arXiv subjects

Dongmei Zhang

Publications and source records attributed to Dongmei Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

CI-Work: Benchmarking Contextual Integrity in Enterprise LLM Agents

Enterprise LLM agents can dramatically improve workplace productivity, but their core capability, retrieving and using internal context to act on a user's behalf, also creates new risks for sensitive information leakage. We introduce CI-Work, a Contextual Integrity (CI)-grounded benchmark that simulates enterprise workflows across five information-flow directions and evaluates whether agents can convey essential content while withholding sensitive context in dense retrieval settings. Our evaluation of frontier models reveals that privacy failures are prevalent (violation rates range from 15.8%-50.9%, with leakage reaching up to 26.7%) and uncovers a counterintuitive trade-off critical for industrial deployment: higher task utility often correlates with increased privacy violations. Moreover, the massive scale of enterprise data and potential user behavior further amplify this vulnerability. Simply increasing model size or reasoning depth fails to address the problem. We conclude that safeguarding enterprise workflows requires a paradigm shift, moving beyond model-centric scaling toward context-centric architectures.

cs.CR↗

arXiv2026

CAST: Achieving Stable LLM-based Text Analysis for Data Analytics

Text analysis of tabular data relies on two core operations: \emph{summarization} for corpus-level theme extraction and \emph{tagging} for row-level labeling. A critical limitation of employing large language models (LLMs) for these tasks is their inability to meet the high standards of output stability demanded by data analytics. To address this challenge, we introduce \textbf{CAST} (\textbf{C}onsistency via \textbf{A}lgorithmic Prompting and \textbf{S}table \textbf{T}hinking), a framework that enhances output stability by constraining the model's latent reasoning path. CAST combines (i) Algorithmic Prompting to impose a procedural scaffold over valid reasoning transitions and (ii) Thinking-before-Speaking to enforce explicit intermediate commitments before final generation. To measure progress, we introduce \textbf{CAST-S} and \textbf{CAST-T}, stability metrics for bulleted summarization and tagging, and validate their alignment with human judgments. Experiments across publicly available benchmarks on multiple LLM backbones show that CAST consistently achieves the best stability among all baselines, improving Stability Score by up to 16.2\%, while maintaining or improving output quality.

cs.CL↗

arXiv2026

From Task to Tutorial: An Automated GUI Framework for Excel Tutorial Document and Video Creation

Excel is one of the most widely used productivity tools across domains, offering rich functionality but also overwhelming users with its complexity. This creates a persistent demand for tutorials to support effective usage. However, while building and maintaining the Microsoft tutorial corpus, we observed that existing tutorials are manually created by experts, need frequent updates with each software release, and involve substantial human labor. Moreover, prior work has not achieved fully automated tutorial generation. In this paper, we present the first framework for automatically generating Excel tutorials directly from natural language task descriptions. Our framework first instantiates the task. Then a central component of this framework, Execution Agent, plans and executes the solution in Excel, and collects the intermediate artifacts required for tutorial construction. These artifacts are then transformed into both structured Excel documents and video demonstrations. To build a comprehensive tutorial corpus, we collected 1,559 task descriptions from real-world scenarios. In addition, we designed a systematic evaluation framework that integrates assessments from both large language models (LLMs) and human reviewers. Experimental results show that our framework improves task execution success rates by 8.5% over state-of-the-art baselines. Moreover, the generated tutorials demonstrate superior readability and instructional effectiveness, often approaching or surpassing expert-authored materials. Importantly, the automated pipeline eliminates manual labor and reduces time costs to 1/20 of expert authoring, making scalable and high-quality tutorial generation practical for the first time.

cs.SE↗

arXiv2026

StepFly: Agentic Troubleshooting Guide Automation for Incident Diagnosis

Effective incident management in large-scale IT systems relies on troubleshooting guides (TSGs), but their manual execution is slow and error-prone. While recent advances in LLMs offer promise for automating incident management tasks, existing LLM-based solutions lack specialized support for several key challenges, including managing TSG quality issues, interpreting complex control flow, handling data-intensive queries, and exploiting execution parallelism. We first conducted an empirical study on 92 real-world TSGs, and, guided by our findings, we present StepFly, a novel end-to-end agentic framework for troubleshooting guide automation. Our approach features a three-stage workflow: the first stage provides a comprehensive guide together with a tool, TSG Mentor, to assist site reliability engineers (SREs) in improving TSG quality; the second stage performs offline preprocessing using LLMs to extract structured execution directed acyclic graphs (DAGs) from unstructured TSGs and to create dedicated Query Preparation Plugins (QPPs); and the third stage executes online using a DAG-guided scheduler-executor framework with a memory system to ensure correct workflow and support parallel execution of independent steps. Our empirical evaluation on a collection of real-world TSGs and incidents demonstrates that StepFly achieves a ~94% success rate on GPT-4.1, outperforming baselines with less time and token consumption. Furthermore, it achieves a remarkable execution time reduction of 32.9% to 70.4% for parallelizable TSGs. Our code and sample data are publicly available at https://github.com/microsoft/StepFly.

cs.AI↗

arXiv2026

Contrastive Attribution in the Wild: An Interpretability Analysis of LLM Failures on Realistic Benchmarks

Interpretability tools are increasingly used to analyze failures of Large Language Models (LLMs), yet prior work largely focuses on short prompts or toy settings, leaving their behavior on commonly used benchmarks underexplored. To address this gap, we study contrastive, LRP-based attribution as a practical tool for analyzing LLM failures in realistic settings. We formulate failure analysis as \textit{contrastive attribution}, attributing the logit difference between an incorrect output token and a correct alternative to input tokens and internal model states, and introduce an efficient extension that enables construction of cross-layer attribution graphs for long-context inputs. Using this framework, we conduct a systematic empirical study across benchmarks, comparing attribution patterns across datasets, model sizes, and training checkpoints. Our results show that this token-level contrastive attribution can yield informative signals in some failure cases, but is not universally applicable, highlighting both its utility and its limitations for realistic LLM failure analysis. Our code is available at: https://aka.ms/Debug-XAI.

cs.AI↗

arXiv2026

Not All Tokens Matter: Towards Efficient LLM Reasoning via Token Significance in Reinforcement Learning

Large language models (LLMs) show strong reasoning abilities but often produce unnecessarily long explanations that reduce efficiency. Although reinforcement learning (RL) has been used to improve reasoning, most methods focus on accuracy and rely on uniform length-based rewards that overlook the differing contributions of individual tokens, often harming correctness. We revisit length optimization in RL through the perspective of token significance. Observing that many chain-of-thought (CoT) tokens contribute little to the final answer, we introduce a significance-aware length reward that selectively penalizes insignificance tokens, reducing redundancy while preserving essential reasoning. We also propose a dynamic length reward that encourages more detailed reasoning early in training and gradually shifts toward conciseness as learning progresses. Integrating these components into standard policy optimization yields a framework that improves both reasoning efficiency and accuracy. Experiments across multiple benchmarks demonstrate substantial reductions in response length while preserving or improving correctness, highlighting the importance of modeling token significance for efficient LLM reasoning.

cs.LG↗

arXiv2026

RepoGenesis: Benchmarking End-to-End Microservice Generation from Readme to Repository

Large language models and agents have achieved remarkable progress in code generation. However, existing benchmarks focus on isolated function/class-level generation (e.g., ClassEval) or modifications to existing codebases (e.g., SWE-Bench), neglecting complete microservice repository generation that reflects real-world 0-to-1 development workflows. To bridge this gap, we introduce RepoGenesis, the first multilingual benchmark for repository-level end-to-end web microservice generation, comprising 106 repositories (60 Python, 46 Java) across 18 domains and 11 frameworks, with 1,258 API endpoints and 2,335 test cases verified through a "review-rebuttal" quality assurance process. We evaluate open-source agents (e.g., DeepCode) and commercial IDEs (e.g., Cursor) using Pass@1, API Coverage (AC), and Deployment Success Rate (DSR). Results reveal that despite high AC (up to 73.91%) and DSR (up to 100%), the best-performing system achieves only 23.67% Pass@1 on Python and 21.45% on Java, exposing deficiencies in architectural coherence, dependency management, and cross-file consistency. Notably, GenesisAgent-8B, fine-tuned on RepoGenesis (train), achieves performance comparable to GPT-5 mini, demonstrating the quality of RepoGenesis for advancing microservice generation. We release our benchmark at https://github.com/pzy2000/RepoGenesis.

cs.SE↗

arXiv2026

DUET: Joint Exploration of User Item Profiles in Recommendation System

Traditional recommendation systems represent users and items as dense vectors and learn to align them in a shared latent space for relevance estimation. Recent LLM-based recommenders instead leverage natural-language representations that are easier to interpret and integrate with downstream reasoning modules. This paper studies how to construct effective textual profiles for users and items, and how to align them for recommendation. A central difficulty is that the best profile format is not known a priori: manually designed templates can be brittle and misaligned with task objectives. Moreover, generating user and item profiles independently may produce descriptions that are individually plausible yet semantically inconsistent for a specific user--item pair. We propose Duet, an interaction-aware profile generator that jointly produces user and item profiles conditioned on both user history and item evidence. Duet follows a three-stage procedure: it first turns raw histories and metadata into compact cues, then expands these cues into paired profile prompts and then generate profiles, and finally optimizes the generation policy with reinforcement learning using downstream recommendation performance as feedback. Experiments on three real-world datasets show that Duet consistently outperforms strong baselines, demonstrating the benefits of template-free profile exploration and joint user-item textual alignment.

cs.IR↗

arXiv2026

LLM Reasoning as Trajectories: Step-Specific Representation Geometry and Correctness Signals

This work characterizes large language models' chain-of-thought generation as a structured trajectory through representation space. We show that mathematical reasoning traverses functionally ordered, step-specific subspaces that become increasingly separable with layer depth. This structure already exists in base models, while reasoning training primarily accelerates convergence toward termination-related subspaces rather than introducing new representational organization. While early reasoning steps follow similar trajectories, correct and incorrect solutions diverge systematically at late stages. This late-stage divergence enables mid-reasoning prediction of final-answer correctness with ROC-AUC up to 0.87. Furthermore, we introduce trajectory-based steering, an inference-time intervention framework that enables reasoning correction and length control based on derived ideal trajectories. Together, these results establish reasoning trajectories as a geometric lens for interpreting, predicting, and controlling LLM reasoning behavior.

cs.CL↗

arXiv2026

Table-LLM-Specialist: Language Model Specialists for Tables using Iterative Generator-Validator Fine-tuning

Language models such as GPT and Llama have shown remarkable ability on diverse natural language tasks, yet their performance on complex table tasks (e.g., NL-to-Code and data cleaning) remains suboptimal. Improving performance typically requires task-specific fine-tuning, which depends on expensive human labeling and is prone to overfitting. In this work, we propose Table-LLM-Specialist, a self-trained fine-tuning paradigm designed for table tasks. Our key insight is that many table tasks admit two dual formulations: a generative version and a classification version. Leveraging this duality, we introduce a Generator-Validator paradigm that iteratively generates and validates training data using language models, enabling effective fine-tuning without manually labeled data. Extensive evaluations on Llama, GPT-3.5, and GPT-4 show that Table-LLM-Specialist achieves (1) strong performance across diverse tasks compared to base models, for example, models fine-tuned on GPT-3.5 often surpass GPT-4 level quality; (2) lower deployment cost by enabling smaller models to reach high quality with reduced latency and cost; and (3) better generalization across multiple benchmarks, due to training on diverse, systematically generated data from real-world tables. Our code is available at https://github.com/microsoft/Table-Specialist. Models fine-tuned with Table-LLM-Specialist have been integrated into Microsoft Excel and are deployed in production for automated table data cleaning.

cs.CL↗

arXiv2026

Formula-R1: Incentivizing LLM Reasoning over Complex Tables with Numerical Computation via Formula-Driven Reinforcement Learning

Tables are a fundamental medium for organizing and analyzing data, making table reasoning a critical capability for intelligent systems. Although large language models (LLMs) exhibit strong general reasoning abilities, they still struggle with accurate numerical reasoning over tabular data, particularly in complex table settings beyond simple relational lookup. Spreadsheet formulas provide a powerful and expressive interface for executable symbolic operations, enabling rich reasoning patterns that remain largely underexplored by existing LLMs. In this paper, we introduce Formula-R1, a model trained via Formula Tuning (Fortune), a formula-driven reinforcement learning (RL) framework for table reasoning. Formula Tuning trains LLMs to generate executable spreadsheet formulas for question answering over general tabular data, using execution success and answer correctness as reward signals, thereby reducing reliance on supervised formula annotations. We demonstrate the effectiveness of Formula Tuning through extensive experiments on seven table reasoning benchmarks. It substantially improves LLM performance on table reasoning, particularly for tasks involving complex tables and multi-step numerical computation. Moreover, Formula-R1 consistently outperforms prior methods under controlled comparison settings. Beyond empirical gains, our extensive analyses provide insights into the role of RL in formula-driven table reasoning, highlighting the broader potential of formula-driven RL to enhance reasoning capabilities in LLMs.

cs.AI↗

arXiv2026

Sirens' Whisper: Inaudible Near-Ultrasonic Jailbreaks of Speech-Driven LLMs

Speech-driven large language models (LLMs) are increasingly accessed through speech interfaces, introducing new security risks via open acoustic channels. We present Sirens' Whisper (SWhisper), the first practical framework for covert prompt-based attacks against speech-driven LLMs under realistic black-box conditions using commodity hardware. SWhisper enables robust, inaudible delivery of arbitrary target baseband audio-including long and structured prompts-on commodity devices by encoding it into near-ultrasound waveforms that demodulate faithfully after acoustic transmission and microphone nonlinearity. This is achieved through a simple yet effective approach to modeling nonlinear channel characteristics across devices and environments, combined with lightweight channel-inversion pre-compensation. Building on this high-fidelity covert channel, we design a voice-aware jailbreak generation method that ensures intelligibility, brevity, and transferability under speech-driven interfaces. Experiments across both commercial and open-source speech-driven LLMs demonstrate strong black-box effectiveness. On commercial models, SWhisper achieves up to 0.94 non-refusal (NR) and 0.925 specific-convincing (SC). A controlled user study further shows that the injected jailbreak audio is perceptually indistinguishable from background-only playback for human listeners. Although jailbreaks serve as a case study, the underlying covert acoustic channel enables a broader class of high-fidelity prompt-injection and commandexecution attacks.

cs.CR↗

arXiv2026

MMTU: A Massive Multi-Task Table Understanding and Reasoning Benchmark

Tables and table-based use cases play a crucial role in many important real-world applications, such as spreadsheets, databases, and computational notebooks, which traditionally require expert-level users like data engineers, data analysts, and database administrators to operate. Although LLMs have shown remarkable progress in working with tables (e.g., in spreadsheet and database copilot scenarios), comprehensive benchmarking of such capabilities remains limited. In contrast to an extensive and growing list of NLP benchmarks, evaluations of table-related tasks are scarce, and narrowly focus on tasks like NL-to-SQL and Table-QA, overlooking the broader spectrum of real-world tasks that professional users face. This gap limits our understanding and model progress in this important area. In this work, we introduce MMTU, a large-scale benchmark with over 28K questions across 25 real-world table tasks, designed to comprehensively evaluate models ability to understand, reason, and manipulate real tables at the expert-level. These tasks are drawn from decades' worth of computer science research on tabular data, with a focus on complex table tasks faced by professional users. We show that MMTU require a combination of skills -- including table understanding, reasoning, and coding -- that remain challenging for today's frontier models, where even frontier reasoning models like OpenAI GPT-5 and DeepSeek R1 score only around 69\% and 57\% respectively, suggesting significant room for improvement. We highlight key findings in our evaluation using MMTU and hope that this benchmark drives further advances in understanding and developing foundation models for structured data processing and analysis. Our code and data are available at https://github.com/MMTU-Benchmark/MMTU and https://huggingface.co/datasets/MMTU-benchmark/MMTU.

cs.AI↗

arXiv2026

UFO3: Weaving the Digital Agent Galaxy

Large language model (LLM)-powered agents are transforming digital devices from passive tools into proactive intelligent collaborators. However, most existing frameworks remain confined to a single OS or device, making cross-device workflows brittle and largely manual. We present UFO$^3$, a system that unifies heterogeneous endpoints, desktops, servers, mobile devices, and edge, into a single orchestration fabric. UFO$^3$ models each user request as a mutable TaskConstellation: a distributed DAG of atomic subtasks (TaskStars) with explicit control and data dependencies (TaskStarLines). The TaskConstellation continuously evolves as results stream in from distributed devices, enabling asynchronous execution, adaptive recovery, and dynamic optimization. A Constellation Orchestrator} executes tasks safely and asynchronously while applying dynamic DAG updates, and the Agent Interaction Protocol (AIP) provides persistent, low-latency channels for reliable task dispatch and result streaming. These designs dissolve the traditional boundaries between devices and platforms, allowing agents to collaborate seamlessly and amplify their collective intelligence. We evaluate UFO$^3$ on NebulaBench, a benchmark of 55 cross-device tasks across 5 machines and 10 categories. UFO$^3$ achieves 83.3% subtask completion, 70.9% task success, exposes parallelism with an average width of 1.72, and reduces end-to-end latency by 31% relative to a sequential baseline. Fault-injection experiments demonstrate graceful degradation and recovery under transient and permanent agent failures. These results show that UFO$^3$ achieves accurate, efficient, and resilient task orchestration across heterogeneous devices, uniting isolated agents into a coherent, adaptive computing fabric that extends across the landscape of ubiquitous computing.

cs.DC↗

arXiv2026

Zipage: Maintain High Request Concurrency for LLM Reasoning through Compressed PagedAttention

With reasoning becoming the generative paradigm for large language models (LLMs), the memory bottleneck caused by KV cache during the decoding phase has become a critical factor limiting high-concurrency service. Although existing KV cache eviction methods address the memory issue, most of them are impractical for industrial-grade applications. This paper introduces Compressed PagedAttention, a method that combines token-wise KV cache eviction with PagedAttention. We propose a comprehensive scheduling strategy and support prefix caching and asynchronous compression for Compressed PagedAttention. Based on this, we have developed a high-concurrency LLM inference engine, Zipage. On large-scale mathematical reasoning tasks, Zipage achieves around 95\% of the performance of Full KV inference engines while delivering over 2.1$\times$ speedup.

cs.DC↗

arXiv2026

Test-Time Learning of Causal Structure from Interventional Data

Supervised causal learning has shown promise in causal discovery, yet it often struggles with generalization across diverse interventional settings, particularly when intervention targets are unknown. To address this, we propose TICL (Test-time Interventional Causal Learning), a novel method that synergizes Test-Time Training with Joint Causal Inference. Specifically, we design a self-augmentation strategy to generate instance-specific training data at test time, effectively avoiding distribution shifts. Furthermore, by integrating joint causal inference, we developed a PC-inspired two-phase supervised learning scheme, which effectively leverages self-augmented training data while ensuring theoretical identifiability. Extensive experiments on bnlearn benchmarks demonstrate TICL's superiority in multiple aspects of causal discovery and intervention target detection.

cs.LG↗

arXiv2026

Computer-Using World Model

Agents operating in complex software environments benefit from reasoning about the consequences of their actions, as even a single incorrect user interface (UI) operation can derail long, artifact-preserving workflows. This challenge is particularly acute for computer-using scenarios, where real execution does not support counterfactual exploration, making large-scale trial-and-error learning and planning impractical despite the environment being fully digital and deterministic. We introduce the Computer-Using World Model (CUWM), a world model for desktop software that predicts the next UI state given the current state and a candidate action. CUWM adopts a two-stage factorization of UI dynamics: it first predicts a textual description of agent-relevant state changes, and then realizes these changes visually to synthesize the next screenshot. CUWM is trained on offline UI transitions collected from agents interacting with real Microsoft Office applications, and further refined with a lightweight reinforcement learning stage that aligns textual transition predictions with the structural requirements of computer-using environments. We evaluate CUWM via test-time action search, where a frozen agent uses the world model to simulate and compare candidate actions before execution. Across a range of Office tasks, world-model-guided test-time scaling improves decision quality and execution robustness.

cs.SE↗

arXiv2026

AdNanny: One Reasoning LLM for All Offline Ads Recommendation Tasks

Large Language Models (LLMs) have shown strong capabilities in Natural Language Understanding and Generation, but deploying them directly in online advertising systems is often impractical due to strict millisecond-level latency constraints. This has motivated the use of LLMs offline to improve retrieval, ranking, and recommendation models. Existing solutions typically fine-tune separate LLMs for individual tasks such as query-ad relevance labeling, keyword-based query generation, and user profiling. This results in redundant models, high maintenance cost, and limited performance gains despite substantial overlap in domain knowledge and reasoning patterns. We introduce AdNanny, a unified reasoning-centric LLM that serves as a shared backbone for offline advertising tasks. AdNanny is obtained by fine-tuning a public 671B-parameter DeepSeek-R1 checkpoint using a scalable training system that supports hybrid dense-MoE parallelism. We construct reasoning-augmented corpora that pair structured supervision with step-by-step natural language explanations. A multi-task supervised fine-tuning stage with adaptive reweighting enables AdNanny to handle diverse labeling and generation tasks in a consistent reasoning format. This is followed by reinforcement learning using downstream advertising metrics to align model behavior with online retrieval and ranking objectives. AdNanny is deployed in production within Bing Ads, where it significantly reduces manual labeling effort and improves accuracy across multiple offline tasks. By consolidating many task-specific models into a single reasoning-centric foundation model, AdNanny provides a scalable and cost-effective solution for large-scale advertising systems.

cs.SE↗