Haoxiang Zhang

Haoxiang Zhang, Qixin Xu, Zhuofeng Li et al.

Long-horizon search agents accumulate large amounts of retrieved content across many tool calls, making context-budget efficiency increasingly important. A minimal intervention is to mask stale observations from the context as the trajectory progresses, but it remains unclear when this form of context management helps and why. We study observation masking through a systematic sweep over various agent backbones (4B to 284B parameters) and three retrievers on offline and live-web agentic search benchmarks. We find that the accuracy gain from masking follows an asymmetric inverted-U shape when plotted against the model's accuracy without context management: a plateau under weak retrievers, a peak when a strong retriever meets a mid-capacity model, and a sharp collapse when the model is saturated. This pattern reflects the interaction between retriever recall and the model's implicit filtering capacity, rather than either factor in isolation. Mechanistically, masking implements a token-for-turn trade-off: it removes observations the model has largely stopped attending to and pages the agent rarely re-opens. The added turns help when they convert failures into successes, but they fail when masking removes evidence the model would otherwise have used. We therefore reframe context management as a regime-dependent intervention and provide a holistic perspective for analyzing context use in agentic deep search. We release our scaffold and trajectories here (https://github.com/i-DeepSearch/observation-masking) to support future research.

Zhuofeng Li, Yi Lu, Dongfu Jiang et al. · utoronto

The rapid rise in AI conference submissions has driven increasing exploration of large language models (LLMs) for peer review support. However, LLM-based reviewers often generate superficial, formulaic comments lacking substantive, evidence-grounded feedback. We attribute this to the underutilization of two key components of human reviewing: explicit rubrics and contextual grounding in existing work. To address this, we introduce REVIEWBENCH, a benchmark evaluating review text according to paper-specific rubrics derived from official guidelines, the paper's content, and human-written reviews. We further propose REVIEWGROUNDER, a rubric-guided, tool-integrated multi-agent framework that decomposes reviewing into drafting and grounding stages, enriching shallow drafts via targeted evidence consolidation. Experiments on REVIEWBENCH show that REVIEWGROUNDER, using a Phi-4-14B-based drafter and a GPT-OSS-120B-based grounding stage, consistently outperforms baselines with substantially stronger/larger backbones (e.g., GPT-4.1 and DeepSeek-R1-670B) in both alignment with human judgments and rubric-based review quality across 8 dimensions. The code is available \href{https://github.com/EigenTom/ReviewGrounder}{here}.

Zirong Chen, Isaac Li, Haoxiang Zhang et al.

An increasing number of monitoring systems have been developed in smart cities to ensure that real-time operations of a city satisfy safety and performance requirements. However, many existing city requirements are written in English with missing, inaccurate, or ambiguous information. There is a high demand for assisting city policy makers in converting human-specified requirements to machine-understandable formal specifications for monitoring systems. To tackle this limitation, we build CitySpec, the first intelligent assistant system for requirement specification in smart cities. To create CitySpec, we first collect over 1,500 real-world city requirements across different domains from over 100 cities and extract city-specific knowledge to generate a dataset of city vocabulary with 3,061 words. We also build a translation model and enhance it through requirement synthesis and develop a novel online learning framework with validation under uncertainty. The evaluation results on real-world city requirements show that CitySpec increases the sentence-level accuracy of requirement specification from 59.02% to 86.64%, and has strong adaptability to a new city and a new domain (e.g., F1 score for requirements in Seattle increases from 77.6% to 93.75% with online learning).

Zirong Chen, Issa Li, Haoxiang Zhang et al.

An increasing number of monitoring systems have been developed in smart cities to ensure that the real-time operations of a city satisfy safety and performance requirements. However, many existing city requirements are written in English with missing, inaccurate, or ambiguous information. There is a high demand for assisting city policymakers in converting human-specified requirements to machine-understandable formal specifications for monitoring systems. To tackle this limitation, we build CitySpec, the first intelligent assistant system for requirement specification in smart cities. To create CitySpec, we first collect over 1,500 real-world city requirements across different domains (e.g., transportation and energy) from over 100 cities and extract city-specific knowledge to generate a dataset of city vocabulary with 3,061 words. We also build a translation model and enhance it through requirement synthesis and develop a novel online learning framework with shielded validation. The evaluation results on real-world city requirements show that CitySpec increases the sentence-level accuracy of requirement specification from 59.02% to 86.64%, and has strong adaptability to a new city and a new domain (e.g., the F1 score for requirements in Seattle increases from 77.6% to 93.75% with online learning). After the enhancement from the shield function, CitySpec is now immune to most known textual adversarial inputs (e.g., the attack success rate of DeepWordBug after the shield function is reduced to 0% from 82.73%). We test the CitySpec with 18 participants from different domains. CitySpec shows its strong usability and adaptability to different domains, and also its robustness to malicious inputs.

Zhuofeng Li, Dongfu Jiang, Xueguang Ma et al.

Training deep research agents requires long-horizon trajectories that interleave search, evidence aggregation, and multi-step reasoning. However, existing data collection pipelines typically rely on proprietary web APIs, making large-scale trajectory synthesis costly, unstable, and difficult to reproduce. We present OpenResearcher, a reproducible pipeline that decouples one-time corpus bootstrapping from multi-turn trajectory synthesis and executes the search-and-browse loop entirely offline using three explicit browser primitives: search, open, and find, over a 15M-document corpus. Using GPT-OSS-120B as the teacher model, we synthesize over 97K trajectories, including a substantial long-horizon tail with 100+ tool calls. Supervised fine-tuning a 30B-A3B backbone on these trajectories achieves 54.8\% accuracy on BrowseComp-Plus, a +34.0 point improvement over the base model, while remaining competitive on BrowseComp, GAIA, and xbench-DeepSearch. Because the environment is offline and fully instrumented, it also enables controlled analysis, where our study reveals practical insights into deep research pipeline design, including data filtering strategies, agent configuration choices, and how retrieval success relates to final answer accuracy. We release the pipeline, synthesized trajectories, model checkpoints, and the offline search environment at https://github.com/TIGER-AI-Lab/OpenResearcher.

Zirong Chen, Isaac Li, Haoxiang Zhang et al.

As more and more monitoring systems have been deployed to smart cities, there comes a higher demand for converting new human-specified requirements to machine-understandable formal specifications automatically. However, these human-specific requirements are often written in English and bring missing, inaccurate, or ambiguous information. In this paper, we present CitySpec, an intelligent assistant system for requirement specification in smart cities. CitySpec not only helps overcome the language differences brought by English requirements and formal specifications, but also offers solutions to those missing, inaccurate, or ambiguous information. The goal of this paper is to demonstrate how CitySpec works. Specifically, we present three demos: (1) interactive completion of requirements in CitySpec; (2) human-in-the-loop correction while CitySepc encounters exceptions; (3) online learning in CitySpec.

Haoxiang Zhang, He Jiang, Ziqiang Wang et al.

Zero-Shot Sketch-Based Image Retrieval (ZSSBIR) is an emerging task. The pioneering work focused on the modal gap but ignored inter-class information. Although recent work has begun to consider the triplet-based or contrast-based loss to mine inter-class information, positive and negative samples need to be carefully selected, or the model is prone to lose modality-specific information. To respond to these issues, an Ontology-Aware Network (OAN) is proposed. Specifically, the smooth inter-class independence learning mechanism is put forward to maintain inter-class peculiarity. Meanwhile, distillation-based consistency preservation is utilized to keep modality-specific information. Extensive experiments have demonstrated the superior performance of our algorithm on two challenging Sketchy and Tu-Berlin datasets.

Haoxiang Zhang, Juliana Freire, Yash Garg

Recent advancements in software and hardware technologies have enabled the use of AI/ML models in everyday applications has significantly improved the quality of service rendered. However, for a given application, finding the right AI/ML model is a complex and costly process, that involves the generation, training, and evaluation of multiple interlinked steps (called pipelines), such as data pre-processing, feature engineering, selection, and model tuning. These pipelines are complex (in structure) and costly (both in compute resource and time) to execute end-to-end, with a hyper-parameter associated with each step. AutoML systems automate the search of these hyper-parameters but are slow, as they rely on optimizing the pipeline's end output. We propose the eTOP Framework which works on top of any AutoML system and decides whether or not to execute the pipeline to the end or terminate at an intermediate step. Experimental evaluation on 26 benchmark datasets and integration of eTOPwith MLBox4 reduces the training time of the AutoML system upto 40x than baseline MLBox.

Zhuofeng Li, Haoxiang Zhang, Cong Wei et al.

Modern retrieval systems, whether lexical or semantic, expose a corpus through a fixed similarity interface that compresses access into a single top-k retrieval step before reasoning. This abstraction is efficient, but for agentic search, it becomes a bottleneck: exact lexical constraints, sparse clue conjunctions, local context checks, and multi-step hypothesis refinement are difficult to implement by calling a conventional off-the-shelf retriever, and evidence filtered out early cannot be recovered by stronger downstream reasoning. Agentic tasks further exacerbate this limitation because they require agents to orchestrate multiple steps, including discovering intermediate entities, combining weak clues, and revising the plan after observing partial evidence. To tackle the limitation, we study direct corpus interaction (DCI), where an agent searches the raw corpus directly with general-purpose terminal tools (e.g., grep, file reads, shell commands, lightweight scripts), without any embedding model, vector index, or retrieval API. This approach requires no offline indexing and adapts naturally to evolving local corpora. Across IR benchmarks and end-to-end agentic search tasks, this simple setup substantially outperforms strong sparse, dense, and reranking baselines on several BRIGHT and BEIR datasets, and attains strong accuracy on BrowseComp-Plus and multi-hop QA without relying on any conventional semantic retriever. Our results indicate that as language agents become stronger, retrieval quality depends not only on reasoning ability but also on the resolution of the interface through which the model interacts with the corpus, with which DCI opens a broader interface-design space for agentic search.

CocoaBench Team, Shibo Hao, Zhining Zhang et al.

LLM agents now perform strongly in software engineering, deep research, GUI automation, and various other applications, while recent agent scaffolds and models are increasingly integrating these capabilities into unified systems. Yet, most evaluations still test these capabilities in isolation, which leaves a gap for more diverse use cases that require agents to combine different capabilities. We introduce CocoaBench, a benchmark for unified digital agents built from human-designed, long-horizon tasks that require flexible composition of vision, search, and coding. Tasks are specified only by an instruction and an automatic evaluation function over the final output, enabling reliable and scalable evaluation across diverse agent infrastructures. We also present CocoaAgent, a lightweight shared scaffold for controlled comparison across model backbones. Experiments show that current agents remain far from reliable on CocoaBench, with the best evaluated system achieving only 45.1% success rate. Our analysis further points to substantial room for improvement in reasoning and planning, tool use and execution, and visual grounding.

Changzhu Liu, Ruisi He, Haoxiang Zhang et al.

Reconfigurable intelligent surface (RIS) has recently been gained attention as an effective technique improving the coverage and performance of communication systems by creating additional communication links. Deployment of RIS is crucial for overcoming signal coverage limitations, especially in high-speed train (HST) scenarios. Considerable research has been performed assuming perfect channel state information (CSI). However, due to the rapidly time-varying fading channels and feedback delays, achieving perfect CSI at the base station (BS) is not feasible in the HST scenarios. To tackle this problem, this paper investigates a robust design strategy for RIS-aided HST communication coverage enhancement, particularly focusing on cascaded BS-RIS-user channels at BS (CBRUB). The study explores the optimization problem under two types distinct of models: centered on minimizing transmit power subject to worst-case rate constraints within the bounded CSI error (BCSIE) model, and the other focusing on outage probability (OP) constraints under the statistical CSI error (SCSIE) model. We use the S-procedure to approximate the non-convex (NC) constraints, converting the worst-case rate constraints into linear matrix inequalities. Additionally, the Bernstein-type inequality is applied to transform the OP constraints into second-order cone constraints and linear inequalities. The simulation analysis results show that CBRUB errors have a significant effect on system performance compared to direct CSI errors.

Hao Jiang, Haoxiang Zhang, Qingshan Hou et al.

Providing high-quality item recall for text queries is crucial in large-scale e-commerce search systems. Current Embedding-based Retrieval Systems (ERS) embed queries and items into a shared low-dimensional space, but uni-modality ERS rely too heavily on textual features, making them unreliable in complex contexts. While multi-modality ERS incorporate various data sources, they often overlook individual preferences for different modalities, leading to suboptimal results. To address these issues, we propose MRSE, a Multi-modality Retrieval System that integrates text, item images, and user preferences through lightweight mixture-of-expert (LMoE) modules to better align features across and within modalities. MRSE also builds user profiles at a multi-modality level and introduces a novel hybrid loss function that enhances consistency and robustness using hard negative sampling. Experiments on a large-scale dataset from Shopee and online A/B testing show that MRSE achieves an 18.9% improvement in offline relevance and a 3.7% gain in online core metrics compared to Shopee's state-of-the-art uni-modality system.

Naye Ji, Fan Zhang, Haoxiang Zhang et al.

Mixed reality applications often require virtual objects that are partly occluded by real objects. However, previous research and commercial products have limitations in terms of performance and efficiency. To address these challenges, we propose a novel depth contour occlusion (DCO) algorithm. The proposed method is based on the sensitivity of contour occlusion and a binocular stereoscopic vision device. In this method, a depth contour map is combined with a sparse depth map obtained from a two-stage adaptive filter area stereo matching algorithm and the depth contour information of the objects extracted by a digital image stabilisation optical flow method. We also propose a quadratic optimisation model with three constraints to generate an accurate dense map of the depth contour for high-quality real-virtual occlusion. The whole process is accelerated by GPU. To evaluate the effectiveness of the algorithm, we demonstrate a time con-sumption statistical analysis for each stage of the DCO algorithm execution. To verify the relia-bility of the real-virtual occlusion effect, we conduct an experimental analysis on single-sided, enclosed, and complex occlusions; subsequently, we compare it with the occlusion method without quadratic optimisation. With our GPU implementation for real-time DCO, the evaluation indicates that applying the presented DCO algorithm can enhance the real-time performance and the visual quality of real-virtual occlusion.

Gustavo A. Oliva, Gopi Krishnan Rajbahadur, Aaditya Bhatia et al.

High-quality labeled datasets are crucial for training and evaluating foundation models in software engineering, but creating them is often prohibitively expensive and labor-intensive. We introduce SPICE, a scalable, automated pipeline for labeling SWE-bench-style datasets with annotations for issue clarity, test coverage, and effort estimation. SPICE combines context-aware code navigation, rationale-driven prompting, and multi-pass consensus to produce labels that closely approximate expert annotations. SPICE's design was informed by our own experience and frustration in labeling more than 800 instances from SWE-Gym. SPICE achieves strong agreement with human-labeled SWE-bench Verified data while reducing the cost of labeling 1,000 instances from around \$100,000 (manual annotation) to just \$5.10. These results demonstrate SPICE's potential to enable cost-effective, large-scale dataset creation for SE-focused FMs. To support the community, we release both SPICE tool and SPICE Bench, a new dataset of 6,802 SPICE-labeled instances curated from 291 open-source projects in SWE-Gym (over 13x larger than SWE-bench Verified).

He Jiang, Mujtaba Asad, Jingjing Liu et al.

Traditional image detail enhancement is local filter-based or global filter-based. In both approaches, the original image is first divided into the base layer and the detail layer, and then the enhanced image is obtained by amplifying the detail layer. Our method is different, and its innovation lies in the special way to get the image detail layer. The detail layer in our method is obtained by updating the residual features, and the updating mechanism is usually based on searching and matching similar patches. However, due to the diversity of image texture features, perfect matching is often not possible. In this paper, the process of searching and matching is treated as a thermodynamic process, where the Metropolis theorem can minimize the internal energy and get the global optimal solution of this task, that is, to find a more suitable feature for a better detail enhancement performance. Extensive experiments have proven that our algorithm can achieve better results in quantitative metrics testing and visual effects evaluation. The source code can be obtained from the link.

Hao Li, Haoxiang Zhang, Ahmed E. Hassan

The future of software engineering--SE 3.0--is unfolding with the rise of AI teammates: autonomous, goal-driven systems collaborating with human developers. Among these, autonomous coding agents are especially transformative, now actively initiating, reviewing, and evolving code at scale. This paper introduces AIDev, the first large-scale dataset capturing how such agents operate in the wild. Spanning over 456,000 pull requests by five leading agents--OpenAI Codex, Devin, GitHub Copilot, Cursor, and Claude Code--across 61,000 repositories and 47,000 developers, AIDev provides an unprecedented empirical foundation for studying autonomous teammates in software development. Unlike prior work that has largely theorized the rise of AI-native software engineering, AIDev offers structured, open data to support research in benchmarking, agent readiness, optimization, collaboration modeling, and AI governance. The dataset includes rich metadata on PRs, authorship, review timelines, code changes, and integration outcomes--enabling exploration beyond synthetic benchmarks like SWE-bench. For instance, although agents often outperform humans in speed, their PRs are accepted less frequently, revealing a trust and utility gap. Furthermore, while agents accelerate code submission--one developer submitted as many PRs in three days as they had in three years--these are structurally simpler (via code complexity metrics). We envision AIDev as a living resource: extensible, analyzable, and ready for the SE and AI communities. Grounding SE 3.0 in real-world evidence, AIDev enables a new generation of research into AI-native workflows and supports building the next wave of symbiotic human-AI collaboration. The dataset is publicly available at https://github.com/SAILResearch/AI_Teammates_in_SE3. > AI Agent, Agentic AI, Coding Agent, Agentic Coding, Software Engineering Agent

Ankur Tagra, Haoxiang Zhang, Gopi Krishnan Rajbahadur et al.

Reopened bugs can degrade the overall quality of a software system since they require unnecessary rework by developers. Moreover, reopened bugs also lead to a loss of trust in the end-users regarding the quality of the software. Thus, predicting bugs that might be reopened could be extremely helpful for software developers to avoid rework. Prior studies on reopened bug prediction focus only on three open source projects (i.e., Apache, Eclipse, and OpenOffice) to generate insights. We observe that one out of the three projects (i.e., Apache) has a data leak issue -- the bug status of reopened was included as training data to predict reopened bugs. In addition, prior studies used an outdated prediction model pipeline (i.e., with old techniques for constructing a prediction model) to predict reopened bugs. Therefore, we revisit the reopened bugs study on a large scale dataset consisting of 47 projects tracked by JIRA using the modern techniques such as SMOTE, permutation importance together with 7 different machine learning models. We study the reopened bugs using a mixed methods approach (i.e., both quantitative and qualitative study). We find that: 1) After using an updated reopened bug prediction model pipeline, only 34% projects give an acceptable performance with AUC >= 0.7. 2) There are four major reasons for a bug getting reopened, that is, technical (i.e., patch/integration issues), documentation, human (i.e., due to incorrect bug assessment), and reasons not shown in the bug reports. 3) In projects with an acceptable AUC, 94% of the reopened bugs are due to patch issues (i.e., the usage of an incorrect patch) identified before bug reopening. Our study revisits reopened bugs and provides new insights into developer's bug reopening activities.

Ahmed E. Hassan, Dayi Lin, Gopi Krishnan Rajbahadur et al.

Foundation models (FMs), such as Large Language Models (LLMs), have revolutionized software development by enabling new use cases and business models. We refer to software built using FMs as FMware. The unique properties of FMware (e.g., prompts, agents, and the need for orchestration), coupled with the intrinsic limitations of FMs (e.g., hallucination) lead to a completely new set of software engineering challenges. Based on our industrial experience, we identified 10 key SE4FMware challenges that have caused enterprise FMware development to be unproductive, costly, and risky. In this paper, we discuss these challenges in detail and state the path for innovation that we envision. Next, we present FMArts, which is our long-term effort towards creating a cradle-to-grave platform for the engineering of trustworthy FMware. Finally, we (i) show how the unique properties of FMArts enabled us to design and develop a complex FMware for a large customer in a timely manner and (ii) discuss the lessons that we learned in doing so. We hope that the disclosure of the aforementioned challenges and our associated efforts to tackle them will not only raise awareness but also promote deeper and further discussions, knowledge sharing, and innovative solutions across the software engineering discipline.

Shahla Shaan Ahmed, Shaowei Wang, Yuan Tian et al.

Context: Navigating the knowledge of Stack Overflow (SO) remains challenging. To make the posts vivid to users, SO allows users to write and edit posts with Markdown or HTML so that users can leverage various formatting styles (e.g., bold, italic, and code) to highlight the important information. Nonetheless, there have been limited studies on the highlighted information. Objective: We carried out the first large-scale exploratory study on the information highlighted in SO answers in our recent study. To extend our previous study, we develop approaches to automatically recommend highlighted content with formatting styles using neural network architectures initially designed for the Named Entity Recognition task. Method: In this paper, we studied 31,169,429 answers of Stack Overflow. For training recommendation models, we choose CNN-based and BERT-based models for each type of formatting (i.e., Bold, Italic, Code, and Heading) using the information highlighting dataset we collected from SO answers. Results: Our models achieve a precision ranging from 0.50 to 0.72 for different formatting types. It is easier to build a model to recommend Code than other types. Models for text formatting types (i.e., Heading, Bold, and Italic) suffer low recall. Our analysis of failure cases indicates that the majority of the failure cases are due to missing identification. One explanation is that the models are easy to learn the frequent highlighted words while struggling to learn less frequent words (i.g., long-tail knowledge). Conclusion: Our findings suggest that it is possible to develop recommendation models for highlighting information for answers with different formatting styles on Stack Overflow.

Zhuofeng Li, Haoxiang Zhang, Seungju Han et al. · stanford

Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a single, monolithic policy that interleaves thoughts and tool calls under full context; this scales poorly with long horizons and diverse tools and generalizes weakly to new scenarios. Agentic systems offer a promising alternative by decomposing work across specialized modules, yet most remain training-free or rely on offline training decoupled from the live dynamics of multi-turn interaction. We introduce AgentFlow, a trainable, in-the-flow agentic framework that coordinates four modules (planner, executor, verifier, generator) through an evolving memory and directly optimizes its planner inside the multi-turn loop. To train on-policy in live environments, we propose Flow-based Group Refined Policy Optimization (Flow-GRPO), which tackles long-horizon, sparse-reward credit assignment by converting multi-turn optimization into a sequence of tractable single-turn policy updates. It broadcasts a single, verifiable trajectory-level outcome to every turn to align local planner decisions with global success and stabilizes learning with group-normalized advantages. Across ten benchmarks, AgentFlow with a 7B-scale backbone outperforms top-performing baselines with average accuracy gains of 14.9% on search, 14.0% on agentic, 14.5% on mathematical, and 4.1% on scientific tasks, even surpassing larger proprietary models like GPT-4o. Further analyses confirm the benefits of in-the-flow optimization, showing improved planning, enhanced tool-calling reliability, and positive scaling with model size and reasoning turns.

Haoxiang Zhang, Shi Chang, Arthur Leung et al.

The rise of Foundation Models (FMs) like Large Language Models (LLMs) is revolutionizing software development. Despite the impressive prototypes, transforming FMware into production-ready products demands complex engineering across various domains. A critical but overlooked aspect is performance engineering, which aims at ensuring FMware meets performance goals such as throughput and latency to avoid user dissatisfaction and financial loss. Often, performance considerations are an afterthought, leading to costly optimization efforts post-deployment. FMware's high computational resource demands highlight the need for efficient hardware use. Continuous performance engineering is essential to prevent degradation. This paper highlights the significance of Software Performance Engineering (SPE) in FMware, identifying four key challenges: cognitive architecture design, communication protocols, tuning and optimization, and deployment. These challenges are based on literature surveys and experiences from developing an in-house FMware system. We discuss problems, current practices, and innovative paths for the software engineering community.

Hao Li, Haoxiang Zhang, Ahmed E. Hassan

AI coding agents are rapidly transforming software engineering by performing tasks such as feature development, debugging, and testing. Despite their growing impact, the research community lacks a comprehensive dataset capturing how these agents are used in real-world projects. To address this gap, we introduce AIDev, a large-scale dataset focused on agent-authored pull requests (Agentic-PRs) in real-world GitHub repositories. AIDev aggregates 932,791 Agentic-PRs produced by five agents: OpenAI Codex, Devin, GitHub Copilot, Cursor, and Claude Code. These PRs span 116,211 repositories and involve 72,189 developers. In addition, AIDev includes a curated subset of 33,596 Agentic-PRs from 2,807 repositories with over 100 stars, providing further information such as comments, reviews, commits, and related issues. This dataset offers a foundation for future research on AI adoption, developer productivity, and human-AI collaboration in the new era of software engineering. > AI Agent, Agentic AI, Coding Agent, Agentic Coding, Agentic Software Engineering, Agentic Engineering

Haoxiang Zhang, Ruihao Yuan, Lihui Zhang et al.

The industrial adoption of Artificial Intelligence for Engineering (AI4E) faces two fundamental bottlenecks: scarce high-quality data and the lack of interpretability in black-box models-particularly critical in safety-sensitive sectors like aerospace. We present an explainable, few-shot AI4E framework that is systematically informed by physics and expert knowledge throughout its architecture. Starting from only 32 experimental samples in an aerial K439B superalloy castings repair welding case, we first augment physically plausible synthetic data through a three-stage protocol: differentiated noise injection calibrated to process variabilities, enforcement of hard physical constraints, and preservation of inter-parameter relationships. We then employ a nested optimization strategy for constitutive model discovery, where symbolic regression explores equation structures while differential evolution optimizes parameters, followed by intensive parameter refinement using hybrid global-local optimization. The resulting interpretable constitutive equation achieves 88% accuracy in predicting hot-cracking tendency. This equation not only provides quantitative predictions but also delivers explicit physical insight, revealing how thermal, geometric, and metallurgical mechanisms couple to drive cracking-thereby advancing engineers' cognitive understanding of the process. Furthermore, the constitutive equation serves as a multi-functional tool for process optimization and high-fidelity virtual data generation, enabling accuracy improvements in other data-driven models. Our approach provides a general blueprint for developing trustworthy AI systems that embed engineering domain knowledge directly into their architecture, enabling reliable adoption in high-stakes industrial applications where data is limited but physical understanding is available.

Zhijie Lin, Zhuofeng Li, Chenglei Dai et al.

In a multi-stage recommendation system, reranking plays a crucial role in modeling intra-list correlations among items. A key challenge lies in exploring optimal sequences within the combinatorial space of permutations. Recent research follows a two-stage (generator-evaluator) paradigm, where a generator produces multiple feasible sequences, and an evaluator selects the best one. In practice, the generator is typically implemented as an autoregressive model. However, these two-stage methods face two main challenges. First, the separation of the generator and evaluator hinders end-to-end training. Second, autoregressive generators suffer from inference efficiency. In this work, we propose a Unified Generative Efficient Reranking Framework (GReF) to address the two primary challenges. Specifically, we introduce Gen-Reranker, an autoregressive generator featuring a bidirectional encoder and a dynamic autoregressive decoder to generate causal reranking sequences. Subsequently, we pre-train Gen-Reranker on the item exposure order for high-quality parameter initialization. To eliminate the need for the evaluator while integrating sequence-level evaluation during training for end-to-end optimization, we propose post-training the model through Rerank-DPO. Moreover, for efficient autoregressive inference, we introduce ordered multi-token prediction (OMTP), which trains Gen-Reranker to simultaneously generate multiple future items while preserving their order, ensuring practical deployment in real-time recommender systems. Extensive offline experiments demonstrate that GReF outperforms state-of-the-art reranking methods while achieving latency that is nearly comparable to non-autoregressive models. Additionally, GReF has also been deployed in a real-world video app Kuaishou with over 300 million daily active users, significantly improving online recommendation quality.

Kirill Vasilevski, Benjamin Rombaut, Gopi Krishnan Rajbahadur et al.

Foundation Models (FMs) such as Large Language Models (LLMs) are reshaping the software industry by enabling FMware, systems that integrate these FMs as core components. In this KDD 2025 tutorial, we present a comprehensive exploration of FMware that combines a curated catalogue of challenges with real-world production concerns. We first discuss the state of research and practice in building FMware. We further examine the difficulties in selecting suitable models, aligning high-quality domain-specific data, engineering robust prompts, and orchestrating autonomous agents. We then address the complex journey from impressive demos to production-ready systems by outlining issues in system testing, optimization, deployment, and integration with legacy software. Drawing on our industrial experience and recent research in the area, we provide actionable insights and a technology roadmap for overcoming these challenges. Attendees will gain practical strategies to enable the creation of trustworthy FMware in the evolving technology landscape.

Benjamin T. Brown, Haoxiang Zhang, Daniel L. Lau et al.

In graph signal processing, learning the weighted connections between nodes from a set of sample signals is a fundamental task when the underlying relationships are not known a priori. This task is typically addressed by finding a graph Laplacian on which the observed signals are smooth. With the extension of graphs to hypergraphs - where edges can connect more than two nodes - graph learning methods have similarly been generalized to hypergraphs. However, the absence of a unified framework for calculating total variation has led to divergent definitions of smoothness and, consequently, differing approaches to hyperedge recovery. We confront this challenge through generalization of several previously proposed hypergraph total variations, subsequently allowing ease of substitution into a vector based optimization. To this end, we propose a novel hypergraph learning method that recovers a hypergraph topology from time-series signals based on a smoothness prior. Our approach, designated as Hypergraph Structure Learning with Smoothness (HSLS), addresses key limitations in prior works, such as hyperedge selection and convergence issues, by formulating the problem as a convex optimization solved via a forward-backward-forward algorithm, ensuring guaranteed convergence. Additionally, we introduce a process that simultaneously limits the span of the hyperedge search and maintains a valid hyperedge selection set. In doing so, our method becomes scalable in increasingly complex network structures. The experimental results demonstrate improved performance, in terms of accuracy, over other state-of-the-art hypergraph inference methods; furthermore, we empirically show our method to be robust to total variation terms, biased towards global smoothness, and scalable to larger hypergraphs.

Keying Guo, Ruisi He, Mi Yang et al.

Time-varying non-stationary channels, with complex dynamic variations and temporal evolution characteristics, have significant challenges in channel modeling and communication system performance evaluation. Most existing methods of time-varying channel modeling focus on predicting channel state at a given moment or simulating short-term channel fluctuations, which are unable to capture the long-term evolution of the channel. This paper emphasizes the generation of long-term dynamic channel to fully capture evolution of non-stationary channel properties. The generated channel not only reflects temporal dynamics but also ensures consistent stationarity. We propose a hybrid deep learning framework that combines conditional generative adversarial networks (CGAN) with long short-term memory (LSTM) networks. A stationarity-constrained approach is designed to ensure temporal correlation of the generated time-series channel. This method can generate channel with required temporal non-stationarity. The model is validated by comparing channel statistical features, and the results show that the generated channel is in good agreement with raw channel and provides good performance in terms of non-stationarity.

Wenhan Zhu, Haoxiang Zhang, Ahmed E. Hassan et al.

Stack Overflow provides a means for developers to exchange knowledge. While much previous research on Stack Overflow has focused on questions and answers (Q&A), recent work has shown that discussions in comments also contain rich information. On Stack Overflow, discussions through comments and chat rooms can be tied to questions or answers. In this paper, we conduct an empirical study that focuses on the nature of question discussions. We observe that: (1) Question discussions occur at all phases of the Q&A process, with most beginning before the first answer is received. (2) Both askers and answerers actively participate in question discussions; the likelihood of their participation increases as the number of comments increases. (3) There is a strong correlation between the number of question comments and the question answering time (i.e., more discussed questions receive answers more slowly); also, questions with a small number of comments are likely to be answered more quickly than questions with no discussion. Our findings suggest that question discussions contain a rich trove of data that is integral to the Q&A processes on Stack Overflow. We further suggest how future research can leverage the information in question discussions, along with the commonly studied Q&A information.

Jiakun Liu, Haoxiang Zhang, Xin Xia et al.

On Stack Overflow, users reuse 11,926,354 external links to share the resources hosted outside the Stack Overflow website. The external links connect to the existing programming-related knowledge and extend the crowdsourced knowledge on Stack Overflow. Some of the external links, so-called as repeated external links, can be shared for multiple times. We observe that 82.5% of the link sharing activities (i.e., sharing links in any question, answer, or comment) on Stack Overflow share external resources, and 57.0% of the occurrences of the external links are sharing the repeated external links. However, it is still unclear what types of external resources are repeatedly shared. To help users manage their knowledge, we wish to investigate the characteristics of the repeated external links in knowledge sharing on Stack Overflow. In this paper, we analyze the repeated external links on Stack Overflow. We observe that external links that point to the text resources (hosted in documentation websites, tutorial websites, etc.) are repeatedly shared the most. We observe that: 1) different users repeatedly share the same knowledge in the form of repeated external links, thus increasing the maintenance effort of knowledge (e.g., update invalid links in multiple posts), 2) the same users can repeatedly share the external links for the purpose of promotion, and 3) external links can point to webpages with an overload of information that is difficult for users to retrieve relevant information. Our findings provide insights to Stack Overflow moderators and researchers. For example, we encourage Stack Overflow to centrally manage the commonly occurring knowledge in the form of repeated external links in order to better maintain the crowdsourced knowledge on Stack Overflow.

Jiakun Liu, Xin Xia, David Lo et al.

Stack Overflow hosts valuable programming-related knowledge with 11,926,354 links that reference to the third-party websites. The links that reference to the resources hosted outside the Stack Overflow websites extend the Stack Overflow knowledge base substantially. However, with the rapid development of programming-related knowledge, many resources hosted on the Internet are not available anymore. Based on our analysis of the Stack Overflow data that was released on Jun. 2, 2019, 14.2% of the links on Stack Overflow are broken links. The broken links on Stack Overflow can obstruct viewers from obtaining desired programming-related knowledge, and potentially damage the reputation of the Stack Overflow as viewers might regard the posts with broken links as obsolete. In this paper, we characterize the broken links on Stack Overflow. 65% of the broken links in our sampled questions are used to show examples, e.g., code examples. 70% of the broken links in our sampled answers are used to provide supporting information, e.g., explaining a certain concept and describing a step to solve a problem. Only 1.67% of the posts with broken links are highlighted as such by viewers in the posts' comments. Only 5.8% of the posts with broken links removed the broken links. Viewers cannot fully rely on the vote scores to detect broken links, as broken links are common across posts with different vote scores. The websites that host resources that can be maintained by their users are referenced by broken links the most on Stack Overflow -- a prominent example of such websites is GitHub. The posts and comments related to the web technologies, i.e., JavaScript, HTML, CSS, and jQuery, are associated with more broken links. Based on our findings, we shed lights for future directions and provide recommendations for practitioners and researchers.

Chen Xia, Haoxiang Zhang, Jacob Moghtader et al.

There are tons of news articles generated every day reflecting the activities of key roles such as people, organizations and political parties. Analyzing these key roles allows us to understand the trends in news. In this paper, we present a demonstration system that visualizes the trend of key roles in news articles based on natural language processing techniques. Specifically, we apply a semantic role labeler and the dynamic word embedding technique to understand relationships between key roles in the news across different time periods and visualize the trends of key role and news topics change over time.

Haoxiang Zhang, Shaowei Wang, Tse-Hsun Peter Chen et al.

Stack Overflow has accumulated millions of answers. Informative comments can strengthen their associated answers (e.g., providing additional information). Currently, Stack Overflow hides comments that are ranked beyond the top 5. Stack Overflow aims to display more informative comments (i.e., the ones with higher scores) and hide less informative ones using this mechanism. As a result, 4.4 million comments are hidden under their answer threads. Therefore, it is very important to understand how well the current comment hiding mechanism works. In this study, we investigate whether the mechanism can effectively deliver informative comments while hiding uninformative comments. We find that: 1) Hidden comments are as informative as displayed comments; more than half of the comments (both hidden and displayed) are informative (e.g., providing alternative answers, or pointing out flaws in their associated answers). 2) The current comment hiding mechanism tends to rank and hide comments based on their creation time instead of their score in most cases due to the large amount of tie-scored comments (e.g., 87% of the comments have 0-score). 3) In 97.3% of answers that have hidden comments, at least one comment is hidden while there is another comment with the same score is displayed (i.e., we refer to such cases as unfairly hidden comments). Among such unfairly hidden comments, the longest unfairly hidden comment is more likely to be informative than the shortest unfairly displayed comments. Our findings suggest that Stack Overflow should consider adjusting their current comment hiding mechanism, e.g., displaying longer unfairly hidden comments to replace shorter unfairly displayed comments. We also recommend that users examine all comments, in case they would miss informative details such as software obsolescence, code error reports, or notices of security vulnerability in hidden comments.

Haoxiang Zhang, Shaowei Wang, Tse-Hsun et al.

Stack Overflow accumulates an enormous amount of software engineering knowledge. However, as time passes, certain knowledge in answers may become obsolete. Such obsolete answers, if not identified or documented clearly, may mislead answer seekers and cause unexpected problems (e.g., using an out-dated security protocol). In this paper, we investigate how the knowledge in answers becomes obsolete and identify the characteristics of such obsolete answers. We find that: 1) More than half of the obsolete answers (58.4%) were probably already obsolete when they were first posted. 2) When an obsolete answer is observed, only a small proportion (20.5%) of such answers are ever updated. 3) Answers to questions in certain tags (e.g., node.js, ajax, android, and objective-c) are more likely to become obsolete. Our findings suggest that Stack Overflow should develop mechanisms to encourage the whole community to maintain answers (to avoid obsolete answers) and answer seekers are encouraged to carefully go through all information (e.g., comments) in answer threads.