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 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.

In e-commerce platforms, there is an emerging type of content that tells a “story” about some merchandise in the form of multimedia (text, images, video), which is namedstorytelling. Well told stories, like advertisements, can inspire users to purchase the related products. Thus, e-commerce service provider is keen to disseminate storytelling items to potentially interested users. We address this requirement by a cross-domain personalized recommendation approach. Because storytelling is a new type of content, its related user actions are much less, more sparse than product-related user actions, thus we propose to use product-domain user actions to assist the identification of user preferences and to make storytelling recommendations. Our method has two technical contributions. First, since the user behavior patterns are different across the storytelling domain and the product domain, we propose an attentional domain-transfer network, which effectively selects the relevant items in the two domains to characterize user preferences. Second, although storytelling is about product, between the two domains there is a large gap: product description is objective and categorical, like “keywords,” but storytelling is close to human language. To bridge the domain gap, we propose a dual-domain contrastive adversarial learning method to pre-train the feature extractors for storytelling and product simultaneously. We conduct experiments on two industrial datasets, and the results demonstrate the advantage of our proposed method that consistently outperforms the state-of-the-art methods. Besides, our method can be used to recommend storytelling to products, which is a desired functionality for product providers. Our code and models are publicly available.

Deep learning is known to be effective at automating the generation of representations, which eliminates the need for handcrafted features. For the task of personalized recommendation, deep learning-based methods have achieved great success by learning efficient representations of multimedia items, especially images and videos. Previous works usually adopt simple, single-modality representations of user interest, such as user embeddings, which cannot fully characterize the diversity and volatility of user interest. To address this problem, in this paper we focus on learning and fusing multiple kinds of user interest representations by leveraging deep networks. Specifically, we consider efficient representations of four aspects of user interest: first, we use latent representation, i.e. user embedding, to profile the overall interest; second, we propose item-level representation, which is learned from and integrates the features of a user's historical items; third, we investigate neighbor-assisted representation, i.e. using neighboring users' information to characterize user interest collaboratively; fourth, we propose category-level representation, which is learned from the categorical attributes of a user's historical items. In order to integrate these multiple user interest representations, we study both early fusion and late fusion; where for early fusion, we study different fusion functions. We validate the proposed method on two real-world video recommendation datasets for micro-video and movie recommendations, respectively. Experimental results demonstrate that our method outperforms existing state-of-the-arts by a significant margin. Our code is publicly available.

This paper describes our solution for the Content-Based Video Relevance Prediction (CBVRP) challenge, where the task is to predict user click-through behavior on new TV series or new movies according to the user's historical behavior. We consider the task as a session-based recommendation problem and we focus on the modeling of the session. Thus, we use the Bidirectional Encoder Representations from Transformer (BERT) methodology and propose a BERT for session-based recommendation (BERT4SessRec) method. Our method has two stages: in the pre-training stage, we use all sessions as training data and train the bidirectional session encoder with the masking trick; in the fine-tuning stage, we use the provided click-through data and train the click-through prediction network. Our method achieves session representations with the help of BERT, which effectively captures the bidirectional correlation in each session. In addition, the pre-training stage makes full use of all sessions, overcoming the positive-negative imbalance problem of the click-through data. We report the results of using different kinds of features on the test set of the challenge, which verify the effectiveness of our method.

Micro-video sharing gains great popularity in recent years, which calls for effective recommendation algorithm to help user find their interested micro-videos. Compared with traditional online (e.g. YouTube) videos, micro-videos contributed by grass-root users and taken by smartphones are much shorter (tens of seconds) and more short of tags or descriptive text, making the recommendation of micro-videos a challenging task. In this paper, we investigate how to model user's historical behaviors so as to predict the user's click-through of micro-videos. Inspired by the recent deep network-based methods, we propose a Temporal Hierarchical Attention at Category- and Item-Level (THACIL) network for user behavior modeling. First, we use temporal windows to capture the short-term dynamics of user interests; Second, we leverage a category-level attention mechanism to characterize user's diverse interests, as well as an item-level attention mechanism for fine-grained profiling of user interests; Third, we adopt forward multi-head self-attention to capture the long-term correlation within user behaviors. Our proposed THACIL network was tested on MicroVideo-1.7M, a new dataset of 1.7 million micro-videos, coming from real data of a micro-video sharing service in China. Experimental results demonstrate the effectiveness of the proposed method in comparison with the state-of-the-art solutions.

This paper describes our proposed method for the Content-Based Video Relevance Prediction (CBVRP) challenge. Our method is based on deep learning, i.e. we train a deep network to predict the relevance between two video sequences from their features. We explore the usage of second-order relevance, both in preparing training data, and in extending the deep network. Second-order relevance refers to e.g. the relevance between x and z if x is relevant to y and y is relevant to z. In our proposed method, we use second-order relevance to increase positive samples and decrease negative samples, when preparing training data. We further extend the deep network with an attention module, where the attention mechanism is designed for second-order relevant video sequences. We verify the effectiveness of our method on the validation set of the CBVRP challenge.