-commerce has become an integral part of daily life, and the ability to effectively retrieve items relevant to a user's query is crucial for enhancing the shopping experience. Embedding-based retrieval (EBR) has proven to be an effective approach in industrial e-commerce search systems. This method involves training models to generate high-quality representations of queries and items, followed by the use of efficient approximate nearest neighbor (ANN) search techniques to find relevant items. However, current EBR methods face several critical limitations: (1) multiple EBR branches often retrieve overlapping or redundant result sets; (2) allocation and utilization of computational resources remain suboptimal, hindering performance; (3) the differentiation modeling of features is somewhat neglected, which restricts the system's ability to retrieve diverse and representative results. These issues hinder the retrieval performance of online search systems. In this paper, we introduce a novel e-commerce search framework called the Uniform Embedding-based Retrieval Framework (UniERF). This framework is meticulously designed to incorporate diverse samples for joint model training, enabling the model to effectively leverage both the semantic information of queries and the personalized features of different users. Extensive offline and online experiments demonstrate that UniERF surpasses baseline methods across various evaluation metrics. UniERF has been successfully implemented in the existing retrieval system at JD.COM, a renowned online shopping website.

Relevance modeling in e-commerce search remains challenged by semantic gaps in term-matching methods (e.g., BM25) and neural models' reliance on the scarcity of domain-specific hard samples. We propose ADORE, a self-sustaining framework that synergizes three innovations: (1) A Rule-aware Relevance Discrimination module, where a Chain-of-Thought LLM generates intent-aligned training data, refined via Kahneman-Tversky Optimization (KTO) to align with user behavior; (2) An Error-type-aware Data Synthesis module that auto-generates adversarial examples to harden robustness; and (3) A Key-attribute-enhanced Knowledge Distillation module that injects domain-specific attribute hierarchies into a deployable student model. ADORE automates annotation, adversarial generation, and distillation, overcoming data scarcity while enhancing reasoning. Large-scale experiments and online A/B testing verify the effectiveness of ADORE. The framework establishes a new paradigm for resource-efficient, cognitively aligned relevance modeling in industrial applications.

The retrieval system is a crucial module in e-commerce search advertising that matches user queries with ads. The diverse expressions of users often produce massive tail queries that cannot match merchant bidwords, leading to poor retrieval efficiency. Existing methods, such as query log mining and vector matching, fail to optimize relevance, authenticity, and ad revenue of the rewrite.

Accurate Click-Through Rate (CTR) prediction is vital for search engines and recommendation systems, yet it is often hindered by the ''cold start problem'', which arises from insufficient historical data for new users. Recent approaches have sought to tackle this by training encoder-decoder networks on data from warm users to generate virtual behavior embeddings for cold users. However, these methods have shortcomings in terms of simplistic encoding techniques for warm user behaviors and direct utilization of virtual behavior embeddings, leading to limitations in user interest expression and generalization. To address these challenges, we propose a novel method that leverages Graph Isomorphism Networks (GIN) for cohort modeling within CTR prediction. GIN effectively captures high-order user-item interactions, providing a more nuanced understanding of users' diverse interests. Additionally, the cohort modeling strategy minimizes deviations in constructed embeddings, enhancing the model's generalization abilities. We validate our approach through experiments on public and industrial datasets, demonstrating significant improvements for both warm and cold users compared to existing methodologies. Furthermore, we implemented the GIN Cohort Modeling (GINCM) in a large-scale online advertising system, optimizing for both pre-computation and real-time processing to reduce latency. The implementation yields notable enhancements of 2.13% in CTR and Revenue Per Mille(RPM), showcasing the practical effectiveness and real-world applicability of our model.

Accurate click-through rate (CTR) prediction is critical for online advertising, relying on regular features like browsing history and demographics and post-event features such as exposed position and detailed page behaviors. However, post-event features, unavailable during inference, often face training-inference inconsistency and low coverage issues, especially post-click features like dwell time that are available only for clicked items. To address these challenges, we propose Causal Optimal Transport (COT), a novel framework that (1) generates pseudo post-click features via semi-supervised pseudo-labeling (2) causally generates accurate feature distributions using a Causal Distribution Shaper (CDS), and (3) refines generated features through optimal transport to minimize distributional divergence, facilitating further knowledge transfer. Experiments on real-world data confirm COT's superiority and practical efficacy in enhancing CTR prediction via improved user interest modeling and bias mitigation. Theoretical guarantees underpin the framework's robustness.

In the dynamic landscape of online advertising, decoding user intent remains a pivotal challenge, particularly in the context of query classification. Swift classification models, exemplified by FastText, cater to the demand for real-time responses but encounter limitations in handling intricate queries. Conversely, accuracy-centric models like BERT introduce challenges associated with increased latency. This paper undertakes a nuanced exploration, navigating the delicate balance between efficiency and accuracy. It unveils FastText's latent potential as an 'online dictionary' for historical queries while harnessing the semantic robustness of BERT for novel and complex scenarios. The proposed Distribution-Diverse Multi-Expert (DDME) framework employs multiple teacher models trained from diverse data distributions. Through meticulous data categorization and enrichment, it elevates the classification performance across the query spectrum. Empirical results within the JD ads search system validate the superiority of our proposed approaches.

Online recommenders have attained growing interest and created great revenue for businesses. Given numerous users and items, incremental update becomes a mainstream paradigm for learning large-scale models in industrial scenarios, where only newly arrived data within a sliding window is fed into the model, meeting the strict requirements of quick response. However, this strategy would be prone to overfitting to newly arrived data. When there exists a significant drift of data distribution, the long-term information would be discarded, which harms the recommendation performance. Conventional methods address this issue through native model-based continual learning methods, without analyzing the data characteristics for online recommenders. To address the aforementioned issue, we propose an incremental update framework for online recommenders with Data-Driven Prior (DDP), which is composed of Feature Prior (FP) and Model Prior (MP). The FP performs the click estimation for each specific value to enhance the stability of the training process. The MP incorporates previous model output into the current update while strictly following the Bayes rules, resulting in a theoretically provable prior for the robust update. In this way, both the FP and MP are well integrated into the unified framework, which is model-agnostic and can accommodate various advanced interaction models. Extensive experiments on two publicly available datasets as well as an industrial dataset demonstrate the superior performance of the proposed framework. © 2023 Copyright held by the owner/author(s). Publication rights licensed to ACM.

It is noted that Click-Through-Rate(CTR) prediction plays an important part in recommendation systems and online advertising. Over the past few years, numerous studies have been conducted to improve the accuracy of CTR prediction by exploring data inherent patterns. These studies indicate that training CTR models with group-specific parameters on divided data groups can lead to significant improvements. However, most works generally divide groups manually with some prior knowledge, and such a fixed group division method may hinder the expression of user common interests. To address this limitation, we propose a novel group parameter modeling method, where the user group division and group parameter learning processes are completed in an automatic and dynamic way. Our method employs a three-stage approach, consisting of group information selection, group representation learning, and group parameter generation, which allows the efficient expression of user common interests. We conduct experiments on both public datasets and industrial datasets, and the experimental results demonstrate the effectiveness of our method. We have also deployed the model in an online advertising system and observed significant improvements in both CTR and Revenue Per Mille (RPM).

Recently, Graph Convolutional Network (GCN) based methods have become novel state-of-the-arts for Collaborative Filtering (CF) based Recommender Systems. To obtain users' preferences over different items, it is a common practice to learn representations of users and items by performing embedding propagation on a user-item bipartite graph, and then calculate the preference scores based on the representations. However, in most existing algorithms, user/item representations are generated independently of target items/users. To address this problem, we propose a novel graph attention model named Bilateral Interactive GCN (BI-GCN), which introduces bilateral interactive guidance into each user-item pair and thus leads to target-aware representations for preference prediction. Specifically, to learn the user/item representation from its neighborhood, we assign higher attention weights to those neighbors similar to the target item/user. By this manner, we can obtain target-aware representations, i.e., the information of the target item/user is explicitly encoded in the corresponding user/item representation, for more precise matching. Extensive experiments on three benchmark datasets demonstrate the effectiveness and robustness of BI-GCN.

Large-scale industrial systems often include multiple scenarios to satisfy diverse user needs. The common approach of using one model per scenario does not scale well and not suitable for minor scenarios with limited samples. An solution is to train a model on all scenarios, which can introduce domination and bias from the main scenario. MMoE-like structures have been proposed for multi-scenario prediction, but they do not explicitly address the issue of gradient unbalancing. This work proposes an adaptive loss harmonizing (ALH) algorithm for multi-scenario CTR prediction. It dynamically adjusts the learning speed for balanced training and improved performance. Experiments on real industrial datasets and rigorous A/B testing prove our method’s superiority.

Industrial systems such as recommender systems and online advertising, have been widely equipped with multi-stage architectures, which are divided into several cascaded modules, including matching, pre-ranking, ranking and re-ranking. As a critical bridge between matching and ranking, existing pre-ranking approaches mainly endure sample selection bias (SSB) problem owing to ignoring the entire-chain data dependence, resulting in sub-optimal performances. In this paper, we rethink pre-ranking system from the perspective of the entire sample space, and propose Entire-chain Cross-domain Models (ECM), which leverage samples from the whole cascaded stages to effectively alleviate SSB problem. Besides, we design a fine-grained neural structure named ECMM to further improve the pre-ranking accuracy. Specifically, we propose a cross-domain multi-tower neural network to comprehensively predict for each stage result, and introduce the sub-networking routing strategy with L0 regularization to reduce computational costs. Evaluations on real-world large-scale traffic logs demonstrate that our pre-ranking models outperform SOTA methods while time consumption is maintained within an acceptable level, which achieves better trade-off between efficiency and effectiveness.

Click-through rate (CTR) Prediction is of great importance in real-world online ads systems. One challenge for the CTR prediction task is to capture the real interest of users from their clicked items, which is inherently influenced by presented positions of items, i.e., more front positions tend to obtain higher CTR values. Therefore, It is crucial to make CTR models aware of the exposed position of the items. A popular line of existing works focuses on explicitly model exposed position by result randomization which is expensive and inefficient, or by inverse propensity weighting (IPW) which relies heavily on the quality of the propensity estimation. Another common solution is modeling position as features during offline training and simply adopting fixed value or dropout tricks when serving. However, training-inference inconsistency can lead to sub-optimal performance. This work proposes a simple yet efficient knowledge distillation framework to model the impact of exposed position and leverage position information to improve CTR prediction. We demonstrate the performance of our proposed method on a real-world production dataset and online A/B tests, achieving significant improvements over competing baseline models. The proposed method has been deployed in the real world online ads systems of JD, serving main traffic of hundreds of millions of active users.

The exposure sequence is being actively studied for user interest modeling in Click-Through Rate (CTR) prediction. However, the existing methods for exposure sequence modeling bring extensive computational burden and neglect noise problems, resulting in an excessively latency and the limited performance in online recommenders. In this paper, we propose to address the high latency and noise problems via Gating-adapted wavelet multiresolution analysis (Gama), which can effectively denoise the extremely long exposure sequence and adaptively capture the implied multi-dimension user interest with linear computational complexity. This is the first attempt to integrate non-parametric multiresolution analysis technique into deep neural network to model user exposure sequence. Extensive experiments on large scale benchmark dataset and real production dataset confirm the effectiveness of Gama for exposure sequence modeling, especially in cold-start scenarios. Benefited from its low latency and high effecitveness, Gama has been deployed in our real large-scale industrial recommender, successfully serving over hundreds of millions users.

We propose a general Variational Embedding Learning Framework (VELF) for alleviating the severe cold-start problem in CTR prediction. VELF addresses the cold start problem via alleviating over-fits caused by data-sparsity in two ways: learning probabilistic embedding, and incorporating trainable and regularized priors which utilize the rich side information of cold start users and advertisements (Ads). The two techniques are naturally integrated into a variational inference framework, forming an end-to-end training process. Abundant empirical tests on benchmark datasets well demonstrate the advantages of our proposed VELF. Besides, extended experiments confirmed that our parameterized and regularized priors provide more generalization capability than traditional fixed priors.

Click-through rate (CTR) prediction plays a crucial role in sponsored search advertising (search ads). User click behavior usually showcases strong comparison patterns among relevant/competing items within the user awareness. Explicit user awareness could be characterized by user behavior sequence modeling, which however suffers from issues such as cold start, behavior noise and hidden channels. Instead, in this paper, we study the problem of modeling implicit user awareness about relevant/competing items. We notice that candidate items of the CTR prediction model could play as surrogates for relevant/competing items within the user awareness. Motivated by this finding, we propose a novel framework, named CIM (Candidate Item Modeling), to characterize users’ awareness on candidate items. CIM introduces an additional module to encode candidate items into a context vector and therefore is plug-and-play for existing neural network-based CTR prediction models. Offline experiments on a ten-billion-scale production dataset collected from the real traffic of a search advertising system, together with the corresponding online A/B testing, demonstrate CIM’s superior performance. Notably, CIM has been deployed in production at JD.com, serving the main traffic of hundreds of millions of users, which shows great application value. Our code and dataset are available at https://github.com/kaifuzheng/cim.

Click-through rate (CTR) prediction based on deep neural networks has made significant progress in recommendation systems. However, these methods often suffer from CTR underestimation due to insufficient impressions for long-tail items. When formalizing CTR prediction as a contextual bandit problem, exploration methods provide a natural solution addressing this issue. In this paper, we first benchmark state-of-the-art exploration methods in the recommendation system setting. We find that the combination of gradient-based uncertainty modeling and Thompson Sampling achieves a significant advantage. On the basis of the benchmark, we further propose a general enhancement strategy, Underestimation Refinement (UR), which explicitly incorporates the prior knowledge that insufficient impressions likely leads to CTR underestimation. This strategy is applicable to almost all the existing exploration methods. Experimental results validate UR's effectiveness, achieving consistent improvement across all baseline exploration methods.

Embedding mechanism plays an important role in Click-Through-Rate (CTR) prediction. Essentially, it tries to learn a new feature space with some learned latent properties as the basis, and maps the high dimensional and categorical raw data to dense, rich and expressive representations, i.e., the embedding features. Current researches usually focus on learning the interactions through operations on the whole embedding features without considering the relations among the learned latent properties. In this paper, we find it has clear positive effects on CTR prediction to model such relations and propose a novel Dimension Relation Module (DRM) to capture them through dimension recalibration. We show that DRM can improve the performance of existing models consistently and the improvements are more obvious when the embedding dimension is higher. We further boost Field-wise and Element-wise embedding methods with our DRM and name this new model FED network. Extensive experiments demonstrate that FED is very powerful in CTR prediction task and achieves new state-of-the-art results on Criteo, Avazu and JD.com datasets.

Targeting system is an essential part of computational advertising. It allows advertisers to select and reach their targeted users. Due to various advertising goals and the demand for making budget plans, advertisers have a strong will to configure the final targeting results, or they can become very cautious in spending money on advertising campaigns. Meanwhile, to guarantee the advertising performance, the targeted users should also be relevant to the ads of the advertisers. Recent targeting methods are mainly based on tags produced by the Data Management Platform (DMP) which is easy for the advertisers to configure the targeting results. However, in such methods, the relevance between the targeted users and ads is not technically evaluated and cannot be guaranteed. The biggest challenge is that it is hard for a machine learning model to both model the relevance and take account of the advertiser’s configuration demands. In this paper, we propose a novel relevance-driven and configurable targeting framework called Smart Targeting to solve the problem. Specifically, different from Tag-wise Targeting, we first use a relevance model to retrieve the most relevant users for the ads. To further enable the advertisers to configure the final results, we develop a Delay Intervention Mechanism to leverage the power of DMP. As far as we know, this is the first attempt of combining relevance modeling and advertiser intervention into a unified targeting system. We implement and evaluate our framework on JD.com platform with over 300 million users and the results show that it can bring significant improvements to the core indicators such as CTR and eCPM. The long term monitoring also demonstrates that Smart Targeting gradually becomes the most popular targeting tool after its release.

The ad-ranking module is the core of the advertising recommender system. Existing ad-ranking modules are mainly based on the deep neural network click-through rate prediction model. Recently an innovative ad-ranking paradigm called DNN-MAB has been introduced to address DNN-only paradigms’ weakness in perceiving highly dynamic user intent over time. We introduce the DNN-MAB paradigm into our ad-ranking system to alleviate the Matthew effect that harms the user experience. Due to data sparsity, however, the actual performance of DNN-MAB is lower than expected. In this paper, we propose an innovative ad-ranking paradigm called DNN-VMAB to solve these problems. Based on vectorization and clustering, it utilizes latent collaborative information in user behavior data to find a set of ads with higher relativity and diversity. As an integration of the essences of classical collaborative filtering, deep click-through rate prediction model, and contextual multi-armed bandit, it can improve platform revenue and user experience. Both offline and online experiments show the advantage of our new algorithm over DNN-MAB and some other existing algorithms.