Large-scale search and recommendation systems typically adopt a cascaded architecture of retrieval, pre-ranking, ranking, and re-ranking to balance efficiency and accuracy. However, pre-ranking still faces challenges of behavioral sparsity, limited interest diversity, and computational latency. We propose the Query-Attention Dual-Stream (QADS) framework to address these issues. QADS partitions user behaviors into strongly and weakly correlated streams and further decomposes them into fine-grained subsequences guided by domain knowledge. A query-centric attention mechanism reduces complexity from O(N) to O(1), enabling efficient inter- and intra-sequence modeling. A contrastive cross-category transfer module propagates dense patterns from weakly correlated to sparse domains, while a latency-aware parallel inference architecture further reduces delay by 36%. Experiments on public and industrial datasets show that QADS delivers significant performance improvements and has been successfully deployed in large-scale e-commerce search systems.

Generative retrieval formulates document retrieval as an identifier generation task. While prevailing methods increasingly adopt Semantic IDs (SIDs), their opaque nature and rigid mappings struggle with the dynamic inventory and strict interpretability requirements of E-commerce search. Furthermore, generating accurate targets from brief queries against noisy, loosely structured item titles remains a practical challenge. To address these issues, we propose a Query-to-Multi-Span generative retrieval framework tailored for E-commerce. Instead of relying on opaque SIDs or raw titles, our method simplifies the process by generating interpretable multispan identifiers from queries. We align the autoregressive model with user preferences using click logs, and employ a constraintbased beam search to isolate key spans for final item retrieval. This approach explicitly bridges generative models with robust constraint matching, ensuring both matching accuracy and transparency. Extensive offline evaluations demonstrate competitive retrieval performance, and online A/B tests confirm its effectiveness in delivering measurable conversion gains in a production environment.

Multi-agent collaboration is the latest trending method to build conversational recommender systems (CRS), especially with the widespread use of Large Language Models (LLMs) recently. Typically, these systems employ several LLM agents, each serving distinct roles to meet user needs. In an industrial setting, it’s essential for a CRS to exhibit low first token latency (i.e., the time taken from a user’s input until the system outputs its first response token.) and high scalability—for instance, minimizing the number of LLM inferences per user request—to enhance user experience and boost platform profit. For example, JD.com’s baseline CRS features two LLM agents and a search API but suffers from high first token latency and requires two LLM inferences per request (LIPR), hindering its performance. To address these issues, we introduce a Hybrid Multi-Agent Collaborative Recommender System (Hybrid-MACRS). It includes a central agent powered by a fine-tuned proprietary LLM and a search agent combining a related search module with a search engine. This hybrid system notably reduces first token latency by about 70% and cuts the LIPR from 2 to 1. We conducted thorough online A/B testing to confirm this approach’s efficiency.

One in 36 children in the United States has autism. Numerous robotic intervention systems have been proposed for children with autism. It is widely acknowledged that engagement with the robotic system is important for intervention success. Visual attention toward the robot can be used as a proxy for engagement. However, less is known about how to maintain and enhance visual attention within a robotic system, and how the variation of visual attention may influence adaptation of dynamic intervention protocols to improve outcomes. Therefore, in this work, we propose a new metric, System Capture Ratio (SCR), that can be automatically quantified in real-time, to measure a participant’s visual attention within a robotic system designed for joint attention intervention for children with autism. Then, we demonstrate that compared to a static intervention protocol, a dynamic intervention protocol can help sustain visual attention and thus achieve a significant performance improvement. The results support the implementation of adaptive and dynamic robotic intervention protocols for autistic children that are based on SCR, offering suggestions on designing effective multi-session studies for autism intervention.

Most existing methods adopt the quadrilateral or rotated rectangle representation to detect multi-oriented objects. Yet, the same oriented object may correspond to several different representations, due to different vertex ordering, or angular periodicity and edge exchangeability. To ensure the uniqueness of the representation, some engineered rules are usually added. This makes these methods suffer from discontinuity problem, resulting in degraded performance for objects around some orientation. In this article, we propose to encode the multi-oriented object with double horizontal rectangles (DHRec) to solve the discontinuity problem. Specifically, for an oriented object, we arrange the horizontal and vertical coordinates of its four vertices in left-right and top-down order, respectively. The first (resp. second) horizontal box is given by two diagonal points with smallest (resp. second) and third (resp. largest) coordinates in both horizontal and vertical dimensions. We then regress three factors given by area ratios between different regions, helping to guide the oriented object decoding from the predicted DHRec. Inherited from the uniqueness of horizontal rectangle representation, the proposed method is free of discontinuity issue, and can accurately detect objects of arbitrary orientation. Extensive experimental results show that the proposed method significantly improves the existing baseline representation, and outperforms state-of-the-art methods. The code is available at: https://github.com/lightbillow/DHRec.

Autism Spectrum Disorders (ASD) are characterized by deficits in social communication skills, such as response to joint attention (RJA). Robotic systems have been designed and applied to help children with ASD improve their RJA skills. One of the most important goals of robot-assisted intervention is helping children generalize social interaction skills to interact with other people. Thus predicting children's human-human interaction (HHI) performance based on their human-robot interaction (HRI) process is an important task. However, to the best of our knowledge, little research exists exploring this topic. The Early Social-Communication Scales (ESCS) test is a measurement of nonverbal social skills, including RJA, for young children. We conducted two longitudinal user studies with a robot-mediated RJA system in young children with ASD, followed by HHI sessions consisting of ESCS administration. In this paper, we present findings regarding how to predict participants' RJA performance in HHI based on their head pose patterns in HRI, under a semi-supervised machine learning framework. As a three-class classification problem, we achieved a micro-averaged accuracy of 73.5%, which indicates the potential effectiveness of the proposed method.