Publications
Most cited and implemented works with links to their production landings.
ORPR: An OR-Guided Pretrain-then-Reinforce Learning Model for Inventory Management
Abstract: As the pursuit of synergy between Artificial Intelligence (AI) and Operations Research (OR) gains momentum in handling complex inventory systems, a critical challenge persists: how to effectively reconcile AI's adaptive perception with OR's structural rigor. To bridge this gap, we propose a novel OR-Guided "Pretrain-then-Reinforce" framework. To provide structured guidance, we propose a simulation-augmented OR model that generates high-quality reference decisions, implicitly capturing complex business constraints and managerial preferences. Leveraging these OR-derived decisions as foundational training labels, we design a domain-informed deep learning foundation model to establish foundational decision-making capabilities, followed by a reinforcement learning (RL) fine-tuning stage. Uniquely, we position RL as a deep alignment mechanism that enables the AI agent to internalize the optimality principles of OR, while simultaneously leveraging exploration for general policy refinement and allowing expert guidance for scenario-specific adaptation (e.g., promotional events). Validated through extensive numerical experiments and a field deployment at JD.com augmented by a Difference-in-Differences (DiD) analysis, our model significantly outperforms incumbent industrial practices, delivering real-world gains of a 5.27-day reduction in turnover and a 2.29% increase in in-stock rates, alongside a 29.95% decrease in holding costs. Contrary to the prevailing trend of brute-force model scaling, our study demonstrates that a lightweight, domain-informed model can deliver state-of-the-art performance and robust transferability when guided by structured OR logic. This approach offers a scalable and cost-effective paradigm for intelligent supply chain management, highlighting the value of deeply aligning AI with OR.
Efficient Multi-Adapter LLM Serving via Cross-Model KV-Cache Reuse with Activated LoRA
Abstract: Modern large language model (LLM) systems increasingly rely on multi-turn pipelines that are composed of multiple task-specific adapters, yet existing serving frameworks remain inefficient, incurring substantial recomputation overhead when switching between adapters. We present the first LLM serving engine that supports cross-model prefix cache reuse between base and adapted models via Activated LoRA (aLoRA), enabling efficient and fine-grained adapter switching during inference. Our design extends the vLLM framework by introducing base-aligned block hashing and activation-aware masking within the model execution path, permitting cache reuse across models while preserving compatibility with existing serving engine optimizations. Integrated into a production-grade inference stack, this approach supports dynamic adapter activation without excessive key-value tensor recomputation. Evaluation across representative multi-turn, multi-adapter pipelines demonstrates up to 58x end-to-end latency reduction and over 100x time-to-first-token improvement relative to standard LoRA baselines, with benefits that scale with model size and sequence length and manifest across all stages of the request lifecycle. This work bridges parameter-efficient model adaptation with high-performance serving, providing the first complete realization of cross-model KV-cache reuse in modern LLM inference engines.
Application of machine learning to predict food processing level using Open Food Facts
Abstract: Ultra-processed foods are increasingly linked to health issues like obesity, cardiovascular disease, type 2 diabetes, and mental health disorders due to poor nutritional quality. This first-of-its-kind study at such a scale uses machine learning to classify food processing levels (NOVA) based on the Open Food Facts dataset of over 900,000 products. Models including LightGBM, Random Forest, and CatBoost were trained on nutrient concentration data. LightGBM performed best, achieving 80-85% accuracy across different nutrient panels and effectively distinguishing minimally from ultra-processed foods. Exploratory analysis revealed strong associations between higher NOVA classes and lower Nutri-Scores, indicating poorer nutritional quality. Products in NOVA 3 and 4 also had higher carbon footprints and lower Eco-Scores, suggesting greater environmental impact. Allergen analysis identified gluten and milk as common in ultra-processed items, posing risks to sensitive individuals. Categories like Cakes and Snacks were dominant in higher NOVA classes, which also had more additives, highlighting the role of ingredient modification. This study, leveraging the largest dataset of NOVA-labeled products, emphasizes the health, environmental, and allergenic implications of food processing and showcases machine learning's value in scalable classification. A user-friendly web tool is available for NOVA prediction using nutrient data: https://cosylab.iiitd.edu.in/foodlabel/.
Staggered Batch Scheduling: Co-optimizing Time-to-First-Token and Throughput for High-Efficiency LLM Inference
Abstract: The evolution of Large Language Model (LLM) serving towards complex, distributed architectures--specifically the P/D-separated, large-scale DP+EP paradigm--introduces distinct scheduling challenges. Unlike traditional deployments where schedulers can treat instances as black boxes, DP+EP architectures exhibit high internal synchronization costs. We identify that immediate request dispatching in such systems leads to severe in-engine queuing and parallelization bubbles, degrading Time-to-First-Token (TTFT). To address this, we propose Staggered Batch Scheduling (SBS), a mechanism that deliberately buffers requests to form optimal execution batches. This temporal decoupling eliminates internal queuing bubbles without compromising throughput. Furthermore, leveraging the scheduling window created by buffering, we introduce a Load-Aware Global Allocation strategy that balances computational load across DP units for both Prefill and Decode phases. Deployed on a production H800 cluster serving Deepseek-V3, our system reduces TTFT by 30%-40% and improves throughput by 15%-20% compared to state-of-the-art immediate scheduling baselines.
Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length
Abstract: Existing diffusion-based video generation methods are fundamentally constrained by sequential computation and long-horizon inconsistency, limiting their practical adoption in real-time, streaming audio-driven avatar synthesis. We present Live Avatar, an algorithm-system co-designed framework that enables efficient, high-fidelity, and infinite-length avatar generation using a 14-billion-parameter diffusion model. Our approach introduces Timestep-forcing Pipeline Parallelism (TPP), a distributed inference paradigm that pipelines denoising steps across multiple GPUs, effectively breaking the autoregressive bottleneck and ensuring stable, low-latency real-time streaming. To further enhance temporal consistency and mitigate identity drift and color artifacts, we propose the Rolling Sink Frame Mechanism (RSFM), which maintains sequence fidelity by dynamically recalibrating appearance using a cached reference image. Additionally, we leverage Self-Forcing Distribution Matching Distillation to facilitate causal, streamable adaptation of large-scale models without sacrificing visual quality. Live Avatar demonstrates state-of-the-art performance, reaching 20 FPS end-to-end generation on 5 H800 GPUs, and, to the best of our knowledge, is the first to achieve practical, real-time, high-fidelity avatar generation at this scale. Our work establishes a new paradigm for deploying advanced diffusion models in industrial long-form video synthesis applications.
PrivateXR: Defending Privacy Attacks in Extended Reality Through Explainable AI-Guided Differential Privacy
Abstract: The convergence of artificial AI and XR technologies (AI XR) promises innovative applications across many domains. However, the sensitive nature of data (e.g., eye-tracking) used in these systems raises significant privacy concerns, as adversaries can exploit these data and models to infer and leak personal information through membership inference attacks (MIA) and re-identification (RDA) with a high success rate. Researchers have proposed various techniques to mitigate such privacy attacks, including differential privacy (DP). However, AI XR datasets often contain numerous features, and applying DP uniformly can introduce unnecessary noise to less relevant features, degrade model accuracy, and increase inference time, limiting real-time XR deployment. Motivated by this, we propose a novel framework combining explainable AI (XAI) and DP-enabled privacy-preserving mechanisms to defend against privacy attacks. Specifically, we leverage post-hoc explanations to identify the most influential features in AI XR models and selectively apply DP to those features during inference. We evaluate our XAI-guided DP approach on three state-of-the-art AI XR models and three datasets: cybersickness, emotion, and activity classification. Our results show that the proposed method reduces MIA and RDA success rates by up to 43% and 39%, respectively, for cybersickness tasks while preserving model utility with up to 97% accuracy using Transformer models. Furthermore, it improves inference time by up to ~2x compared to traditional DP approaches. To demonstrate practicality, we deploy the XAI-guided DP AI XR models on an HTC VIVE Pro headset and develop a user interface (UI), namely PrivateXR, allowing users to adjust privacy levels (e.g., low, medium, high) while receiving real-time task predictions, protecting user privacy during XR gameplay.
BookReconciler: An Open-Source Tool for Metadata Enrichment and Work-Level Clustering
Abstract: We present BookReconciler, an open-source tool for enhancing and clustering book data. BookReconciler allows users to take spreadsheets with minimal metadata, such as book title and author, and automatically 1) add authoritative, persistent identifiers like ISBNs 2) and cluster related Expressions and Manifestations of the same Work, e.g., different translations or editions. This enhancement makes it easier to combine related collections and analyze books at scale. The tool is currently designed as an extension for OpenRefine -- a popular software application -- and connects to major bibliographic services including the Library of Congress, VIAF, OCLC, HathiTrust, Google Books, and Wikidata. Our approach prioritizes human judgment. Through an interactive interface, users can manually evaluate matches and define the contours of a Work (e.g., to include translations or not). We evaluate reconciliation performance on datasets of U.S. prize-winning books and contemporary world fiction. BookReconciler achieves near-perfect accuracy for U.S. works but lower performance for global texts, reflecting structural weaknesses in bibliographic infrastructures for non-English and global literature. Overall, BookReconciler supports the reuse of bibliographic data across domains and applications, contributing to ongoing work in digital libraries and digital humanities.
Prediction of Respiratory Syncytial Virus-Associated Hospitalizations Using Machine Learning Models Based on Environmental Data
Abstract: Respiratory syncytial virus (RSV) is a leading cause of hospitalization among young children, with outbreaks strongly influenced by environmental conditions. This study developed a machine learning framework to predict RSV-associated hospitalizations in the United States (U.S.) by integrating wastewater surveillance, meteorological, and air quality data. The dataset combined weekly hospitalization rates, wastewater RSV levels, daily meteorological measurements, and air pollutant concentrations. Classification models, including CART, Random Forest, and Boosting, were trained to predict weekly RSV-associated hospitalization rates classified as \textit{Low risk}, \textit{Alert}, and \textit{Epidemic} levels. The wastewater RSV level was identified as the strongest predictor, followed by meteorological and air quality variables such as temperature, ozone levels, and specific humidity. Notably, the analysis also revealed significantly higher RSV-associated hospitalization rates among Native Americans and Alaska Natives. Further research is needed to better understand the drivers of RSV disparity in these communities to improve prevention strategies. Furthermore, states at high altitudes, characterized by lower surface pressure, showed consistently higher RSV-associated hospitalization rates. These findings highlight the value of combining environmental and community surveillance data to forecast RSV outbreaks, enabling more timely public health interventions and resource allocation. In order to provide accessibility and practical use of the models, we have developed an interactive R Shiny dashboard (https://f6yxlu-eric-guo.shinyapps.io/rsv_app/), which allows users to explore RSV-associated hospitalization risk levels across different states, visualize the impact of key predictors, and interactively generate RSV outbreak forecasts.
Early Warning Index for Patient Deteriorations in Hospitals
Abstract: Hospitals lack automated systems to harness the growing volume of heterogeneous clinical and operational data to effectively forecast critical events. Early identification of patients at risk for deterioration is essential not only for patient care quality monitoring but also for physician care management. However, translating varied data streams into accurate and interpretable risk assessments poses significant challenges due to inconsistent data formats. We develop a multimodal machine learning framework, the Early Warning Index (EWI), to predict the aggregate risk of ICU admission, emergency response team dispatch, and mortality. Key to EWI's design is a human-in-the-loop process: clinicians help determine alert thresholds and interpret model outputs, which are enhanced by explainable outputs using Shapley Additive exPlanations (SHAP) to highlight clinical and operational factors (e.g., scheduled surgeries, ward census) driving each patient's risk. We deploy EWI in a hospital dashboard that stratifies patients into three risk tiers. Using a dataset of 18,633 unique patients at a large U.S. hospital, our approach automatically extracts features from both structured and unstructured electronic health record (EHR) data and achieves C-statistics of 0.796. It is currently used as a triage tool for proactively managing at-risk patients. The proposed approach saves physicians valuable time by automatically sorting patients of varying risk levels, allowing them to concentrate on patient care rather than sifting through complex EHR data. By further pinpointing specific risk drivers, the proposed model provides data-informed adjustments to caregiver scheduling and allocation of critical resources. As a result, clinicians and administrators can avert downstream complications, including costly procedures or high readmission rates and improve overall patient flow.
A data-physics hybrid generative model for patient-specific post-stroke motor rehabilitation using wearable sensor data
Abstract: Dynamic prediction of locomotor capacity after stroke is crucial for tailoring rehabilitation, yet current assessments provide only static impairment scores and do not indicate whether patients can safely perform specific tasks such as slope walking or stair climbing. Here, we develop a data-physics hybrid generative framework that reconstructs an individual stroke survivor's neuromuscular control from a single 20 m level-ground walking trial and predicts task-conditioned locomotion across rehabilitation scenarios. The system combines wearable-sensor kinematics, a proportional-derivative physics controller, a population Healthy Motion Atlas, and goal-conditioned deep reinforcement learning with behaviour cloning and generative adversarial imitation learning to generate physically plausible, patient-specific gait simulations for slopes and stairs. In 11 stroke survivors, the personalized controllers preserved idiosyncratic gait patterns while improving joint-angle and endpoint fidelity by 4.73% and 12.10%, respectively, and reducing training time to 25.56% relative to a physics-only baseline. In a multicentre pilot involving 21 inpatients, clinicians who used our locomotion predictions to guide task selection and difficulty obtained larger gains in Fugl-Meyer lower-extremity scores over 28 days of standard rehabilitation than control clinicians (mean change 6.0 versus 3.7 points). These findings indicate that our generative, task-predictive framework can augment clinical decision-making in post-stroke gait rehabilitation and provide a template for dynamically personalized motor recovery strategies.
Towards Practical Large-scale Dynamical Heterogeneous Graph Embedding: Cold-start Resilient Recommendation
Abstract: Deploying dynamic heterogeneous graph embeddings in production faces key challenges of scalability, data freshness, and cold-start. This paper introduces a practical, two-stage solution that balances deep graph representation with low-latency incremental updates. Our framework combines HetSGFormer, a scalable graph transformer for static learning, with Incremental Locally Linear Embedding (ILLE), a lightweight, CPU-based algorithm for real-time updates. HetSGFormer captures global structure with linear scalability, while ILLE provides rapid, targeted updates to incorporate new data, thus avoiding costly full retraining. This dual approach is cold-start resilient, leveraging the graph to create meaningful embeddings from sparse data. On billion-scale graphs, A/B tests show HetSGFormer achieved up to a 6.11% lift in Advertiser Value over previous methods, while the ILLE module added another 3.22% lift and improved embedding refresh timeliness by 83.2%. Our work provides a validated framework for deploying dynamic graph learning in production environments.
Seedance 1.5 pro: A Native Audio-Visual Joint Generation Foundation Model
Abstract: Recent strides in video generation have paved the way for unified audio-visual generation. In this work, we present Seedance 1.5 pro, a foundational model engineered specifically for native, joint audio-video generation. Leveraging a dual-branch Diffusion Transformer architecture, the model integrates a cross-modal joint module with a specialized multi-stage data pipeline, achieving exceptional audio-visual synchronization and superior generation quality. To ensure practical utility, we implement meticulous post-training optimizations, including Supervised Fine-Tuning (SFT) on high-quality datasets and Reinforcement Learning from Human Feedback (RLHF) with multi-dimensional reward models. Furthermore, we introduce an acceleration framework that boosts inference speed by over 10X. Seedance 1.5 pro distinguishes itself through precise multilingual and dialect lip-syncing, dynamic cinematic camera control, and enhanced narrative coherence, positioning it as a robust engine for professional-grade content creation. Seedance 1.5 pro is now accessible on Volcano Engine at https://console.volcengine.com/ark/region:ark+cn-beijing/experience/vision?type=GenVideo.
AI as a Teaching Partner: Early Lessons from Classroom Codesign with Secondary Teachers
Abstract: This report presents a comprehensive account of the Colleague AI Classroom pilot, a collaborative design (co-design) study that brought generative AI technology directly into real classrooms. In this study, AI functioned as a third agent, an active participant that mediated feedback, supported inquiry, and extended teachers' instructional reach while preserving human judgment and teacher authority. Over seven weeks in spring 2025, 21 in-service teachers from four Washington State public school districts and one independent school integrated four AI-powered features of the Colleague AI Classroom into their instruction: Teaching Aide, Assessment and AI Grading, AI Tutor, and Student Growth Insights. More than 600 students in grades 6-12 used the platform in class at the direction of their teachers, who designed and facilitated the AI activities. During the Classroom pilot, teachers were co-design partners: they planned activities, implemented them with students, and provided weekly reflections on AI's role in classroom settings. The teachers' feedback guided iterative improvements for Colleague AI. The research team captured rich data through surveys, planning and reflection forms, group meetings, one-on-one interviews, and platform usage logs to understand where AI adds instructional value and where it requires refinement.
Monitoring Deployed AI Systems in Health Care
Abstract: Post-deployment monitoring of artificial intelligence (AI) systems in health care is essential to ensure their safety, quality, and sustained benefit-and to support governance decisions about which systems to update, modify, or decommission. Motivated by these needs, we developed a framework for monitoring deployed AI systems grounded in the mandate to take specific actions when they fail to behave as intended. This framework, which is now actively used at Stanford Health Care, is organized around three complementary principles: system integrity, performance, and impact. System integrity monitoring focuses on maximizing system uptime, detecting runtime errors, and identifying when changes to the surrounding IT ecosystem have unintended effects. Performance monitoring focuses on maintaining accurate system behavior in the face of changing health care practices (and thus input data) over time. Impact monitoring assesses whether a deployed system continues to have value in the form of benefit to clinicians and patients. Drawing on examples of deployed AI systems at our academic medical center, we provide practical guidance for creating monitoring plans based on these principles that specify which metrics to measure, when those metrics should be reviewed, who is responsible for acting when metrics change, and what concrete follow-up actions should be taken-for both traditional and generative AI. We also discuss challenges to implementing this framework, including the effort and cost of monitoring for health systems with limited resources and the difficulty of incorporating data-driven monitoring practices into complex organizations where conflicting priorities and definitions of success often coexist. This framework offers a practical template and starting point for health systems seeking to ensure that AI deployments remain safe and effective over time.
The Adoption and Usage of AI Agents: Early Evidence from Perplexity
Abstract: This paper presents the first large-scale field study of the adoption, usage intensity, and use cases of general-purpose AI agents operating in open-world web environments. Our analysis centers on Comet, an AI-powered browser developed by Perplexity, and its integrated agent, Comet Assistant. Drawing on hundreds of millions of anonymized user interactions, we address three fundamental questions: Who is using AI agents? How intensively are they using them? And what are they using them for? Our findings reveal substantial heterogeneity in adoption and usage across user segments. Earlier adopters, users in countries with higher GDP per capita and educational attainment, and individuals working in digital or knowledge-intensive sectors -- such as digital technology, academia, finance, marketing, and entrepreneurship -- are more likely to adopt or actively use the agent. To systematically characterize the substance of agent usage, we introduce a hierarchical agentic taxonomy that organizes use cases across three levels: topic, subtopic, and task. The two largest topics, Productivity & Workflow and Learning & Research, account for 57% of all agentic queries, while the two largest subtopics, Courses and Shopping for Goods, make up 22%. The top 10 out of 90 tasks represent 55% of queries. Personal use constitutes 55% of queries, while professional and educational contexts comprise 30% and 16%, respectively. In the short term, use cases exhibit strong stickiness, but over time users tend to shift toward more cognitively oriented topics. The diffusion of increasingly capable AI agents carries important implications for researchers, businesses, policymakers, and educators, inviting new lines of inquiry into this rapidly emerging class of AI capabilities.
SnapStream: Efficient Long Sequence Decoding on Dataflow Accelerators
Abstract: The proliferation of 100B+ parameter Large Language Models (LLMs) with 100k+ context length support have resulted in increasing demands for on-chip memory to support large KV caches. Techniques such as StreamingLLM and SnapKV demonstrate how to control KV cache size while maintaining model accuracy. Yet, these techniques are not commonly used within industrial deployments using frameworks like vLLM or SGLang. The reason is twofold: on one hand, the static graphs and continuous batching methodology employed by these frameworks make it difficult to admit modifications to the standard multi-head attention algorithm, while on the other hand, the accuracy implications of such techniques on modern instruction-following and reasoning models are not well understood, obfuscating the need for implementing these techniques. In this paper, we explore these accuracy implications on Llama-3.1-8B-Instruct and DeepSeek-R1, and develop SnapStream, a KV cache compression method that can be deployed at scale. We demonstrate the efficacy of SnapStream in a 16-way tensor-parallel deployment of DeepSeek-671B on SambaNova SN40L accelerators running at 128k context length and up to 1832 tokens per second in a real production setting. SnapStream enables $4\times$ improved on-chip memory usage and introduces minimal accuracy degradation on LongBench-v2, AIME24 and LiveCodeBench. To the best of our knowledge, this is the first implementation of sparse KV attention techniques deployed in a production inference system with static graphs and continuous batching.
A Field Guide to Deploying AI Agents in Clinical Practice
Abstract: Large language models (LLMs) integrated into agent-driven workflows hold immense promise for healthcare, yet a significant gap exists between their potential and practical implementation within clinical settings. To address this, we present a practitioner-oriented field manual for deploying generative agents that use electronic health record (EHR) data. This guide is informed by our experience deploying the "irAE-Agent", an automated system to detect immune-related adverse events from clinical notes at Mass General Brigham, and by structured interviews with 21 clinicians, engineers, and informatics leaders involved in the project. Our analysis reveals a critical misalignment in clinical AI development: less than 20% of our effort was dedicated to prompt engineering and model development, while over 80% was consumed by the sociotechnical work of implementation. We distill this effort into five "heavy lifts": data integration, model validation, ensuring economic value, managing system drift, and governance. By providing actionable solutions for each of these challenges, this field manual shifts the focus from algorithmic development to the essential infrastructure and implementation work required to bridge the "valley of death" and successfully translate generative AI from pilot projects into routine clinical care.
SmartAlert: Implementing Machine Learning-Driven Clinical Decision Support for Inpatient Lab Utilization Reduction
Abstract: Repetitive laboratory testing unlikely to yield clinically useful information is a common practice that burdens patients and increases healthcare costs. Education and feedback interventions have limited success, while general test ordering restrictions and electronic alerts impede appropriate clinical care. We introduce and evaluate SmartAlert, a machine learning (ML)-driven clinical decision support (CDS) system integrated into the electronic health record that predicts stable laboratory results to reduce unnecessary repeat testing. This case study describes the implementation process, challenges, and lessons learned from deploying SmartAlert targeting complete blood count (CBC) utilization in a randomized controlled pilot across 9270 admissions in eight acute care units across two hospitals between August 15, 2024, and March 15, 2025. Results show significant decrease in number of CBC results within 52 hours of SmartAlert display (1.54 vs 1.82, p <0.01) without adverse effect on secondary safety outcomes, representing a 15% relative reduction in repetitive testing. Implementation lessons learned include interpretation of probabilistic model predictions in clinical contexts, stakeholder engagement to define acceptable model behavior, governance processes for deploying a complex model in a clinical environment, user interface design considerations, alignment with clinical operational priorities, and the value of qualitative feedback from end users. In conclusion, a machine learning-driven CDS system backed by a deliberate implementation and governance process can provide precision guidance on inpatient laboratory testing to safely reduce unnecessary repetitive testing.
Breaking Determinism: Stochastic Modeling for Reliable Off-Policy Evaluation in Ad Auctions
Abstract: Online A/B testing, the gold standard for evaluating new advertising policies, consumes substantial engineering resources and risks significant revenue loss from deploying underperforming variations. This motivates the use of Off-Policy Evaluation (OPE) for rapid, offline assessment. However, applying OPE to ad auctions is fundamentally more challenging than in domains like recommender systems, where stochastic policies are common. In online ad auctions, it is common for the highest-bidding ad to win the impression, resulting in a deterministic, winner-takes-all setting. This results in zero probability of exposure for non-winning ads, rendering standard OPE estimators inapplicable. We introduce the first principled framework for OPE in deterministic auctions by repurposing the bid landscape model to approximate the propensity score. This model allows us to derive robust approximate propensity scores, enabling the use of stable estimators like Self-Normalized Inverse Propensity Scoring (SNIPS) for counterfactual evaluation. We validate our approach on the AuctionNet simulation benchmark and against 2-weeks online A/B test from a large-scale industrial platform. Our method shows remarkable alignment with online results, achieving a 92\% Mean Directional Accuracy (MDA) in CTR prediction, significantly outperforming the parametric baseline. MDA is the most critical metric for guiding deployment decisions, as it reflects the ability to correctly predict whether a new model will improve or harm performance. This work contributes the first practical and validated framework for reliable OPE in deterministic auction environments, offering an efficient alternative to costly and risky online experiments.
Learning to Comparison-Shop
Abstract: In online marketplaces like Airbnb, users frequently engage in comparison shopping before making purchase decisions. Despite the prevalence of this behavior, a significant disconnect persists between mainstream e-commerce search engines and users' comparison needs. Traditional ranking models often evaluate items in isolation, disregarding the context in which users compare multiple items on a search results page. While recent advances in deep learning have sought to improve ranking accuracy, diversity, and fairness by encoding listwise context, the challenge of aligning search rankings with user comparison shopping behavior remains inadequately addressed. In this paper, we propose a novel ranking architecture - Learning-to-Comparison-Shop (LTCS) System - that explicitly models and learns users' comparison shopping behaviors. Through extensive offline and online experiments, we demonstrate that our approach yields statistically significant gains in key business metrics - improving NDCG by 1.7% and boosting booking conversion rate by 0.6% in A/B testing - while also enhancing user experience. We also compare our model against state-of-the-art approaches and demonstrate that LTCS significantly outperforms them.
AR-Med: Automated Relevance Enhancement in Medical Search via LLM-Driven Information Augmentation
Abstract: Accurate and reliable search on online healthcare platforms is critical for user safety and service efficacy. Traditional methods, however, often fail to comprehend complex and nuanced user queries, limiting their effectiveness. Large language models (LLMs) present a promising solution, offering powerful semantic understanding to bridge this gap. Despite their potential, deploying LLMs in this high-stakes domain is fraught with challenges, including factual hallucinations, specialized knowledge gaps, and high operational costs. To overcome these barriers, we introduce \textbf{AR-Med}, a novel framework for \textbf{A}utomated \textbf{R}elevance assessment for \textbf{Med}ical search that has been successfully deployed at scale on the Online Medical Delivery Platforms. AR-Med grounds LLM reasoning in verified medical knowledge through a retrieval-augmented approach, ensuring high accuracy and reliability. To enable efficient online service, we design a practical knowledge distillation scheme that compresses large teacher models into compact yet powerful student models. We also introduce LocalQSMed, a multi-expert annotated benchmark developed to guide model iteration and ensure strong alignment between offline and online performance. Extensive experiments show AR-Med achieves an offline accuracy of over 93\%, a 24\% absolute improvement over the original online system, and delivers significant gains in online relevance and user satisfaction. Our work presents a practical and scalable blueprint for developing trustworthy, LLM-powered systems in real-world healthcare applications.
Crossing the Sim2Real Gap Between Simulation and Ground Testing to Space Deployment of Autonomous Free-flyer Control
Abstract: Reinforcement learning (RL) offers transformative potential for robotic control in space. We present the first on-orbit demonstration of RL-based autonomous control of a free-flying robot, the NASA Astrobee, aboard the International Space Station (ISS). Using NVIDIA's Omniverse physics simulator and curriculum learning, we trained a deep neural network to replace Astrobee's standard attitude and translation control, enabling it to navigate in microgravity. Our results validate a novel training pipeline that bridges the simulation-to-reality (Sim2Real) gap, utilizing a GPU-accelerated, scientific-grade simulation environment for efficient Monte Carlo RL training. This successful deployment demonstrates the feasibility of training RL policies terrestrially and transferring them to space-based applications. This paves the way for future work in In-Space Servicing, Assembly, and Manufacturing (ISAM), enabling rapid on-orbit adaptation to dynamic mission requirements.
Cross-embodied Co-design for Dexterous Hands
Abstract: Dexterous manipulation is limited by both control and design, without consensus as to what makes manipulators best for performing dexterous tasks. This raises a fundamental challenge: how should we design and control robot manipulators that are optimized for dexterity? We present a co-design framework that learns task-specific hand morphology and complementary dexterous control policies. The framework supports 1) an expansive morphology search space including joint, finger, and palm generation, 2) scalable evaluation across the wide design space via morphology-conditioned cross-embodied control, and 3) real-world fabrication with accessible components. We evaluate the approach across multiple dexterous tasks, including in-hand rotation with simulation and real deployment. Our framework enables an end-to-end pipeline that can design, train, fabricate, and deploy a new robotic hand in under 24 hours. The full framework will be open-sourced and available on our website.
Autonomous Planning In-space Assembly Reinforcement-learning free-flYer (APIARY) International Space Station Astrobee Testing
Abstract: The US Naval Research Laboratory's (NRL's) Autonomous Planning In-space Assembly Reinforcement-learning free-flYer (APIARY) experiment pioneers the use of reinforcement learning (RL) for control of free-flying robots in the zero-gravity (zero-G) environment of space. On Tuesday, May 27th 2025 the APIARY team conducted the first ever, to our knowledge, RL control of a free-flyer in space using the NASA Astrobee robot on-board the International Space Station (ISS). A robust 6-degrees of freedom (DOF) control policy was trained using an actor-critic Proximal Policy Optimization (PPO) network within the NVIDIA Isaac Lab simulation environment, randomizing over goal poses and mass distributions to enhance robustness. This paper details the simulation testing, ground testing, and flight validation of this experiment. This on-orbit demonstration validates the transformative potential of RL for improving robotic autonomy, enabling rapid development and deployment (in minutes to hours) of tailored behaviors for space exploration, logistics, and real-time mission needs.
An AI Implementation Science Study to Improve Trustworthy Data in a Large Healthcare System
Abstract: The rapid growth of Artificial Intelligence (AI) in healthcare has sparked interest in Trustworthy AI and AI Implementation Science, both of which are essential for accelerating clinical adoption. However, strict regulations, gaps between research and clinical settings, and challenges in evaluating AI systems continue to hinder real-world implementation. This study presents an AI implementation case study within Shriners Childrens (SC), a large multisite pediatric system, showcasing the modernization of SCs Research Data Warehouse (RDW) to OMOP CDM v5.4 within a secure Microsoft Fabric environment. We introduce a Python-based data quality assessment tool compatible with SCs infrastructure, extending OHDsi's R/Java-based Data Quality Dashboard (DQD) and integrating Trustworthy AI principles using the METRIC framework. This extension enhances data quality evaluation by addressing informative missingness, redundancy, timeliness, and distributional consistency. We also compare systematic and case-specific AI implementation strategies for Craniofacial Microsomia (CFM) using the FHIR standard. Our contributions include a real-world evaluation of AI implementations, integration of Trustworthy AI principles into data quality assessment, and insights into hybrid implementation strategies that blend systematic infrastructure with use-case-driven approaches to advance AI in healthcare.
A Large Scale Heterogeneous Treatment Effect Estimation Framework and Its Applications of Users' Journey at Snap
Abstract: Heterogeneous Treatment Effect (HTE) and Conditional Average Treatment Effect (CATE) models relax the assumption that treatment effects are the same for every user. We present a large scale industrial framework for estimating HTE using experimental data from hundreds of millions of Snapchat users. By combining results across many experiments, the framework uncovers latent user characteristics that were previously unmeasurable and produces stable treatment effect estimates at scale. We describe the core components that enabled this system, including experiment selection, base learner design, and incremental training. We also highlight two applications: user influenceability to ads and user sensitivity to ads. An online A/B test using influenceability scores for targeting showed an improvement on key business metrics that is more than six times larger than what is typically considered significant.
Dynamic Content Moderation in Livestreams: Combining Supervised Classification with MLLM-Boosted Similarity Matching
Abstract: Content moderation remains a critical yet challenging task for large-scale user-generated video platforms, especially in livestreaming environments where moderation must be timely, multimodal, and robust to evolving forms of unwanted content. We present a hybrid moderation framework deployed at production scale that combines supervised classification for known violations with reference-based similarity matching for novel or subtle cases. This hybrid design enables robust detection of both explicit violations and novel edge cases that evade traditional classifiers. Multimodal inputs (text, audio, visual) are processed through both pipelines, with a multimodal large language model (MLLM) distilling knowledge into each to boost accuracy while keeping inference lightweight. In production, the classification pipeline achieves 67% recall at 80% precision, and the similarity pipeline achieves 76% recall at 80% precision. Large-scale A/B tests show a 6-8% reduction in user views of unwanted livestreams}. These results demonstrate a scalable and adaptable approach to multimodal content governance, capable of addressing both explicit violations and emerging adversarial behaviors.
AskNearby: An LLM-Based Application for Neighborhood Information Retrieval and Personalized Cognitive-Map Recommendations
Abstract: The "15-minute city" envisions neighborhoods where residents can meet daily needs via a short walk or bike ride. Realizing this vision requires not only physical proximity but also efficient and reliable access to information about nearby places, services, and events. Existing location-based systems, however, focus mainly on city-level tasks and neglect the spatial, temporal, and cognitive factors that shape localized decision-making. We conceptualize this gap as the Local Life Information Accessibility (LLIA) problem and introduce AskNearby, an AI-driven community application that unifies retrieval and recommendation within the 15-minute life circle. AskNearby integrates (i) a three-layer Retrieval-Augmented Generation (RAG) pipeline that synergizes graph-based, semantic-vector, and geographic retrieval with (ii) a cognitive-map model that encodes each user's neighborhood familiarity and preferences. Experiments on real-world community datasets demonstrate that AskNearby significantly outperforms LLM-based and map-based baselines in retrieval accuracy and recommendation quality, achieving robust performance in spatiotemporal grounding and cognitive-aware ranking. Real-world deployments further validate its effectiveness. By addressing the LLIA challenge, AskNearby empowers residents to more effectively discover local resources, plan daily activities, and engage in community life.
IACT: A Self-Organizing Recursive Model for General AI Agents: A Technical White Paper on the Architecture Behind kragent.ai
Abstract: This technical white paper introduces the Interactive Agents Call Tree (IACT), a computational model designed to address the limitations of static, hard-coded agent workflows. Unlike traditional systems that require pre-defined graphs or specialized programming, IACT operates as a general-purpose autonomous system driven purely by user dialogue. Given a high-level objective, the system autonomously grows a dynamic, recursive agent topology incrementally tailored to the problem's structure. This allows it to scale its organizational complexity to match open-ended tasks. To mitigate the error propagation inherent in unidirectional function calls, IACT introduces interactional redundancy by replacing rigid invocations with bidirectional, stateful dialogues. This mechanism enables runtime error correction and ambiguity resolution. We describe the architecture, design principles, and practical lessons behind the production deployment of this model in the kragent.ai system, presenting qualitative evidence from real-world workflows rather than exhaustive benchmark results.
ADORE: Autonomous Domain-Oriented Relevance Engine for E-commerce
Abstract: 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.
LORE: A Large Generative Model for Search Relevance
Abstract: Achievement. We introduce LORE, a systematic framework for Large Generative Model-based relevance in e-commerce search. Deployed and iterated over three years, LORE achieves a cumulative +27\% improvement in online GoodRate metrics. This report shares the valuable experience gained throughout its development lifecycle, spanning data, features, training, evaluation, and deployment. Insight. While existing works apply Chain-of-Thought (CoT) to enhance relevance, they often hit a performance ceiling. We argue this stems from treating relevance as a monolithic task, lacking principled deconstruction. Our key insight is that relevance comprises distinct capabilities: knowledge and reasoning, multi-modal matching, and rule adherence. We contend that a qualitative-driven decomposition is essential for breaking through current performance bottlenecks. Contributions. LORE provides a complete blueprint for the LLM relevance lifecycle. Key contributions include: (1) A two-stage training paradigm combining progressive CoT synthesis via SFT with human preference alignment via RL. (2) A comprehensive benchmark, RAIR, designed to evaluate these core capabilities. (3) A query frequency-stratified deployment strategy that efficiently transfers offline LLM capabilities to the online system. LORE serves as both a practical solution and a methodological reference for other vertical domains.
First On-Orbit Demonstration of a Geospatial Foundation Model
Abstract: Geospatial foundation models (GeoFMs) promise broad generalisation capacity for Earth observation (EO) tasks, particularly under data-limited conditions. However, their large size poses a barrier to deployment on resource-constrained space hardware. To address this, we present compact variants of a Vision Transformer (ViT)-based GeoFM that preserve downstream task performance while enabling onboard execution. Evaluation across five downstream tasks and validation in two representative flight environments show that model compression and domain adaptation are critical to reducing size and resource demands while maintaining high performance under operational conditions. We further demonstrate reliable on-orbit inference with the IMAGIN-e payload aboard the International Space Station. These results establish a pathway from large GeoFMs to flight-ready, resource-efficient deployments, expanding the feasibility of onboard AI for EO missions.
Can-SAVE: Deploying Low-Cost and Population-Scale Cancer Screening via Survival Analysis Variables and EHR
arXiv:2309.15039v4 Announce Type: replace Abstract: Conventional medical cancer screening methods are costly, labor-intensive, and extremely difficult to scale. Although AI can improve cancer detection, most systems rely on complex or specialized medical data, making them impractical for large-scale screening. We introduce Can-SAVE, a lightweight AI system that ranks population-wide cancer risks solely based on medical history events. By integrating survival model outputs into a gradient-boosting framework, our approach detects subtle, long-term patient risk patterns - often well before clinical symptoms manifest. Can-SAVE was rigorously evaluated on a real-world dataset of 2.5 million adults spanning five Russian regions, marking the study as one of the largest and most comprehensive deployments of AI-driven cancer risk assessment. In a retrospective oncologist-supervised study over 1.9M patients, Can-SAVE achieves a 4-10x higher detection rate at identical screening volumes and an Average Precision (AP) of 0.228 vs. 0.193 for the best baseline (LoRA-tuned Qwen3-Embeddings via DeepSeek-R1 summarization). In a year-long prospective pilot (426K patients), our method almost doubled the cancer detection rate (+91%) and increased population coverage by 36% over the national screening protocol. The system demonstrates practical scalability: a city-wide population of 1 million patients can be processed in under three hours using standard hardware, enabling seamless clinical integration. This work proves that Can-SAVE achieves nationally significant cancer detection improvements while adhering to real-world public healthcare constraints, offering immediate clinical utility and a replicable framework for population-wide screening. Code for training and feature engineering is available at https://github.com/sb-ai-lab/Can-SAVE.
How Similar Are Grokipedia and Wikipedia? A Multi-Dimensional Textual and Structural Comparison
arXiv:2510.26899v3 Announce Type: replace-cross Abstract: The launch of Grokipedia - an AI-generated encyclopedia developed by Elon Musk's xAI - was presented as a response to perceived ideological and structural biases in Wikipedia, aiming to produce "truthful" entries using the Grok large language model. Yet whether an AI-driven alternative can escape the biases and limitations of human-edited platforms remains unclear. This study conducts a large-scale computational comparison of more than 17,000 matched article pairs from the 20,000 most-edited English Wikipedia pages. Using metrics spanning lexical richness, readability, reference density, structural features, and semantic similarity, we assess how closely the two platforms align in form and substance. We find that Grokipedia articles are substantially longer and contain significantly fewer references per word. Moreover, Grokipedia's content divides into two distinct groups: one that remains semantically and stylistically aligned with Wikipedia, and another that diverges sharply. Among the dissimilar articles, we observe a systematic rightward shift in the political bias of cited news sources, concentrated primarily in entries related to politics, history, and religion. These findings suggest that AI-generated encyclopedic content diverges from established editorial norms-favouring narrative expansion over citation-based verification. The implications highlight emerging tensions around transparency, provenance, and the governance of knowledge in an era of automated text generation.
Optimizing Generative Ranking Relevance via Reinforcement Learning in Xiaohongshu Search
Abstract: Ranking relevance is a fundamental task in search engines, aiming to identify the items most relevant to a given user query. Traditional relevance models typically produce scalar scores or directly predict relevance labels, limiting both interpretability and the modeling of complex relevance signals. Inspired by recent advances in Chain-of-Thought (CoT) reasoning for complex tasks, we investigate whether explicit reasoning can enhance both interpretability and performance in relevance modeling. However, existing reasoning-based Generative Relevance Models (GRMs) primarily rely on supervised fine-tuning on large amounts of human-annotated or synthetic CoT data, which often leads to limited generalization. Moreover, domain-agnostic, free-form reasoning tends to be overly generic and insufficiently grounded, limiting its potential to handle the diverse and ambiguous cases prevalent in open-domain search. In this work, we formulate relevance modeling in Xiaohongshu search as a reasoning task and introduce a Reinforcement Learning (RL)-based training framework to enhance the grounded reasoning capabilities of GRMs. Specifically, we incorporate practical business-specific relevance criteria into the multi-step reasoning prompt design and propose Stepwise Advantage Masking (SAM), a lightweight process-supervision strategy which facilitates effective learning of these criteria through improved credit assignment. To enable industrial deployment, we further distill the large-scale RL-tuned model to a lightweight version suitable for real-world search systems. Extensive experiments on industrial datasets, along with online A/B tests, demonstrate the effectiveness of our approach.
Enhancing Talent Search Ranking with Role-Aware Expert Mixtures and LLM-based Fine-Grained Job Descriptions
Abstract: Talent search is a cornerstone of modern recruitment systems, yet existing approaches often struggle to capture nuanced job-specific preferences, model recruiter behavior at a fine-grained level, and mitigate noise from subjective human judgments. We present a novel framework that enhances talent search effectiveness and delivers substantial business value through two key innovations: (i) leveraging LLMs to extract fine-grained recruitment signals from job descriptions and historical hiring data, and (ii) employing a role-aware multi-gate MoE network to capture behavioral differences across recruiter roles. To further reduce noise, we introduce a multi-task learning module that jointly optimizes click-through rate (CTR), conversion rate (CVR), and resume matching relevance. Experiments on real-world recruitment data and online A/B testing show relative AUC gains of 1.70% (CTR) and 5.97% (CVR), and a 17.29% lift in click-through conversion rate. These improvements reduce dependence on external sourcing channels, enabling an estimated annual cost saving of millions of CNY.
Multilingual Conversational AI for Financial Assistance: Bridging Language Barriers in Indian FinTech
Abstract: India's linguistic diversity presents both opportunities and challenges for fintech platforms. While the country has 31 major languages and over 100 minor ones, only 10\% of the population understands English, creating barriers to financial inclusion. We present a multilingual conversational AI system for a financial assistance use case that supports code-mixed languages like Hinglish, enabling natural interactions for India's diverse user base. Our system employs a multi-agent architecture with language classification, function management, and multilingual response generation. Through comparative analysis of multiple language models and real-world deployment, we demonstrate significant improvements in user engagement while maintaining low latency overhead (4-8\%). This work contributes to bridging the language gap in digital financial services for emerging markets.
Heterogeneous Multi-treatment Uplift Modeling for Trade-off Optimization in Short-Video Recommendation
arXiv:2511.18997v2 Announce Type: replace Abstract: The rapid proliferation of short videos on social media platforms presents unique challenges and opportunities for recommendation systems. Users exhibit diverse preferences, and the responses resulting from different strategies often conflict with one another, potentially exhibiting inverse correlations between metrics such as watch time and video view counts. Existing uplift models face limitations in handling the heterogeneous multi-treatment scenarios of short-video recommendations, often failing to effectively capture both the synergistic and individual causal effects of different strategies. Furthermore, traditional fixed-weight approaches for balancing these responses lack personalization and can result in biased decision-making. To address these issues, we propose a novel Heterogeneous Multi-treatment Uplift Modeling (HMUM) framework for trade-off optimization in short-video recommendations. HMUM comprises an Offline Hybrid Uplift Modeling (HUM) module, which captures the synergistic and individual effects of multiple strategies, and an Online Dynamic Decision-Making (DDM) module, which estimates the value weights of different user responses in real-time for personalized decision-making. Evaluated on two public datasets, an industrial dataset, and through online A/B experiments on the Kuaishou platform, our model demonstrated superior offline performance and significant improvements in key metrics. It is now fully deployed on the platform, benefiting hundreds of millions of users.
Mofasa: A Step Change in Metal-Organic Framework Generation
Abstract: Mofasa is an all-atom latent diffusion model with state-of-the-art performance for generating Metal-Organic Frameworks (MOFs). These are highly porous crystalline materials used to harvest water from desert air, capture carbon dioxide, store toxic gases and catalyse chemical reactions. In recognition of their value, the development of MOFs recently received a Nobel Prize in Chemistry. In many ways, MOFs are well-suited for exploiting generative models in chemistry: they are rationally-designable materials with a large combinatorial design space and strong structure-property couplings. And yet, to date, a high performance generative model has been lacking. To fill this gap, we introduce Mofasa, a general-purpose latent diffusion model that jointly samples positions, atom-types and lattice vectors for systems as large as 500 atoms. Mofasa avoids handcrafted assembly algorithms common in the literature, unlocking the simultaneous discovery of metal nodes, linkers and topologies. To help the scientific community build on our work, we release MofasaDB, an annotated library of hundreds of thousands of sampled MOF structures, along with a user-friendly web interface for search and discovery: https://mofux.ai/ .
Orion: A Unified Visual Agent for Multimodal Perception, Advanced Visual Reasoning and Execution
We introduce Orion, a visual agent that integrates vision-based reasoning with tool-augmented execution to achieve powerful, precise, multi-step visual intelligence across images, video, and documents. Unlike traditional vision-language models that generate descriptive outputs, Orion orchestrates a suite of specialized computer vision tools, including object detection, keypoint localization, panoptic segmentation, Optical Character Recognition (OCR), and geometric analysis, to execute complex multi-step visual workflows. The system achieves competitive performance across MMMU, MMBench, DocVQA, and MMLongBench while extending monolithic VLM capabilities to production-grade visual intelligence. Through its agentic, tool-augmented approach, Orion enables autonomous visual reasoning that bridges neural perception with symbolic execution, marking the transition from passive visual understanding to active, tool-driven visual intelligence. Try Orion for free at: https://chat.vlm.run Learn more at: https://www.vlm.run/orion
AIF: Asynchronous Inference Framework for Cost-Effective Pre-Ranking
In industrial recommendation systems, pre-ranking models based on deep neural networks (DNNs) commonly adopt a sequential execution framework: feature fetching and model forward computation are triggered only after receiving candidates from the upstream retrieval stage. This design introduces inherent bottlenecks, including redundant computations of identical users/items and increased latency due to strictly sequential operations, which jointly constrain the model's capacity and system efficiency. To address these limitations, we propose the Asynchronous Inference Framework (AIF), a cost-effective computational architecture that decouples interaction-independent components, those operating within a single user or item, from real-time prediction. AIF reorganizes the model inference process by performing user-side computations in parallel with the retrieval stage and conducting item-side computations in a nearline manner. This means that interaction-independent components are calculated just once and completed before the real-time prediction phase of the pre-ranking stage. As a result, AIF enhances computational efficiency and reduces latency, freeing up resources to significantly improve the feature set and model architecture of interaction-independent components. Moreover, we delve into model design within the AIF framework, employing approximated methods for interaction-dependent components in online real-time predictions. By co-designing both the framework and the model, our solution achieves notable performance gains without significantly increasing computational and latency costs. This has enabled the successful deployment of AIF in the Taobao display advertising system.
Enhancing Multi-Camera Gymnast Tracking Through Domain Knowledge Integration
We present a robust multi-camera gymnast tracking, which has been applied at international gymnastics championships for gymnastics judging. Despite considerable progress in multi-camera tracking algorithms, tracking gymnasts presents unique challenges: (i) due to space restrictions, only a limited number of cameras can be installed in the gymnastics stadium; and (ii) due to variations in lighting, background, uniforms, and occlusions, multi-camera gymnast detection may fail in certain views and only provide valid detections from two opposing views. These factors complicate the accurate determination of a gymnast's 3D trajectory using conventional multi-camera triangulation. To alleviate this issue, we incorporate gymnastics domain knowledge into our tracking solution. Given that a gymnast's 3D center typically lies within a predefined vertical plane during \revised{much of their} performance, we can apply a ray-plane intersection to generate coplanar 3D trajectory candidates for opposing-view detections. More specifically, we propose a novel cascaded data association (DA) paradigm that employs triangulation to generate 3D trajectory candidates when cross-view detections are sufficient, and resort to the ray-plane intersection when they are insufficient. Consequently, coplanar candidates are used to compensate for uncertain trajectories, thereby minimizing tracking failures. The robustness of our method is validated through extensive experimentation, demonstrating its superiority over existing methods in challenging scenarios. Furthermore, our gymnastics judging system, equipped with this tracking method, has been successfully applied to recent Gymnastics World Championships, earning significant recognition from the International Gymnastics Federation.
Comparative Security Performance of Workday Cloud ERP Across Key Dimensions
Workday is a cloud-based Enterprise Resource Planning-ERP system that brings HR, Finance, Supply Chain functions , Prism Analytics and Extend custom built in application together under an integrated software as a service SaaS environment. As every organization that undergoes digital transformation, the importance of securing sensitive enterprise data in cloud ERP systems has always been more challeging. To analyze Workday's security architecture, we present a Security analysis in both CIA Triad Enhanced Framework and Zero Trust Security Architecture. The study examines five key dimensions confidentiality, integrity, availability, authentication, and compliance with weighted sub metric analysis and qualitative document review. The results show Workday delivers a composite score of 0.86 with an overall score that closely matches international standards of best practices like GDPR, HIPAA, SOC 2, etc. The platform uses encryption protocols, granular access controls, network safeguards, and continuous verification mechanisms to enable least-privilege access and adaptive defense. Security groups and business process access rules provide scalable governance across very large organizational structures.Workday's layered security to tackle everyday cloud security weaknesses. The work concludes that Workday's architecture demonstrates the best practices for secure, scalable, and compliant ERP application-oriented deployment, which can make this a standard for enterprise cloud security management. These insights provide important guidance for organizations that wish to bolster their cloud ERP defenses and stay ahead of changing regulatory expectations.
CroPS: Improving Dense Retrieval with Cross-Perspective Positive Samples in Short-Video Search
Dense retrieval has become a foundational paradigm in modern search systems, especially on short-video platforms. However, most industrial systems adopt a self-reinforcing training pipeline that relies on historically exposed user interactions for supervision. This paradigm inevitably leads to a filter bubble effect, where potentially relevant but previously unseen content is excluded from the training signal, biasing the model toward narrow and conservative retrieval. In this paper, we present CroPS (Cross-Perspective Positive Samples), a novel retrieval data engine designed to alleviate this problem by introducing diverse and semantically meaningful positive examples from multiple perspectives. CroPS enhances training with positive signals derived from user query reformulation behavior (query-level), engagement data in recommendation streams (system-level), and world knowledge synthesized by large language models (knowledge-level). To effectively utilize these heterogeneous signals, we introduce a Hierarchical Label Assignment (HLA) strategy and a corresponding H-InfoNCE loss that together enable fine-grained, relevance-aware optimization. Extensive experiments conducted on Kuaishou Search, a large-scale commercial short-video search platform, demonstrate that CroPS significantly outperforms strong baselines both offline and in live A/B tests, achieving superior retrieval performance and reducing query reformulation rates. CroPS is now fully deployed in Kuaishou Search, serving hundreds of millions of users daily.
Driving with Regulation: Trustworthy and Interpretable Decision-Making for Autonomous Driving with Retrieval-Augmented Reasoning
Understanding and adhering to traffic regulations is essential for autonomous vehicles to ensure safety and trustworthiness. However, traffic regulations are complex, context-dependent, and differ between regions, posing a major challenge to conventional rule-based decision-making approaches. We present an interpretable, regulation-aware decision-making framework, DriveReg, which enables autonomous vehicles to understand and adhere to region-specific traffic laws and safety guidelines. The framework integrates a Retrieval-Augmented Generation (RAG)-based Traffic Regulation Retrieval Agent, which retrieves relevant rules from regulatory documents based on the current situation, and a Large Language Model (LLM)-powered Reasoning Agent that evaluates actions for legal compliance and safety. Our design emphasizes interpretability to enhance transparency and trustworthiness. To support systematic evaluation, we introduce the DriveReg Scenarios Dataset, a comprehensive dataset of driving scenarios across Boston, Singapore, and Los Angeles, with both hypothesized text-based cases and real-world driving data, constructed and annotated to evaluate models' capacity for regulation understanding and reasoning. We validate our framework on the DriveReg Scenarios Dataset and real-world deployment, demonstrating strong performance and robustness across diverse environments.
RescueLens: LLM-Powered Triage and Action on Volunteer Feedback for Food Rescue
Food rescue organizations simultaneously tackle food insecurity and waste by working with volunteers to redistribute food from donors who have excess to recipients who need it. Volunteer feedback allows food rescue organizations to identify issues early and ensure volunteer satisfaction. However, food rescue organizations monitor feedback manually, which can be cumbersome and labor-intensive, making it difficult to prioritize which issues are most important. In this work, we investigate how large language models (LLMs) assist food rescue organizers in understanding and taking action based on volunteer experiences. We work with 412 Food Rescue, a large food rescue organization based in Pittsburgh, Pennsylvania, to design RescueLens, an LLM-powered tool that automatically categorizes volunteer feedback, suggests donors and recipients to follow up with, and updates volunteer directions based on feedback. We evaluate the performance of RescueLens on an annotated dataset, and show that it can recover 96% of volunteer issues at 71% precision. Moreover, by ranking donors and recipients according to their rates of volunteer issues, RescueLens allows organizers to focus on 0.5% of donors responsible for more than 30% of volunteer issues. RescueLens is now deployed at 412 Food Rescue and through semi-structured interviews with organizers, we find that RescueLens streamlines the feedback process so organizers better allocate their time.
Cluster-based Adaptive Retrieval: Dynamic Context Selection for RAG Applications
Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by pulling in external material, document, code, manuals, from vast and ever-growing corpora, to effectively answer user queries. The effectiveness of RAG depends significantly on aligning the number of retrieved documents with query characteristics: narrowly focused queries typically require fewer, highly relevant documents, whereas broader or ambiguous queries benefit from retrieving more extensive supporting information. However, the common static top-k retrieval approach fails to adapt to this variability, resulting in either insufficient context from too few documents or redundant information from too many. Motivated by these challenges, we introduce Cluster-based Adaptive Retrieval (CAR), an algorithm that dynamically determines the optimal number of documents by analyzing the clustering patterns of ordered query-document similarity distances. CAR detects the transition point within similarity distances, where tightly clustered, highly relevant documents shift toward less pertinent candidates, establishing an adaptive cut-off that scales with query complexity. On Coinbase's CDP corpus and the public MultiHop-RAG benchmark, CAR consistently picks the optimal retrieval depth and achieves the highest TES score, outperforming every fixed top-k baseline. In downstream RAG evaluations, CAR cuts LLM token usage by 60%, trims end-to-end latency by 22%, and reduces hallucinations by 10% while fully preserving answer relevance. Since integrating CAR into Coinbase's virtual assistant, we've seen user engagement jump by 200%.
Cortex AISQL: A Production SQL Engine for Unstructured Data
Snowflake's Cortex AISQL is a production SQL engine that integrates native semantic operations directly into SQL. This integration allows users to write declarative queries that combine relational operations with semantic reasoning, enabling them to query both structured and unstructured data effortlessly. However, making semantic operations efficient at production scale poses fundamental challenges. Semantic operations are more expensive than traditional SQL operations, possess distinct latency and throughput characteristics, and their cost and selectivity are unknown during query compilation. Furthermore, existing query engines are not designed to optimize semantic operations. The AISQL query execution engine addresses these challenges through three novel techniques informed by production deployment data from Snowflake customers. First, AI-aware query optimization treats AI inference cost as a first-class optimization objective, reasoning about large language model (LLM) cost directly during query planning to achieve 2-8$\times$ speedups. Second, adaptive model cascades reduce inference costs by routing most rows through a fast proxy model while escalating uncertain cases to a powerful oracle model, achieving 2-6$\times$ speedups while maintaining 90-95% of oracle model quality. Third, semantic join query rewriting lowers the quadratic time complexity of join operations to linear through reformulation as multi-label classification tasks, achieving 15-70$\times$ speedups with often improved prediction quality. AISQL is deployed in production at Snowflake, where it powers diverse customer workloads across analytics, search, and content understanding.
Weather Maps as Tokens: Transformers for Renewable Energy Forecasting
Accurate renewable energy forecasting is essential to reduce dependence on fossil fuels and enabling grid decarbonization. However, current approaches fail to effectively integrate the rich spatial context of weather patterns with their temporal evolution. This work introduces a novel approach that treats weather maps as tokens in transformer sequences to predict renewable energy. Hourly weather maps are encoded as spatial tokens using a lightweight convolutional neural network, and then processed by a transformer to capture temporal dynamics across a 45-hour forecast horizon. Despite disadvantages in input initialization, evaluation against ENTSO-E operational forecasts shows a reduction in RMSE of about 60% and 20% for wind and solar respectively. A live dashboard showing daily forecasts is available at: https://www.sardiniaforecast.ifabfoundation.it.
Infrastructuring Pop-Up Cities with "Social Layer": Designing Serendipitous Co-Livings for Temporary Intentional Communities
After the pandemic, a new form of "pop-up city" has emerged -- co-living gatherings of 100-200 people for 4-8 weeks that differ from conferences and hack houses. These temporary intentional communities leverages existing urban infrastructure, blending daily life (housing, meals, care) with self-organized activities like learning, creating, and socializing. They coordinate bottom-up programming through an "unconference" system for identity, calendaring, RSVP, and social discovery that fosters spontaneous, serendipitous, enduring ties. This paper examines the design of "Social Layer," an unconferencing system for pop-up cities. We studied its real-world deployment in ShanHaiWoo (Jilin, China, 2023), muChiangmai (Chiangmai, Thailand, 2023), Edge Esmeralda, Edge Esmeralda (Healdsburg, CA, USA, 2024), Aleph (Buenos Aires, Argentina, 2024), and Gathering of Tribe (Lisbon, Portugal, 2024). Our findings distill: (1) the strong concept "scaffolded spontaneity" -- infrastructural affordances that balance structure with openness, amplifying participant agency while maintaining privacy and lightweight governance; (2) design implications for design researchers working on pop-up cities.
RL-100: Performant Robotic Manipulation with Real-World Reinforcement Learning
Real-world robotic manipulation in homes and factories demands reliability, efficiency, and robustness that approach or surpass the performance of skilled human operators. We present RL-100, a real-world reinforcement learning framework built on diffusion-based visuomotor policies. RL-100 unifies imitation and reinforcement learning under a single PPO-style objective applied within the denoising process, yielding conservative and stable policy improvements across both offline and online stages. To meet deployment latency constraints, we employ a lightweight consistency distillation procedure that compresses multi-step diffusion into a one-step controller for high-frequency control. The framework is task-, embodiment-, and representation-agnostic, and supports both single-action outputs and action-chunking control. We evaluate RL-100 on seven diverse real-robot manipulation tasks, ranging from dynamic pushing and agile bowling to pouring, cloth folding, unscrewing, and multi-stage juicing. RL-100 attains 100% success across evaluated trials, achieving 900 out of 900 successful episodes, including up to 250 out of 250 consecutive trials on one task, and matches or surpasses expert teleoperators in time-to-completion. Without retraining, a single policy attains approximately 90% zero-shot success under environmental and dynamics shifts, adapts in a few-shot regime to significant task variations (86.7%), and remains robust to aggressive human perturbations (about 95%). In a public shopping-mall deployment, the juicing robot served random customers continuously for roughly seven hours without failure. Together, these results suggest a practical path toward deployment-ready robot learning: start from human priors, align training objectives with human-grounded metrics, and reliably extend performance beyond human demonstrations.
SilverTorch: A Unified Model-based System to Democratize Large-Scale Recommendation on GPUs
Serving deep learning based recommendation models (DLRM) at scale is challenging. Existing systems rely on CPU-based ANN indexing and filtering services, suffering from non-negligible costs and forgoing joint optimization opportunities. Such inefficiency makes them difficult to support more complex model architectures, such as learned similarities and multi-task retrieval. In this paper, we propose SilverTorch, a model-based system for serving recommendation models on GPUs. SilverTorch unifies model serving by replacing standalone indexing and filtering services with layers of served models. We propose a Bloom index algorithm on GPUs for feature filtering and a tensor-native fused Int8 ANN kernel on GPUs for nearest neighbor search. We further co-design the ANN search index and filtering index to reduce GPU memory utilization and eliminate unnecessary computation. Benefit from SilverTorch's serving paradigm, we introduce a OverArch scoring layer and a Value Model to aggregate results across multi-tasks. These advancements improve the accuracy for retrieval and enable future studies for serving more complex models. For ranking, SilverTorch's design accelerates item embedding calculation by caching the pre-calculated embeddings inside the serving model. Our evaluation on the industry-scale datasets show that SilverTorch achieves up to 5.6x lower latency and 23.7x higher throughput compared to the state-of-the-art approaches. We also demonstrate that SilverTorch's solution is 13.35x more cost-efficient than CPU-based solution while improving accuracy via serving more complex models. SilverTorch serves over hundreds of models online across major products and recommends contents for billions of daily active users.
FinTRec: Transformer Based Unified Contextual Ads Targeting and Personalization for Financial Applications
Transformer-based architectures are widely adopted in sequential recommendation systems, yet their application in Financial Services (FS) presents distinct practical and modeling challenges for real-time recommendation. These include:a) long-range user interactions (implicit and explicit) spanning both digital and physical channels generating temporally heterogeneous context, b) the presence of multiple interrelated products require coordinated models to support varied ad placements and personalized feeds, while balancing competing business goals. We propose FinTRec, a transformer-based framework that addresses these challenges and its operational objectives in FS. While tree-based models have traditionally been preferred in FS due to their explainability and alignment with regulatory requirements, our study demonstrate that FinTRec offers a viable and effective shift toward transformer-based architectures. Through historic simulation and live A/B test correlations, we show FinTRec consistently outperforms the production-grade tree-based baseline. The unified architecture, when fine-tuned for product adaptation, enables cross-product signal sharing, reduces training cost and technical debt, while improving offline performance across all products. To our knowledge, this is the first comprehensive study of unified sequential recommendation modeling in FS that addresses both technical and business considerations.
Scalable and Efficient Large-Scale Log Analysis with LLMs: An IT Software Support Case Study
IT environments typically have logging mechanisms to monitor system health and detect issues. However, the huge volume of generated logs makes manual inspection impractical, highlighting the importance of automated log analysis in IT Software Support. In this paper, we propose a log analytics tool that leverages Large Language Models (LLMs) for log data processing and issue diagnosis, enabling the generation of automated insights and summaries. We further present a novel approach for efficiently running LLMs on CPUs to process massive log volumes in minimal time without compromising output quality. We share the insights and lessons learned from deployment of the tool - in production since March 2024 - scaled across 70 software products, processing over 2000 tickets for issue diagnosis, achieving a time savings of 300+ man hours and an estimated $15,444 per month in manpower costs compared to the traditional log analysis practices.
Making Evidence Actionable in Adaptive Learning
Adaptive learning often diagnoses precisely yet intervenes weakly, yielding help that is mistimed or misaligned. This study presents evidence supporting an instructor-governed feedback loop that converts concept-level assessment evidence into vetted micro-interventions. The adaptive learning algorithm contains three safeguards: adequacy as a hard guarantee of gap closure, attention as a budgeted constraint for time and redundancy, and diversity as protection against overfitting to a single resource. We formalize intervention assignment as a binary integer program with constraints for coverage, time, difficulty windows informed by ability estimates, prerequisites encoded by a concept matrix, and anti-redundancy enforced through diversity. Greedy selection serves low-richness and tight-latency regimes, gradient-based relaxation serves rich repositories, and a hybrid method transitions along a richness-latency frontier. In simulation and in an introductory physics deployment with one thousand two hundred four students, both solvers achieved full skill coverage for essentially all learners within bounded watch time. The gradient-based method reduced redundant coverage by approximately twelve percentage points relative to greedy and harmonized difficulty across slates, while greedy delivered comparable adequacy with lower computational cost in scarce settings. Slack variables localized missing content and supported targeted curation, sustaining sufficiency across subgroups. The result is a tractable and auditable controller that closes the diagnostic-pedagogical loop and delivers equitable, load-aware personalization at classroom scale.
TaoSearchEmb: A Multi-Objective Reinforcement Learning Framework for Dense Retrieval in Taobao Search
Dense retrieval, as the core component of e-commerce search engines, maps user queries and items into a unified semantic space through pre-trained embedding models to enable large-scale real-time semantic retrieval. Despite the rapid advancement of LLMs gradually replacing traditional BERT architectures for embedding, their training paradigms still adhere to BERT-like supervised fine-tuning and hard negative mining strategies. This approach relies on complex offline hard negative sample construction pipelines, which constrain model iteration efficiency and hinder the evolutionary potential of semantic representation capabilities. Besides, existing multi-task learning frameworks face the seesaw effect when simultaneously optimizing semantic relevance and non-relevance objectives. In this paper, we propose Retrieval-GRPO, a multi-objective reinforcement learning-based dense retrieval framework designed to address these challenges. The method eliminates offline hard negative sample construction by dynamically retrieving Top-K candidate products for each query during training, while introducing a relevance LLM as a reward model to generate real-time feedback. Specifically, the retrieval model dynamically optimizes embedding representations through reinforcement learning, with reward signals combining LLM-generated relevance scores, product quality scores, and multi-way exclusivity metrics to achieve multi-objective user preference alignment and real-time error correction. This mechanism not only removes dependency on hard negatives but also mitigates the seesaw effect through collaborative multi-objective optimization, significantly enhancing the model's semantic generalization capability for complex long-tail queries. Extensive offline and online experiments validate the effectiveness of Retrieval-GRPO, which has been deployed on China's largest e-commerce platform.
From Reasoning LLMs to BERT: A Two-Stage Distillation Framework for Search Relevance
Query-service relevance prediction in e-commerce search systems faces strict latency requirements that prevent the direct application of Large Language Models (LLMs). To bridge this gap, we propose a two-stage reasoning distillation framework to transfer reasoning capabilities from a powerful teacher LLM to a lightweight, deployment-friendly student model. In the first stage, we address the limitations of general-purpose LLMs by constructing a domain-adapted teacher model. This is achieved through a three-step process: domain-adaptive pre-training to inject platform knowledge, supervised fine-tuning to elicit reasoning skills, and preference optimization with a multi-dimensional reward model to ensure the generation of reliable and preference-aligned reasoning paths. This teacher can then automatically annotate massive query-service pairs from search logs with both relevance labels and reasoning chains. In the second stage, to address the challenges of architectural heterogeneity in standard distillation, we introduce Contrastive Reasoning Self-Distillation (CRSD). By modeling the behavior of the same student model under "standard" and "reasoning-augmented" inputs as a teacher-student relationship, CRSD enables the lightweight model to internalize the teacher's complex decision-making mechanisms without needing the explicit reasoning path at inference. Offline evaluations and online A/B testing in the Meituan search advertising system demonstrate that our framework achieves significant improvements across multiple metrics, validating its effectiveness and practical value.
In-IDE Programming Courses: Learning Software Development in a Real-World Setting
While learning programming languages is crucial for software engineers, mastering the necessary tools is equally important. To facilitate this, JetBrains recently released the JetBrains Academy plugin, which customizes the IDE for learners, allowing tutors to create courses entirely within IDE. In this work, we provide the first exploratory study of this learning format. We carried out eight one-hour interviews with students and developers who completed at least one course using the plugin, inquiring about their experience with the format, the used IDE features, and the current shortcomings. Our results indicate that learning inside the IDE is overall welcomed by the learners, allowing them to study in a more realistic setting, using features such as debugging and code analysis, which are crucial for real software development. With the collected results and the analysis of the current drawbacks, we aim to contribute to teaching students more practical skills.
GAEA: Experiences and Lessons Learned from a Country-Scale Environmental Digital Twin
This paper describes the experiences and lessons learned after the deployment of a country-scale environmental digital twin on the island of Cyprus for three years. This digital twin, called GAEA, contains 27 environmental geospatial services and is suitable for urban planners, policymakers, farmers, property owners, real-estate and forestry professionals, as well as insurance companies and banks that have properties in their portfolio. This paper demonstrates the power, potential, current and future challenges of geospatial analytics and environmental digital twins on a large scale.
Deploying Rapid Damage Assessments from sUAS Imagery for Disaster Response
This paper presents the first AI/ML system for automating building damage assessment in uncrewed aerial systems (sUAS) imagery to be deployed operationally during federally declared disasters (Hurricanes Debby and Helene). In response to major disasters, sUAS teams are dispatched to collect imagery of the affected areas to assess damage; however, at recent disasters, teams collectively delivered between 47GB and 369GB of imagery per day, representing more imagery than can reasonably be transmitted or interpreted by subject matter experts in the disaster scene, thus delaying response efforts. To alleviate this data avalanche encountered in practice, computer vision and machine learning techniques are necessary. While prior work has been deployed to automatically assess damage in satellite imagery, there is no current state of practice for sUAS-based damage assessment systems, as all known work has been confined to academic settings. This work establishes the state of practice via the development and deployment of models for building damage assessment with sUAS imagery. The model development involved training on the largest known dataset of post-disaster sUAS aerial imagery, containing 21,716 building damage labels, and the operational training of 91 disaster practitioners. The best performing model was deployed during the responses to Hurricanes Debby and Helene, where it assessed a combined 415 buildings in approximately 18 minutes. This work contributes documentation of the actual use of AI/ML for damage assessment during a disaster and lessons learned to the benefit of the AI/ML research and user communities.
Introducing AI to an Online Petition Platform Changed Outputs but not Outcomes
The rapid integration of AI writing tools into online platforms raises critical questions about their impact on content production and outcomes. We leverage a unique natural experiment on Change.org, a leading social advocacy platform, to causally investigate the effects of an in-platform ''write with AI'' tool. To understand the impact of the AI integration, we collected 1.5 million petitions and employed a difference-in-differences analysis. Our findings reveal that in-platform AI access significantly altered the lexical features of petitions and increased petition homogeneity, but did not improve petition outcomes. We confirmed the results in a separate analysis of repeat petition writers who wrote petitions before and after introduction of the AI tool. The results suggest that while AI writing tools can profoundly reshape online content, their practical utility for improving desired outcomes may be less beneficial than anticipated, and introduce unintended consequences like content homogenization.
Zipf-Gramming: Scaling Byte N-Grams Up to Production Sized Malware Corpora
A classifier using byte n-grams as features is the only approach we have found fast enough to meet requirements in size (sub 2 MB), speed (multiple GB/s), and latency (sub 10 ms) for deployment in numerous malware detection scenarios. However, we've consistently found that 6-8 grams achieve the best accuracy on our production deployments but have been unable to deploy regularly updated models due to the high cost of finding the top-k most frequent n-grams over terabytes of executable programs. Because the Zipfian distribution well models the distribution of n-grams, we exploit its properties to develop a new top-k n-gram extractor that is up to $35\times$ faster than the previous best alternative. Using our new Zipf-Gramming algorithm, we are able to scale up our production training set and obtain up to 30\% improvement in AUC at detecting new malware. We show theoretically and empirically that our approach will select the top-k items with little error and the interplay between theory and engineering required to achieve these results.
MOON Embedding: Multimodal Representation Learning for E-commerce Search Advertising
We introduce MOON, our comprehensive set of sustainable iterative practices for multimodal representation learning for e-commerce applications. MOON has already been fully deployed across all stages of Taobao search advertising system, including retrieval, relevance, ranking, and so on. The performance gains are particularly significant on click-through rate (CTR) prediction task, which achieves an overall +20.00% online CTR improvement. Over the past three years, this project has delivered the largest improvement on CTR prediction task and undergone five full-scale iterations. Throughout the exploration and iteration of our MOON, we have accumulated valuable insights and practical experience that we believe will benefit the research community. MOON contains a three-stage training paradigm of "Pretraining, Post-training, and Application", allowing effective integration of multimodal representations with downstream tasks. Notably, to bridge the misalignment between the objectives of multimodal representation learning and downstream training, we define the exchange rate to quantify how effectively improvements in an intermediate metric can translate into downstream gains. Through this analysis, we identify the image-based search recall as a critical intermediate metric guiding the optimization of multimodal models. Over three years and five iterations, MOON has evolved along four critical dimensions: data processing, training strategy, model architecture, and downstream application. The lessons and insights gained through the iterative improvements will also be shared. As part of our exploration into scaling effects in the e-commerce field, we further conduct a systematic study of the scaling laws governing multimodal representation learning, examining multiple factors such as the number of training tokens, negative samples, and the length of user behavior sequences.
DevPiolt: Operation Recommendation for IoT Devices at Xiaomi Home
Operation recommendation for IoT devices refers to generating personalized device operations for users based on their context, such as historical operations, environment information, and device status. This task is crucial for enhancing user satisfaction and corporate profits. Existing recommendation models struggle with complex operation logic, diverse user preferences, and sensitive to suboptimal suggestions, limiting their applicability to IoT device operations. To address these issues, we propose DevPiolt, a LLM-based recommendation model for IoT device operations. Specifically, we first equip the LLM with fundamental domain knowledge of IoT operations via continual pre-training and multi-task fine-tuning. Then, we employ direct preference optimization to align the fine-tuned LLM with specific user preferences. Finally, we design a confidence-based exposure control mechanism to avoid negative user experiences from low-quality recommendations. Extensive experiments show that DevPiolt significantly outperforms baselines on all datasets, with an average improvement of 69.5% across all metrics. DevPiolt has been practically deployed in Xiaomi Home app for one quarter, providing daily operation recommendations to 255,000 users. Online experiment results indicate a 21.6% increase in unique visitor device coverage and a 29.1% increase in page view acceptance rates.
IndiTag: An Online Media Bias Analysis System Using Fine-Grained Bias Indicators
In the age of information overload and polarized discourse, understanding media bias has become imperative for informed decision-making and fostering a balanced public discourse. However, without the experts' analysis, it is hard for the readers to distinguish bias from the news articles. This paper presents IndiTag, an innovative online media bias analysis system that leverages fine-grained bias indicators to dissect and distinguish bias in digital content. IndiTag offers a novel approach by incorporating large language models, bias indicators, and vector database to detect and interpret bias automatically. Complemented by a user-friendly interface facilitating automated bias analysis for readers, IndiTag offers a comprehensive platform for in-depth bias examination. We demonstrate the efficacy and versatility of IndiTag through experiments on four datasets encompassing news articles from diverse platforms. Furthermore, we discuss potential applications of IndiTag in fostering media literacy, facilitating fact-checking initiatives, and enhancing the transparency and accountability of digital media platforms. IndiTag stands as a valuable tool in the pursuit of fostering a more informed, discerning, and inclusive public discourse in the digital age. We release an online system for end users and the source code is available at https://github.com/lylin0/IndiTag.
Ken Utilization Layer: Hebbian Replay Within a Student's Ken for Adaptive Exercise Recommendation
Adaptive exercise recommendation (ER) aims to choose the next activity that matches a learner's evolving Zone of Proximal Development (ZPD). We present KUL-Rec, a biologically inspired ER system that couples a fast Hebbian memory with slow replay-based consolidation to enable continual, few-shot personalization from sparse interactions. The model operates in an embedding space, allowing a single architecture to handle both tabular knowledge-tracing logs and open-ended short-answer text. We align evaluation with tutoring needs using bidirectional ranking and rank-sensitive metrics (nDCG, Recall@K). Across ten public datasets, KUL-Rec improves macro nDCG (0.316 vs. 0.265 for the strongest baseline) and Recall@10 (0.305 vs. 0.211), while achieving low inference latency and an $\approx99$\% reduction in peak GPU memory relative to a competitive graph-based model. In a 13-week graduate course, KUL-Rec personalized weekly short-answer quizzes generated by a retrieval-augmented pipeline and the personalized quizzes were associated with lower perceived difficulty and higher helpfulness (p < .05). An embedding robustness audit highlights that encoder choice affects semantic alignment, motivating routine audits when deploying open-response assessment. Together, these results indicate that Hebbian replay with bounded consolidation offers a practical path to real-time, interpretable ER that scales across data modalities and classroom settings.
LEMUR: Large scale End-to-end MUltimodal Recommendation
Traditional ID-based recommender systems often struggle with cold-start and generalization challenges. Multimodal recommendation systems, which leverage textual and visual data, offer a promising solution to mitigate these issues. However, existing industrial approaches typically adopt a two-stage training paradigm: first pretraining a multimodal model, then applying its frozen representations to train the recommendation model. This decoupled framework suffers from misalignment between multimodal learning and recommendation objectives, as well as an inability to adapt dynamically to new data. To address these limitations, we propose LEMUR, the first large-scale multimodal recommender system trained end-to-end from raw data. By jointly optimizing both the multimodal and recommendation components, LEMUR ensures tighter alignment with downstream objectives while enabling real-time parameter updates. Constructing multimodal sequential representations from user history often entails prohibitively high computational costs. To alleviate this bottleneck, we propose a novel memory bank mechanism that incrementally accumulates historical multimodal representations throughout the training process. After one month of deployment in Douyin Search, LEMUR has led to a 0.843% reduction in query change rate decay and a 0.81% improvement in QAUC. Additionally, LEMUR has shown significant gains across key offline metrics for Douyin Advertisement. Our results validate the superiority of end-to-end multimodal recommendation in real-world industrial scenarios.
Coordinated Humanoid Robot Locomotion with Symmetry Equivariant Reinforcement Learning Policy
The human nervous system exhibits bilateral symmetry, enabling coordinated and balanced movements. However, existing Deep Reinforcement Learning (DRL) methods for humanoid robots neglect morphological symmetry of the robot, leading to uncoordinated and suboptimal behaviors. Inspired by human motor control, we propose Symmetry Equivariant Policy (SE-Policy), a new DRL framework that embeds strict symmetry equivariance in the actor and symmetry invariance in the critic without additional hyperparameters. SE-Policy enforces consistent behaviors across symmetric observations, producing temporally and spatially coordinated motions with higher task performance. Extensive experiments on velocity tracking tasks, conducted in both simulation and real-world deployment with the Unitree G1 humanoid robot, demonstrate that SE-Policy improves tracking accuracy by up to 40% compared to state-of-the-art baselines, while achieving superior spatial-temporal coordination. These results demonstrate the effectiveness of SE-Policy and its broad applicability to humanoid robots.
Small Models, Big Support: A Local LLM Framework for Educator-Centric Content Creation and Assessment with RAG and CAG
While Large Language Models (LLMs) are increasingly applied in student-facing educational tools, their potential to directly support educators through locally deployable and customizable solutions remains underexplored. Many existing approaches rely on proprietary, cloud-based systems that raise significant cost, privacy, and control concerns for educational institutions. To address these barriers, we introduce an end-to-end, open-source framework that empowers educators using small (3B-7B parameter), locally deployable LLMs. Our system is designed for comprehensive teacher support, including customized teaching material generation and AI-assisted assessment. The framework synergistically combines Retrieval-Augmented Generation (RAG) and Context-Augmented Generation (CAG) to produce factually accurate, pedagogically-styled content. A core feature is an interactive refinement loop, a teacher-in-the-loop mechanism that ensures educator agency and precise alignment of the final output. To enhance reliability and safety, an auxiliary verifier LLM inspects all generated content. We validate our framework through a rigorous evaluation of its content generation capabilities and report on a successful technical deployment in a college physics course, which confirms its feasibility on standard institutional hardware. Our findings demonstrate that carefully engineered, self-hosted systems built on small LLMs can provide robust, affordable, and private support for educators, achieving practical utility comparable to much larger models for targeted instructional tasks. This work presents a practical blueprint for the development of sovereign AI tools tailored to the real-world needs of educational institutions.
OlmoEarth: Stable Latent Image Modeling for Multimodal Earth Observation
Earth observation data presents a unique challenge: it is spatial like images, sequential like video or text, and highly multimodal. We present OlmoEarth: a multimodal, spatio-temporal foundation model that employs a novel self-supervised learning formulation, masking strategy, and loss all designed for the Earth observation domain. OlmoEarth achieves state-of-the-art performance compared to 12 other foundation models across a variety of research benchmarks and real-world tasks from external partners. When evaluating embeddings OlmoEarth achieves the best performance on 15 out of 24 tasks, and with full fine-tuning it is the best on 19 of 29 tasks. We deploy OlmoEarth as the backbone of an end-to-end platform for data collection, labeling, training, and inference of Earth observation models. The OlmoEarth Platform puts frontier foundation models and powerful data management tools into the hands of non-profits and NGOs working to solve the world's biggest problems. OlmoEarth source code, training data, and pre-trained weights are available at $\href{https://github.com/allenai/olmoearth_pretrain}{\text{https://github.com/allenai/olmoearth_pretrain}}$.
RAGPulse: An Open-Source RAG Workload Trace to Optimize RAG Serving Systems
Retrieval-Augmented Generation (RAG) is a critical paradigm for building reliable, knowledge-intensive Large Language Model (LLM) applications. However, the multi-stage pipeline (retrieve, generate) and unique workload characteristics (e.g., knowledge dependency) of RAG systems pose significant challenges for serving performance optimization. Existing generic LLM inference traces fail to capture these RAG-specific dynamics, creating a significant performance gap between academic research and real-world deployment. To bridge this gap, this paper introduces RAGPulse, an open-source RAG workload trace dataset. This dataset was collected from an university-wide Q&A system serving that has served more than 40,000 students and faculties since April 2024. We detail RAGPulse's system architecture, its privacy-preserving hash-based data format, and provide an in-depth statistical analysis. Our analysis reveals that real-world RAG workloads exhibit significant temporal locality and a highly skewed hot document access pattern. RAGPulse provides a high-fidelity foundation for researchers to develop and validate novel optimization strategies for RAG systems, such as content-aware batching and retrieval caching, ultimately enhancing the efficiency and reliability of RAG services. The code is available at https://github.com/flashserve/RAGPulse.
From Scaling to Structured Expressivity: Rethinking Transformers for CTR Prediction
Despite massive investments in scale, deep models for click-through rate (CTR) prediction often exhibit rapidly diminishing returns - a stark contrast to the smooth, predictable gains seen in large language models. We identify the root cause as a structural misalignment: Transformers assume sequential compositionality, while CTR data demand combinatorial reasoning over high-cardinality semantic fields. Unstructured attention spreads capacity indiscriminately, amplifying noise under extreme sparsity and breaking scalable learning. To restore alignment, we introduce the Field-Aware Transformer (FAT), which embeds field-based interaction priors into attention through decomposed content alignment and cross-field modulation. This design ensures model complexity scales with the number of fields F, not the total vocabulary size n >> F, leading to tighter generalization and, critically, observed power-law scaling in AUC as model width increases. We present the first formal scaling law for CTR models, grounded in Rademacher complexity, that explains and predicts this behavior. On large-scale benchmarks, FAT improves AUC by up to +0.51% over state-of-the-art methods. Deployed online, it delivers +2.33% CTR and +0.66% RPM. Our work establishes that effective scaling in recommendation arises not from size, but from structured expressivity-architectural coherence with data semantics.
JobSphere: An AI-Powered Multilingual Career Copilot for Government Employment Platforms
Users of government employment websites commonly face engagement and accessibility challenges linked to navigational complexity, a dearth of language options, and a lack of personalized support. This paper introduces JobSphere, an AI-powered career assistant that is redefining the employment platform in Punjab called PGRKAM. JobSphere employs Retrieval-Augmented Generation (RAG) architecture, and it is multilingual, available in English, Hindi and Punjabi. JobSphere technique uses 4-bit quantization, allowing the platform to deploy on consumer-grade GPUs (i.e., NVIDIA RTX 3050 4GB), making the implementation 89% cheaper than that of cloud-based systems. Key innovations include voice-enabled interaction with the assistant, automated mock tests, resume parsing with skills recognition, and embed-based job recommendation that achieves a precision@10 score of 68%. An evaluation of JobSphere's implementation reveals 94% factual accuracy, a median response time of 1.8 seconds, and a System Usability Scale score of 78.5/100, a 50% improvement compared to the baseline PGRKAM platform context. In conclusion, JobSphere effectively fills significant accessibility gaps for Punjab/Hindi-speaking users in rural locations, while also affirming the users access to trusted job content provided by government agencies.
GRIN Transfer: A production-ready tool for libraries to retrieve digital copies from Google Books
Publicly launched in 2004, the Google Books project has scanned tens of millions of items in partnership with libraries around the world. As part of this project, Google created the Google Return Interface (GRIN). Through this platform, libraries can access their scanned collections, the associated metadata, and the ongoing OCR and metadata improvements that become available as Google reprocesses these collections using new technologies. When downloading the Harvard Library Google Books collection from GRIN to develop the Institutional Books dataset, we encountered several challenges related to rate-limiting and atomized metadata within the GRIN platform. To overcome these challenges and help other libraries make more robust use of their Google Books collections, this technical report introduces the initial release of GRIN Transfer. This open-source and production-ready Python pipeline allows partner libraries to efficiently retrieve their Google Books collections from GRIN. This report also introduces an updated version of our Institutional Books 1.0 pipeline, initially used to analyze, augment, and assemble the Institutional Books 1.0 dataset. We have revised this pipeline for compatibility with the output format of GRIN Transfer. A library could pair these two tools to create an end-to-end processing pipeline for their Google Books collection to retrieve, structure, and enhance data available from GRIN. This report gives an overview of how GRIN Transfer was designed to optimize for reliability and usability in different environments, as well as guidance on configuration for various use cases.
A GPU-Accelerated RAG-Based Telegram Assistant for Supporting Parallel Processing Students
This project addresses a critical pedagogical need: offering students continuous, on-demand academic assistance beyond conventional reception hours. I present a domain-specific Retrieval-Augmented Generation (RAG) system powered by a quantized Mistral-7B Instruct model and deployed as a Telegram bot. The assistant enhances learning by delivering real-time, personalized responses aligned with the "Introduction to Parallel Processing" course materials. GPU acceleration significantly improves inference latency, enabling practical deployment on consumer hardware. This approach demonstrates how consumer GPUs can enable affordable, private, and effective AI tutoring for HPC education.
Align$^3$GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation
Large Language Models (LLMs) demonstrate significant advantages in leveraging structured world knowledge and multi-step reasoning capabilities. However, fundamental challenges arise when transforming LLMs into real-world recommender systems due to semantic and behavioral misalignment. To bridge this gap, we propose Align$^3$GR, a novel framework that unifies token-level, behavior modeling-level, and preference-level alignment. Our approach introduces: Dual tokenization fusing user-item semantic and collaborative signals. Enhanced behavior modeling with bidirectional semantic alignment. Progressive DPO strategy combining self-play (SP-DPO) and real-world feedback (RF-DPO) for dynamic preference adaptation. Experiments show Align$^3$GR outperforms the SOTA baseline by +17.8% in Recall@10 and +20.2% in NDCG@10 on the public dataset, with significant gains in online A/B tests and full-scale deployment on an industrial large-scale recommendation platform.
GovScape: A Public Multimodal Search System for 70 Million Pages of Government PDFs
Efforts over the past three decades have produced web archives containing billions of webpage snapshots and petabytes of data. The End of Term Web Archive alone contains, among other file types, millions of PDFs produced by the federal government. While preservation with web archives has been successful, significant challenges for access and discoverability remain. For example, current affordances for browsing the End of Term PDFs are limited to downloading and browsing individual PDFs, as well as performing basic keyword search across them. In this paper, we introduce GovScape, a public search system that supports multimodal searches across 10,015,993 federal government PDFs from the 2020 End of Term crawl (70,958,487 total PDF pages) - to our knowledge, all renderable PDFs in the 2020 crawl that are 50 pages or under. GovScape supports four primary forms of search over these 10 million PDFs: in addition to providing (1) filter conditions over metadata facets including domain and crawl date and (2) exact text search against the PDF text, we provide (3) semantic text search and (4) visual search against the PDFs across individual pages, enabling users to structure queries such as "redacted documents" or "pie charts." We detail the constituent components of GovScape, including the search affordances, embedding pipeline, system architecture, and open source codebase. Significantly, the total estimated compute cost for GovScape's pre-processing pipeline for 10 million PDFs was approximately $1,500, equivalent to 47,000 PDF pages per dollar spent on compute, demonstrating the potential for immediate scalability. Accordingly, we outline steps that we have already begun pursuing toward multimodal search at the 100+ million PDF scale. GovScape can be found at https://www.govscape.net.
An External Fairness Evaluation of LinkedIn Talent Search
We conduct an independent, third-party audit for bias of LinkedIn's Talent Search ranking system, focusing on potential ranking bias across two attributes: gender and race. To do so, we first construct a dataset of rankings produced by the system, collecting extensive Talent Search results across a diverse set of occupational queries. We then develop a robust labeling pipeline that infers the two demographic attributes of interest for the returned users. To evaluate potential biases in the collected dataset of real-world rankings, we utilize two exposure disparity metrics: deviation from group proportions and MinSkew. Our analysis reveals an under-representation of minority groups in early ranks across many queries. We further examine potential causes of this disparity, and discuss why they may be difficult or, in some cases, impossible to fully eliminate among the early ranks of queries. Beyond static metrics, we also investigate the concept of subgroup fairness over time, highlighting temporal disparities in exposure and retention, which are often more difficult to audit for in practice. In employer recruiting platforms such as LinkedIn Talent Search, the persistence of a particular candidate over multiple days in the ranking can directly impact the probability that the given candidate is selected for opportunities. Our analysis reveals demographic disparities in this temporal stability, with some groups experiencing greater volatility in their ranked positions than others. We contextualize all our findings alongside LinkedIn's published self-audits of its Talent Search system and reflect on the methodological constraints of a black-box external evaluation, including limited observability and noisy demographic inference.
GPR: Towards a Generative Pre-trained One-Model Paradigm for Large-Scale Advertising Recommendation
As an intelligent infrastructure connecting users with commercial content, advertising recommendation systems play a central role in information flow and value creation within the digital economy. However, existing multi-stage advertising recommendation systems suffer from objective misalignment and error propagation, making it difficult to achieve global optimality, while unified generative recommendation models still struggle to meet the demands of practical industrial applications. To address these issues, we propose GPR (Generative Pre-trained Recommender), the first one-model framework that redefines advertising recommendation as an end-to-end generative task, replacing the traditional cascading paradigm with a unified generative approach. To realize GPR, we introduce three key innovations spanning unified representation, network architecture, and training strategy. First, we design a unified input schema and tokenization method tailored to advertising scenarios, mapping both ads and organic content into a shared multi-level semantic ID space, thereby enhancing semantic alignment and modeling consistency across heterogeneous data. Second, we develop the Heterogeneous Hierarchical Decoder (HHD), a dual-decoder architecture that decouples user intent modeling from ad generation, achieving a balance between training efficiency and inference flexibility while maintaining strong modeling capacity. Finally, we propose a multi-stage joint training strategy that integrates Multi-Token Prediction (MTP), Value-Aware Fine-Tuning and the Hierarchy Enhanced Policy Optimization (HEPO) algorithm, forming a complete generative recommendation pipeline that unifies interest modeling, value alignment, and policy optimization. GPR has been fully deployed in the Tencent Weixin Channels advertising system, delivering significant improvements in key business metrics including GMV and CTCVR.
Owlgorithm: Supporting Self-Regulated Learning in Competitive Programming through LLM-Driven Reflection
We present Owlgorithm, an educational platform that supports Self-Regulated Learning (SRL) in competitive programming (CP) through AI-generated reflective questions. Leveraging GPT-4o, Owlgorithm produces context-aware, metacognitive prompts tailored to individual student submissions. Integrated into a second- and third-year CP course, the system-provided reflective prompts adapted to student outcomes: guiding deeper conceptual insight for correct solutions and structured debugging for partial or failed ones. Our exploratory assessment of student ratings and TA feedback revealed both promising benefits and notable limitations. While many found the generated questions useful for reflection and debugging, concerns were raised about feedback accuracy and classroom usability. These results suggest advantages of LLM-supported reflection for novice programmers, though refinements are needed to ensure reliability and pedagogical value for advanced learners. From our experience, several key insights emerged: GenAI can effectively support structured reflection, but careful prompt design, dynamic adaptation, and usability improvements are critical to realizing their potential in education. We offer specific recommendations for educators using similar tools and outline next steps to enhance Owlgorithm's educational impact. The underlying framework may also generalize to other reflective learning contexts.
Does AI-Assisted Coding Deliver? A Difference-in-Differences Study of Cursor's Impact on Software Projects
Large language models (LLMs) have demonstrated the promise to revolutionize the field of software engineering. Among other things, LLM agents are rapidly gaining momentum in their application to software development, with practitioners claiming a multifold productivity increase after adoption. Yet, empirical evidence is lacking around these claims. In this paper, we estimate the causal effect of adopting a widely popular LLM agent assistant, namely Cursor, on development velocity and software quality. The estimation is enabled by a state-of-the-art difference-in-differences design comparing Cursor-adopting GitHub projects with a matched control group of similar GitHub projects that do not use Cursor. We find that the adoption of Cursor leads to a significant, large, but transient increase in project-level development velocity, along with a significant and persistent increase in static analysis warnings and code complexity. Further panel generalized method of moments estimation reveals that the increase in static analysis warnings and code complexity acts as a major factor causing long-term velocity slowdown. Our study carries implications for software engineering practitioners, LLM agent assistant designers, and researchers.
A Brain Cell Type Resource Created by Large Language Models and a Multi-Agent AI System for Collaborative Community Annotation
Single-cell RNA sequencing has transformed our ability to identify diverse cell types and their transcriptomic signatures. However, annotating these signatures-especially those involving poorly characterized genes-remains a major challenge. Traditional methods, such as Gene Set Enrichment Analysis (GSEA), depend on well-curated annotations and often perform poorly in these contexts. Large Language Models (LLMs) offer a promising alternative but struggle to represent complex biological knowledge within structured ontologies. To address this, we present BRAINCELL-AID (BRAINCELL-AID: https://biodataai.uth.edu/BRAINCELL-AID), a novel multi-agent AI system that integrates free-text descriptions with ontology labels to enable more accurate and robust gene set annotation. By incorporating retrieval-augmented generation (RAG), we developed a robust agentic workflow that refines predictions using relevant PubMed literature, reducing hallucinations and enhancing interpretability. Using this workflow, we achieved correct annotations for 77% of mouse gene sets among their top predictions. Applying this approach, we annotated 5,322 brain cell clusters from the comprehensive mouse brain cell atlas generated by the BRAIN Initiative Cell Census Network, enabling novel insights into brain cell function by identifying region-specific gene co-expression patterns and inferring functional roles of gene ensembles. BRAINCELL-AID also identifies Basal Ganglia-related cell types with neurologically meaningful descriptions. Hence, we create a valuable resource to support community-driven cell type annotation.
Democratizing Tabular Data Access with an Open$\unicode{x2013}$Source Synthetic$\unicode{x2013}$Data SDK
Machine learning development critically depends on access to high-quality data. However, increasing restrictions due to privacy, proprietary interests, and ethical concerns have created significant barriers to data accessibility. Synthetic data offers a viable solution by enabling safe, broad data usage without compromising sensitive information. This paper presents the MOSTLY AI Synthetic Data Software Development Kit (SDK), an open-source toolkit designed specifically for synthesizing high-quality tabular data. The SDK integrates robust features such as differential privacy guarantees, fairness-aware data generation, and automated quality assurance into a flexible and accessible Python interface. Leveraging the TabularARGN autoregressive framework, the SDK supports diverse data types and complex multi-table and sequential datasets, delivering competitive performance with notable improvements in speed and usability. Currently deployed both as a cloud service and locally installable software, the SDK has seen rapid adoption, highlighting its practicality in addressing real-world data bottlenecks and promoting widespread data democratization.
Unveiling the Impact of Data and Model Scaling on High-Level Control for Humanoid Robots
Data scaling has long remained a critical bottleneck in robot learning. For humanoid robots, human videos and motion data are abundant and widely available, offering a free and large-scale data source. Besides, the semantics related to the motions enable modality alignment and high-level robot control learning. However, how to effectively mine raw video, extract robot-learnable representations, and leverage them for scalable learning remains an open problem. To address this, we introduce Humanoid-Union, a large-scale dataset generated through an autonomous pipeline, comprising over 260 hours of diverse, high-quality humanoid robot motion data with semantic annotations derived from human motion videos. The dataset can be further expanded via the same pipeline. Building on this data resource, we propose SCHUR, a scalable learning framework designed to explore the impact of large-scale data on high-level control in humanoid robots. Experimental results demonstrate that SCHUR achieves high robot motion generation quality and strong text-motion alignment under data and model scaling, with 37\% reconstruction improvement under MPJPE and 25\% alignment improvement under FID comparing with previous methods. Its effectiveness is further validated through deployment in real-world humanoid robot.
DeepDR: an integrated deep-learning model web server for drug repositioning
Background: Identifying new indications for approved drugs is a complex and time-consuming process that requires extensive knowledge of pharmacology, clinical data, and advanced computational methods. Recently, deep learning (DL) methods have shown their capability for the accurate prediction of drug repositioning. However, implementing DL-based modeling requires in-depth domain knowledge and proficient programming skills. Results: In this application, we introduce DeepDR, the first integrated platform that combines a variety of established DL-based models for disease- and target-specific drug repositioning tasks. DeepDR leverages invaluable experience to recommend candidate drugs, which covers more than 15 networks and a comprehensive knowledge graph that includes 5.9 million edges across 107 types of relationships connecting drugs, diseases, proteins/genes, pathways, and expression from six existing databases and a large scientific corpus of 24 million PubMed publications. Additionally, the recommended results include detailed descriptions of the recommended drugs and visualize key patterns with interpretability through a knowledge graph. Conclusion: DeepDR is free and open to all users without the requirement of registration. We believe it can provide an easy-to-use, systematic, highly accurate, and computationally automated platform for both experimental and computational scientists.
AI-generated podcasts: Synthetic Intimacy and Cultural Translation in NotebookLM's Audio Overviews
This paper analyses AI-generated podcasts produced by Google's NotebookLM, which generates audio podcasts with two chatty AI hosts discussing whichever documents a user uploads. While AI-generated podcasts have been discussed as tools, for instance in medical education, they have not yet been analysed as media. By uploading different types of text and analysing the generated outputs I show how the podcasts' structure is built around a fixed template. I also find that NotebookLM not only translates texts from other languages into a perky standardised Mid-Western American accent, it also translates cultural contexts to a white, educated, middle-class American default. This is a distinct development in how publics are shaped by media, marking a departure from the multiple public spheres that scholars have described in human podcasting from the early 2000s until today, where hosts spoke to specific communities and responded to listener comments, to an abstraction of the podcast genre.
Systems for Scaling Accessibility Efforts in Large Computing Courses
It is critically important to make computing courses accessible for disabled students. This is particularly challenging in large computing courses, which face unique challenges due to the sheer scale of course content and staff. In this experience report, we share our attempts to scale accessibility efforts for a large university-level introductory programming course sequence, with over 3500 enrolled students and 100 teaching assistants (TAs) per year. First, we introduce our approach to auditing and remediating course materials by systematically identifying and resolving accessibility issues. However, remediating content post-hoc is purely reactive and scales poorly. We then discuss two approaches to systems that enable proactive accessibility work. We developed technical systems to manage remediation complexity at scale: redesigning other course content to be web-first and accessible by default, providing alternate accessible views for existing course content, and writing automated tests to receive instant feedback on a subset of accessibility issues. Separately, we established human systems to empower both course staff and students in accessibility best practices: developing and running various TA-targeted accessibility trainings, establishing course-wide accessibility norms, and integrating accessibility topics into core course curriculum. Preliminary qualitative feedback from both staff and students shows increased engagement in accessibility work and accessible technologies. We close by discussing limitations and lessons learned from our work, with advice for others developing similar auditing, remediation, technical, or human systems.
TaoSR-AGRL: Adaptive Guided Reinforcement Learning Framework for E-commerce Search Relevance
Query-product relevance prediction is fundamental to e-commerce search and has become even more critical in the era of AI-powered shopping, where semantic understanding and complex reasoning directly shape the user experience and business conversion. Large Language Models (LLMs) enable generative, reasoning-based approaches, typically aligned via supervised fine-tuning (SFT) or preference optimization methods like Direct Preference Optimization (DPO). However, the increasing complexity of business rules and user queries exposes the inability of existing methods to endow models with robust reasoning capacity for long-tail and challenging cases. Efforts to address this via reinforcement learning strategies like Group Relative Policy Optimization (GRPO) often suffer from sparse terminal rewards, offering insufficient guidance for multi-step reasoning and slowing convergence. To address these challenges, we propose TaoSR-AGRL, an Adaptive Guided Reinforcement Learning framework for LLM-based relevance prediction in Taobao Search Relevance. TaoSR-AGRL introduces two key innovations: (1) Rule-aware Reward Shaping, which decomposes the final relevance judgment into dense, structured rewards aligned with domain-specific relevance criteria; and (2) Adaptive Guided Replay, which identifies low-accuracy rollouts during training and injects targeted ground-truth guidance to steer the policy away from stagnant, rule-violating reasoning patterns toward compliant trajectories. TaoSR-AGRL was evaluated on large-scale real-world datasets and through online side-by-side human evaluations on Taobao Search. It consistently outperforms DPO and standard GRPO baselines in offline experiments, improving relevance accuracy, rule adherence, and training stability. The model trained with TaoSR-AGRL has been successfully deployed in the main search scenario on Taobao, serving hundreds of millions of users.
Beyond the Hype: Embeddings vs. Prompting for Multiclass Classification Tasks
Are traditional classification approaches irrelevant in this era of AI hype? We show that there are multiclass classification problems where predictive models holistically outperform LLM prompt-based frameworks. Given text and images from home-service project descriptions provided by Thumbtack customers, we build embeddings-based softmax models that predict the professional category (e.g., handyman, bathroom remodeling) associated with each problem description. We then compare against prompts that ask state-of-the-art LLM models to solve the same problem. We find that the embeddings approach outperforms the best LLM prompts in terms of accuracy, calibration, latency, and financial cost. In particular, the embeddings approach has 49.5\% higher accuracy than the prompting approach, and its superiority is consistent across text-only, image-only, and text-image problem descriptions. Furthermore, it yields well-calibrated probabilities, which we later use as confidence signals to provide contextualized user experience during deployment. On the contrary, prompting scores are overly uninformative. Finally, the embeddings approach is 14 and 81 times faster than prompting in processing images and text respectively, while under realistic deployment assumptions, it can be up to 10 times cheaper. Based on these results, we deployed a variation of the embeddings approach, and through A/B testing we observed performance consistent with our offline analysis. Our study shows that for multiclass classification problems that can leverage proprietary datasets, an embeddings-based approach may yield unequivocally better results. Hence, scientists, practitioners, engineers, and business leaders can use our study to go beyond the hype and consider appropriate predictive models for their classification use cases.
Computational Blueprints: Generating Isomorphic Mathematics Problems with Large Language Models
Personalized mathematics education is growing rapidly, creating a strong demand for large sets of similar practice problems. Yet existing studies on mathematics problem generation have focused on data augmentation for training neural language models rather than on direct educational deployment. To bridge this gap, we define a new task, Isomorphic Math Problem Generation (IMPG), designed to produce structurally consistent variants of source problems. Subsequently, we explored LLM-based frameworks for automatic IMPG through successive refinements, and established Computational Blueprints for Isomorphic Twins (CBIT). With meta-level generation and template-based selective variation, CBIT achieves high mathematical correctness and structural consistency while reducing the cost of generation. Empirical results across refinements demonstrate that CBIT is superior on generation accuracy and cost-effectiveness at scale. Most importantly, CBIT-generated problems exhibited an error rate 17.8% lower than expert-authored items, with deployment to 6,732 learners on a commercial education platform yielding 186,870 interactions.
Operational machine learning for remote spectroscopic detection of CH$_{4}$ point sources
Mitigating anthropogenic methane sources is one the most cost-effective levers to slow down global warming. While satellite-based imaging spectrometers, such as EMIT, PRISMA, and EnMAP, can detect these point sources, current methane retrieval methods based on matched filters still produce a high number of false detections requiring laborious manual verification. This paper describes the operational deployment of a machine learning system for detecting methane emissions within the Methane Alert and Response System (MARS) of the United Nations Environment Programme's International Methane Emissions Observatory. We created the largest and most diverse global dataset of annotated methane plumes from three imaging spectrometer missions and quantitatively compared different deep learning model configurations. Focusing on the requirements for operational deployment, we extended prior evaluation methodologies from small tiled datasets to full granule evaluation. This revealed that deep learning models still produce a large number of false detections, a problem we address with model ensembling, which reduced false detections by over 74%. Deployed in the MARS pipeline, our system processes scenes and proposes plumes to analysts, accelerating the detection and analysis process. During seven months of operational deployment, it facilitated the verification of 1,351 distinct methane leaks, resulting in 479 stakeholder notifications. We further demonstrate the model's utility in verifying mitigation success through case studies in Libya, Argentina, Oman, and Azerbaijan. Our work represents a critical step towards a global AI-assisted methane leak detection system, which is required to process the dramatically higher data volumes expected from new and current imaging spectrometers.
Enhancing reliability in AI inference services: An empirical study on real production incidents
Hyperscale large language model (LLM) inference places extraordinary demands on cloud systems, where even brief failures can translate into significant user and business impact. To better understand and mitigate these risks, we present one of the first provider-internal, practice-based analysis of LLM inference incidents. We developed a taxonomy and methodology grounded in a year of operational experience, validating it on 156 high-severity incidents, and conducted a focused quantitative study of Apr-Jun 2025 to ensure recency and relevance. Our approach achieves high labeling consistency (Cohen's K ~0.89), identifies dominant failure modes (in our dataset ~60% inference engine failures, within that category ~40% timeouts), and surfaces mitigation levers (~74% auto-detected; ~28% required hotfix). Beyond hotfixes, many incidents were mitigated via traffic routing, node rebalancing, or capacity increase policies, indicating further automation opportunities. We also show how the taxonomy guided targeted strategies such as connection liveness, GPU capacity-aware routing, and per-endpoint isolation and reduced incident impact and accelerated recovery. Finally, we contribute a practitioner-oriented adoption checklist that enables others to replicate our taxonomy, analysis, and automation opportunities in their own systems. This study demonstrates how systematic, empirically grounded analysis of inference operations can drive more reliable and cost-efficient LLM serving at scale.
Closing the Loop: An Instructor-in-the-Loop AI Assistance System for Supporting Student Help-Seeking in Programming Education
Timely and high-quality feedback is essential for effective learning in programming courses; yet, providing such support at scale remains a challenge. While AI-based systems offer scalable and immediate help, their responses can occasionally be inaccurate or insufficient. Human instructors, in contrast, may bring more valuable expertise but are limited in time and availability. To address these limitations, we present a hybrid help framework that integrates AI-generated hints with an escalation mechanism, allowing students to request feedback from instructors when AI support falls short. This design leverages the strengths of AI for scale and responsiveness while reserving instructor effort for moments of greatest need. We deployed this tool in a data science programming course with 82 students. We observe that out of the total 673 AI-generated hints, students rated 146 (22%) as unhelpful. Among those, only 16 (11%) of the cases were escalated to the instructors. A qualitative investigation of instructor responses showed that those feedback instances were incorrect or insufficient roughly half of the time. This finding suggests that when AI support fails, even instructors with expertise may need to pay greater attention to avoid making mistakes. We will publicly release the tool for broader adoption and enable further studies in other classrooms. Our work contributes a practical approach to scaling high-quality support and informs future efforts to effectively integrate AI and humans in education.
OneRec-Think: In-Text Reasoning for Generative Recommendation
The powerful generative capacity of Large Language Models (LLMs) has instigated a paradigm shift in recommendation. However, existing generative models (e.g., OneRec) operate as implicit predictors, critically lacking the capacity for explicit and controllable reasoning-a key advantage of LLMs. To bridge this gap, we propose OneRec-Think, a unified framework that seamlessly integrates dialogue, reasoning, and personalized recommendation. OneRec-Think incorporates: (1) Itemic Alignment: cross-modal Item-Textual Alignment for semantic grounding; (2) Reasoning Activation: Reasoning Scaffolding to activate LLM reasoning within the recommendation context; and (3) Reasoning Enhancement, where we design a recommendation-specific reward function that accounts for the multi-validity nature of user preferences. Experiments across public benchmarks show state-of-the-art performance. Moreover, our proposed "Think-Ahead" architecture enables effective industrial deployment on Kuaishou, achieving a 0.159\% gain in APP Stay Time and validating the practical efficacy of the model's explicit reasoning capability.
Benchmarking LLMs for Fine-Grained Code Review with Enriched Context in Practice
Code review is a cornerstone of software quality assurance, and recent advances in Large Language Models (LLMs) have shown promise in automating this process. However, existing benchmarks for LLM-based code review face three major limitations. (1) Lack of semantic context: most benchmarks provide only code diffs without textual information such as issue descriptions, which are crucial for understanding developer intent. (2) Data quality issues: without rigorous validation, many samples are noisy-e.g., reviews on outdated or irrelevant code-reducing evaluation reliability. (3) Coarse granularity: most benchmarks operate at the file or commit level, overlooking the fine-grained, line-level reasoning essential for precise review. We introduce ContextCRBench, a high-quality, context-rich benchmark for fine-grained LLM evaluation in code review. Our construction pipeline comprises: (1) Raw Data Crawling, collecting 153.7K issues and pull requests from top-tier repositories; (2) Comprehensive Context Extraction, linking issue-PR pairs for textual context and extracting the full surrounding function or class for code context; and (3) Multi-stage Data Filtering, combining rule-based and LLM-based validation to remove outdated, malformed, or low-value samples, resulting in 67,910 context-enriched entries. ContextCRBench supports three evaluation scenarios aligned with the review workflow: (1) hunk-level quality assessment, (2) line-level defect localization, and (3) line-level comment generation. Evaluating eight leading LLMs (four closed-source and four open-source) reveals that textual context yields greater performance gains than code context alone, while current LLMs remain far from human-level review ability. Deployed at ByteDance, ContextCRBench drives a self-evolving code review system, improving performance by 61.98% and demonstrating its robustness and industrial utility.
MetricSynth: Framework for Aggregating DORA and KPI Metrics Across Multi-Platform Engineering
In modern, large-scale software development, engineering leaders face the significant challenge of gaining a holistic and data-driven view of team performance and system health. Data is often siloed across numerous disparate tools, making manual report generation time-consuming and prone to inconsistencies. This paper presents the architecture and implementation of a centralized framework designed to provide near-real-time visibility into developer experience (DevEx) and Key Performance Indicator (KPI) metrics for a software ecosystem. By aggregating data from various internal tools and platforms, the system computes and visualizes metrics across key areas such as Developer Productivity, Quality, and Operational Efficiency. The architecture features a cron-based data ingestion layer, a dual-schema data storage approach, a processing engine for metric pre-computation, a proactive alerting system, and utilizes the open-source BI tool Metabase for visualization, all secured with role-based access control (RBAC). The implementation resulted in a significant reduction in manual reporting efforts, saving an estimated 20 person-hours per week, and enabled faster, data-driven bottleneck identification. Finally, we evaluate the system's scalability and discuss its trade-offs, positioning it as a valuable contribution to engineering intelligence platforms.
On the Mechanisms of Collaborative Learning in VAE Recommenders
Variational Autoencoders (VAEs) are a powerful alternative to matrix factorization for recommendation. A common technique in VAE-based collaborative filtering (CF) consists in applying binary input masking to user interaction vectors, which improves performance but remains underexplored theoretically. In this work, we analyze how collaboration arises in VAE-based CF and show it is governed by latent proximity: we derive a latent sharing radius that informs when an SGD update on one user strictly reduces the loss on another user, with influence decaying as the latent Wasserstein distance increases. We further study the induced geometry: with clean inputs, VAE-based CF primarily exploits \emph{local} collaboration between input-similar users and under-utilizes global collaboration between far-but-related users. We compare two mechanisms that encourage \emph{global} mixing and characterize their trade-offs: (1) $\beta$-KL regularization directly tightens the information bottleneck, promoting posterior overlap but risking representational collapse if too large; (2) input masking induces stochastic geometric contractions and expansions, which can bring distant users onto the same latent neighborhood but also introduce neighborhood drift. To preserve user identity while enabling global consistency, we propose an anchor regularizer that aligns user posteriors with item embeddings, stabilizing users under masking and facilitating signal sharing across related items. Our analyses are validated on the Netflix, MovieLens-20M, and Million Song datasets. We also successfully deployed our proposed algorithm on an Amazon streaming platform following a successful online experiment.
Make It Long, Keep It Fast: End-to-End 10k-Sequence Modeling at Billion Scale on Douyin
Short-video recommenders such as Douyin must exploit extremely long user histories without breaking latency or cost budgets. We present an end-to-end system that scales long-sequence modeling to 10k-length histories in production. First, we introduce Stacked Target-to-History Cross Attention (STCA), which replaces history self-attention with stacked cross-attention from the target to the history, reducing complexity from quadratic to linear in sequence length and enabling efficient end-to-end training. Second, we propose Request Level Batching (RLB), a user-centric batching scheme that aggregates multiple targets for the same user/request to share the user-side encoding, substantially lowering sequence-related storage, communication, and compute without changing the learning objective. Third, we design a length-extrapolative training strategy -- train on shorter windows, infer on much longer ones -- so the model generalizes to 10k histories without additional training cost. Across offline and online experiments, we observe predictable, monotonic gains as we scale history length and model capacity, mirroring the scaling law behavior observed in large language models. Deployed at full traffic on Douyin, our system delivers significant improvements on key engagement metrics while meeting production latency, demonstrating a practical path to scaling end-to-end long-sequence recommendation to the 10k regime.
Kunlun Anomaly Troubleshooter: Enabling Kernel-Level Anomaly Detection and Causal Reasoning for Large Model Distributed Inference
Anomaly troubleshooting for large model distributed inference (LMDI) remains a critical challenge. Resolving anomalies such as inference performance degradation or latency jitter in distributed system demands significant manual efforts from domain experts, resulting in extremely time-consuming diagnosis processes with relatively low accuracy. In this paper, we introduce Kunlun Anomaly Troubleshooter (KAT), the first anomaly troubleshooting framework tailored for LMDI. KAT addresses this problem through two core innovations. First, KAT exploits the synchronicity and consistency of GPU workers, innovatively leverages function trace data to precisely detect kernel-level anomalies and associated hardware components at nanosecond resolution. Second, KAT integrates these detection results into a domain-adapted LLM, delivering systematic causal reasoning and natural language interpretation of complex anomaly symptoms. Evaluations conducted in Alibaba Cloud Service production environment indicate that KAT achieves over 0.884 precision and 0.936 recall in anomaly detection, providing detail anomaly insights that significantly narrow down the diagnostic scope and improve both the efficiency and success rate of troubleshooting.
When AI Meets the Web: Prompt Injection Risks in Third-Party AI Chatbot Plugins
Prompt injection attacks pose a critical threat to large language models (LLMs), with prior work focusing on cutting-edge LLM applications like personal copilots. In contrast, simpler LLM applications, such as customer service chatbots, are widespread on the web, yet their security posture and exposure to such attacks remain poorly understood. These applications often rely on third-party chatbot plugins that act as intermediaries to commercial LLM APIs, offering non-expert website builders intuitive ways to customize chatbot behaviors. To bridge this gap, we present the first large-scale study of 17 third-party chatbot plugins used by over 10,000 public websites, uncovering previously unknown prompt injection risks in practice. First, 8 of these plugins (used by 8,000 websites) fail to enforce the integrity of the conversation history transmitted in network requests between the website visitor and the chatbot. This oversight amplifies the impact of direct prompt injection attacks by allowing adversaries to forge conversation histories (including fake system messages), boosting their ability to elicit unintended behavior (e.g., code generation) by 3 to 8x. Second, 15 plugins offer tools, such as web-scraping, to enrich the chatbot's context with website-specific content. However, these tools do not distinguish the website's trusted content (e.g., product descriptions) from untrusted, third-party content (e.g., customer reviews), introducing a risk of indirect prompt injection. Notably, we found that ~13% of e-commerce websites have already exposed their chatbots to third-party content. We systematically evaluate both vulnerabilities through controlled experiments grounded in real-world observations, focusing on factors such as system prompt design and the underlying LLM. Our findings show that many plugins adopt insecure practices that undermine the built-in LLM safeguards.