arXiv subjects

Zihan Xu

Publications and source records attributed to Zihan Xu.

At least 19 recordsLinked to original sources

arXiv2026

Affective and Functional Needs in AI Use: User Patterns and Their Relation to Real-Life Support and Well-Being

Conversational AI systems are increasingly used to fulfill both functional and affective needs, yet how users perceive the AI-adoption effects varies. Existing research tends to treat these two needs as binary categories and to classify users by a single dominant need. However, this approach fails to capture users who engage with AI for both affective and functional purposes simultaneously, leaving it unclear how these needs combine in real-world AI use. To capture the multifaceted nature of users' AI engagement, we developed the AI-Support measurement with 1,090 participants, grounded in social support theory, which evaluates four dimensions of perceived AI support: emotional/esteem, social companionship, informational, and instrumental support. This study provided (1) an AI-Support Scale developed to measure the perceived support types from human-AI interaction, (2) evidence that users might adopt AI with a high level of both functional and affective use, and (3) exploratory indicators of users' real-life social support and well-being, enabling descriptive results aligning with theoretical expectations. Overall, this study highlights the importance of understanding individual differences in AI affordance use, enabling AI systems to be designed in alignment with different user adoption profiles rather than assuming uniform benefits across all users.

cs.HC↗

arXiv2026

Locating Failure in Multi-Page Visually Rich Document Understanding Systems: An Empirical Attribution

Multi-page visually rich document understanding (MP-VRDU) systems can fail when evidence is not faithfully represented, not selected, or not used correctly, yet final accuracy alone does not reveal where failures occur. We present an empirical attribution framework that examines these failures through controlled interventions in a unified document question answering pipeline. Experiments on MMLongBench-Doc, with cross-dataset and cross-family checks, show that complementary modalities can reduce sensitivity to conversion errors: images mitigate parser errors, while extracted text reduces dependence on visual resolution. Adding related non-gold pages can both repair incorrect answers and break correct ones, even when all annotated evidence is present, revealing instability beneath aggregate accuracy. Abstention trades unsupported answers for false refusals, while its interaction with reasoning instructions differs across backbones. Qualitative examples and accuracy--cost comparisons connect these findings to guidance on representation, evidence selection, and reasoner choice under deployment constraints.

cs.AI↗

arXiv2026

Every Batch Is Its Own Validation Set: Leave-One-Out Gradient Matching for Online Data Selection in LLM Fine-Tuning

Online batch selection fine-tunes a language model on the most useful part of each candidate batch. Selectors that match the gradient of the candidate batch are attractive because they need no held-out data, yet they rarely beat training on the whole batch. We show why. In-sample gradient matching uses every example as part of its own target, so its objective credits each example with its own gradient noise. This is the covariance penalty that makes training error optimistic, now sitting on the diagonal of the gradient Gram matrix: it steers selection toward the noisiest examples and makes the full batch the best solution the objective can reach. The fix costs nothing. For each example, the other candidates form an independent sample of the data distribution, so removing the diagonal turns the matching objective into an unbiased estimate of the update's error with respect to the population gradient. The minimizer of this leave-one-out objective weights examples by their gradient signal-to-noise ratio (SNR), and whenever per-example SNR is heterogeneous enough, half of a batch yields a lower-error update than the whole batch; we give the exact condition. We build \method{} on this principle. It computes the Gram matrix in the metric of the Adam preconditioner during the ordinary backward pass, selects a weighted subset greedily with a $(1-e^{-γ})$ guarantee, and uses no held-out data. Across four fine-tuning tasks and seven backbones from 1.5B to 8B parameters, LOOM improves on full-batch training by 2.3 and 2.4 points on Llama-3.1-8B and Qwen2.5-7B, exceeds every in-sample gradient matcher by 2.4 points and the validation-guided GREATS and OPUS by 1.6--2.0, and selects injected label noise at under a fifth of its base rate.

cs.LG↗

arXiv2026

Information-Computation Inversion in Pseudo-Marginal MCMC

Observation refinement changes posterior uncertainty and the likelihood calculation in pseudo-marginal MCMC. We compare their combined effect through finite-run squared-error risk. A sufficient inversion condition relates information gain to accepted event flow and coarse-kernel contraction. A bootstrap construction realizes inversion at every fixed particle count. We then couple cross-event proposals and refresh same-event proposals independently. A swap identity establishes invariance; continuation identities describe subsequent risk. In the same finite model, a rational certificate proves inversion against optimized constant mixtures and repair by selective allocation over an initialization class at a common action-price budget. Reaction-network experiments measure CPU costs. Under finite-pool initialization, selective allocation reduces event mean-squared error by 40.5% against a tuned mixture at 25 post-initialization CPU seconds. Paired transcription observations show how increased particle effort can raise finite-budget error.

stat.ME↗

arXiv2026

Hyperedge-Triggered Hawkes Processes for Diagnosing Higher-Order Temporal Dependence in Event Streams

Monte Carlo simulation is used to assess a likelihood-based EM procedure for diagnosing higher-order temporal dependence in multivariate event streams. Multivariate Hawkes processes provide a likelihood-based framework for modelling self- and cross-excitation in continuous-time event streams, but their standard pairwise formulation can confound dense dyadic excitation with genuinely group-triggered temporal dependence. We introduce a hyperedge-triggered Hawkes model in which a candidate group contributes additional intensity only after all of its member nodes fire within a prescribed trigger window. A most-recent-anchor convention makes this higher-order component non-accumulating and leads to a corresponding piecewise compensator in the likelihood. The model is estimated by a latent-branching EM algorithm that assigns event-level responsibilities to background, pairwise, and hyperedge-triggered sources. Controlled simulations evaluate recovery, regularisation, convergence, separation from pairwise baselines, trigger-window and kernel-timescale sensitivity, computational scaling, and comparison with a parameter-matched third-order surrogate. The results show that higher-order excitation can be recovered in informative regimes, while pairwise--hyperedge confounding remains the main statistical limitation. Applications to retinal and visual-cortex spike trains are presented as exploratory predictive analyses: the hyperedge-triggered component can improve candidate-penalised held-out fit, but the fitted effects should be interpreted as diagnostics of higher-order temporal dependence rather than calibrated discoveries of individual biological hyperedges.

stat.ME↗

arXiv2026

A Two-Tier Perspective on Inference-Time Parallelism in Multi-Agent LLM Systems

Large language model (LLM)-driven multi-agent systems typically require multiple model invocations and complex coordination during inference, and their execution strategies directly affect system accuracy, latency, and computational cost. Parallel execution provides a means to improve inference-time efficiency. From the perspective of inference-time execution, this paper models parallelism in multi-agent systems as two distinct levels of decision processes: Replica Parallelism, which explores multiple complete solution paths at the task level, and Structural Parallelism, which enables concurrent execution within a single solution path through task decomposition. However, the roles of different forms of parallelism and their interrelationships still lack systematic study in terms of unified organization and coordination. We therefore propose TIPEX, a controllable execution framework that unifies these two levels of parallelism and coordinates their roles within the inference process under a unified execution semantics while supporting systematic combinations and analyses of different parallel strategies and parameter configurations. Systematic experiments on the GAIA benchmark demonstrate that inference-time parallelism can significantly improve accuracy and reduce end-to-end latency at the cost of increased token consumption. Further analysis shows that Replica and Structural Parallelism exhibit complementary effects across task complexities, with tasks of intermediate difficulty benefiting most from their coordination, while overly aggressive parallel strategies do not necessarily yield better performance.

cs.MA↗

arXiv2026

Tencent WorkBuddy Bench: A Multi-Domain Coding-Agent Benchmark with Contamination-Resistant Task Construction

We introduce Tencent WorkBuddy Bench, a multi-domain evaluation suite for coding agents; this report documents its construction methodology, scoring protocol, and a cross-model leaderboard. At its core is a unified evaluation framework for constructing and running distribution-informed coding-agent tasks across four work domains - Code, Web, Office, and Security. Rather than adapting public issue text, every task is reverse-engineered from a real commit, pull request, or business scenario and rewritten as a short, colloquial, role-played request, so that a task's prompt is not recoverable by web-searching the underlying issue, pull request, or commit thread. Because the dataset is released openly - task directories, environment images, evaluation harness, tests, and reference solutions - contamination resistance rests on this construction together with dataset versioning rather than on secrecy. The four subsets - repository-level engineering, front-end development, office and business workflows, and red-/blue-team security - probe complementary facets of real work, each with its own verification style. All are packaged in a uniform task-directory format and run, under a uniform and reproducible protocol, on two agent harnesses (CodeBuddy Code and Claude Code); the full open release makes the benchmark reproducible end to end and directly auditable, since any third party can re-run each task and inspect its content. Because each subset uses a different scoring instrument, scores are not comparable across subsets and the suite reports no suite-wide average. We report a cross-model leaderboard across several model families.

cs.CL↗

arXiv2026

LongMedBench: Benchmarking Medical Agents for Long-Horizon Clinical Decision-Making

In this work, we introduce LongMedBench, a real-world EHR-based benchmark for long-horizon clinical decision-making. Prior evaluations of LLM-based medical agents have largely emphasized short-context knowledge QA and tool use. However, real-world medical care is inherently longitudinal, and clinicians must aggregate evidence across repeated visits, tests, and evolving treatments. Therefore, long-horizon interaction is essential for realistic assessment. LongMedBench is constructed via a reproducible pipeline that integrates MIMIC-IV admission records and clinical notes into time-series event streams and long-context memory datasets, enabling long-horizon, multi-session interactions between agents and a clinical environment. It comprises 335 patients, with 19.72 inpatient visits per patient on average and 44.91 medical events per visit. Guided by the long-horizon decision process, we propose an evaluation taxonomy with three suites: fact-based QA, temporal reasoning, and long-horizon decision-making. This taxonomy measures how agents understand and leverage historical patient information over extended horizons. Our experiments show that while recent LLMs can make good use of explicit timestamps, they have challenges in implicit time inference; The RAG and agent memory system can improve the performance of information retrieval tasks, but the performance of decision-making tasks is highly dependent on the model's immediate context.

cs.AI↗

arXiv2026

X-FEMR: A Token-level Explainable Approach for Electronic Health Records Foundation Models using Transformer-based Models

Foundation Models for Electronic Health Records (FEMRs) are pretrained on large-scale structured patient data, enabling them to convert longitudinal patient trajectories into generalizable representations for diverse clinical prediction tasks. Despite their effectiveness, FEMRs remain black-box models, raising concerns about bias, interpretability, and clinical trust. To address this, we propose the first token-level explainability approach for FEMRs. We train a Transformer-based surrogate model on input-output pairs from the FEMR across two prediction tasks, approximating its behavior while preserving temporal dynamics. We identify the most influential tokens, providing insights into how FEMRs leverage different aspects of patient history for predictions. To evaluate clinical relevance, we introduce a novel clinical alignment metric that quantifies the correspondence between the surrogate model's key tokens and clinically validated features. Our results demonstrate that the surrogate closely approximates FEMR predictions and that token-level explanations align well with clinical knowledge, offering a practical framework for interpretable and trustworthy clinical AI.

cs.LG↗

arXiv2026

Beyond Skepticism: Evaluating LLMs Pedagogical Intent Reasoning with the Adaptive Pedagogical Vigilance Framework

The capacity of Large Language Models (LLMs) to reason about pedagogical intent within instructional communication remains underexplored, particularly in educational domains such as translation pedagogy. To address this, we propose the \textbf{Adaptive Pedagogical Vigilance (APV)} framework, a novel computational formalism that reframes communicative vigilance as an adaptive mechanism for optimizing learning through intent inference. APV formalizes the problem via a Bayesian Pedagogical Intent Inference Engine (PIIE), which models how instructors select content to maximize pedagogical utility and how vigilant learners should inversely reason about latent instructional configurations -- encompassing genre, stance, and incentives. We evaluate APV through a three-tier hierarchy: distinguishing instructional genre, reasoning about structured pedagogical setups, and generalizing to authentic educational discourse. Experiments on leading LLMs (e.g., GPT-4o, Claude 3.5) show that APV substantially improves model vigilance. It achieves the strongest discrimination between pedagogical and exposure-based content, correlates highly with human judgments ($r=0.958$), and maintains robust performance on naturalistic data where baseline methods degrade. This work establishes a unified framework for assessing and enhancing LLMs' understanding of pedagogical motives, advancing the development of more reliable AI-assisted learning systems.

cs.CL↗

arXiv2026

DocOCR-Eval: A Correction-Based Framework for OCR Tool Selection Without Ground Truth

Document parsing is a foundational step for document understanding tasks such as visual question answering and key information extraction, as it transforms unstructured scanned images into structured representations by extracting textual, visual, and layout information. While numerous Optical Character Recognition (OCR) engines and multimodal large language models (MLLMs) have been developed for this purpose, selecting an appropriate document parsing solution for a given document collection remains challenging, particularly in label-scarce settings. In this work, we conduct a systematic evaluation of text recognition performance across a diverse set of OCR engines and state-of-the-art MLLMs on multiple scanned document benchmarks spanning different domains and languages. Motivated by the limited contextual reasoning capabilities of many OCR engines and the high cost of manual annotations, we propose DocOCR-Eval, an annotation-free evaluation framework for automatic OCR assessment and selection. DocOCR-Eval employs a three-staged correction and ranking strategy to approximate annotation-based tool ordering without ground-truth labels. We show that aggregating across multiple MLLMs progressively improves alignment with annotation-based rankings. Extensive experiments further demonstrate that reliable OCR tool selection can be achieved in realistic, label-limited settings, providing practical guidance for deploying document parsing systems across diverse real-world document collections.

cs.LG↗

arXiv2026

Curation and Extraction of Drug-Related Entities from Reddit Platform

Physicians learn primarily about illicit drugs from clinical overdose cases, limiting their understanding of real-world usage. Meanwhile, drug users share first-hand experiences online, offering insights into dosage and effects of drugs. To bridge this gap, we introduce ReDose (REddit Drug DOSe and Effect), a dataset of 6,435 Reddit posts on substance use. A board-certified toxicologist primarily annotated both the training and test sets, while two medical science students contributed to the test set, labeling DRUG, DOSE, and EFFECT entities. We benchmarked 6,267 annotations using BERT-based, large language model (LLM)-based, and Retrieval-Augmented Generation (RAG) models. BiomedBERT achieved an F1-score of 0.843 for DRUG, while Llama-3 70B outperformed GPT-4 (F1 = 0.79 vs. 0.72). EFFECT extraction remains challenging, with GPT-4 achieving a recall of 0.41. ReDose captures patient-curated narratives to advance medical data extraction from social media.

cs.CL↗

arXiv2026

Establishing dermatopathology encyclopedia DermpathNet with Artificial Intelligence-Based Workflow

Accessing high-quality, open-access dermatopathology image datasets for learning and cross-referencing is a common challenge for clinicians and dermatopathology trainees. To establish a comprehensive open-access dermatopathology dataset for educational, cross-referencing, and machine-learning purposes, we employed a hybrid workflow to curate and categorize images from the PubMed Central (PMC) repository. We used specific keywords to extract relevant images, and classified them using a novel hybrid method that combined deep learning-based image modality classification with figure caption analyses. Validation on 651 manually annotated images demonstrated the robustness of our workflow, with an F-score of 89.6% for the deep learning approach, 61.0% for the keyword-based retrieval method, and 90.4% for the hybrid approach. We retrieved over 7,772 images across 166 diagnoses and released this fully annotated dataset, reviewed by board-certified dermatopathologists. Using our dataset as a challenging task, we found the current image analysis algorithm from OpenAI inadequate for analyzing dermatopathology images. In conclusion, we have developed a large, peer-reviewed, open-access dermatopathology image dataset, DermpathNet, which features a semi-automated curation workflow.

cs.CV↗

arXiv2026

SmartSnap: Proactive Evidence Seeking for Self-Verifying Agents

Agentic reinforcement learning (RL) holds great promise for the development of autonomous agents under complex GUI tasks, but its scalability remains severely hampered by the verification of task completion. Existing task verification is treated as a passive, post-hoc process: a verifier (i.e., rule-based scoring script, reward or critic model, and LLM-as-a-Judge) analyzes the agent's entire interaction trajectory to determine if the agent succeeds. Such processing of verbose context that contains irrelevant, noisy history poses challenges to the verification protocols and therefore leads to prohibitive cost and low reliability. To overcome this bottleneck, we propose SmartSnap, a paradigm shift from this passive, post-hoc verification to proactive, in-situ self-verification by the agent itself. We introduce the Self-Verifying Agent, a new type of agent designed with dual missions: to not only complete a task but also to prove its accomplishment with curated snapshot evidences. Guided by our proposed 3C Principles (Completeness, Conciseness, and Creativity), the agent leverages its accessibility to the online environment to perform self-verification on a minimal, decisive set of snapshots. Such evidences are provided as the sole materials for a general LLM-as-a-Judge verifier to determine their validity and relevance. Experiments on mobile tasks across model families and scales demonstrate that our SmartSnap paradigm allows training LLM-driven agents in a scalable manner, bringing performance gains up to 26.08% and 16.66% respectively to 8B and 30B models. The synergizing between solution finding and evidence seeking facilitates the cultivation of efficient, self-verifying agents with competitive performance against DeepSeek V3.1 and Qwen3-235B-A22B. Code is available at: https://github.com/TencentYoutuResearch/SmartSnap

cs.CL↗

arXiv2025

Youtu-Agent: Scaling Agent Productivity with Automated Generation and Hybrid Policy Optimization

Existing Large Language Model (LLM) agent frameworks face two significant challenges: high configuration costs and static capabilities. Building a high-quality agent often requires extensive manual effort in tool integration and prompt engineering, while deployed agents struggle to adapt to dynamic environments without expensive fine-tuning. To address these issues, we propose \textbf{Youtu-Agent}, a modular framework designed for the automated generation and continuous evolution of LLM agents. Youtu-Agent features a structured configuration system that decouples execution environments, toolkits, and context management, enabling flexible reuse and automated synthesis. We introduce two generation paradigms: a \textbf{Workflow} mode for standard tasks and a \textbf{Meta-Agent} mode for complex, non-standard requirements, capable of automatically generating tool code, prompts, and configurations. Furthermore, Youtu-Agent establishes a hybrid policy optimization system: (1) an \textbf{Agent Practice} module that enables agents to accumulate experience and improve performance through in-context optimization without parameter updates; and (2) an \textbf{Agent RL} module that integrates with distributed training frameworks to enable scalable and stable reinforcement learning of any Youtu-Agents in an end-to-end, large-scale manner. Experiments demonstrate that Youtu-Agent achieves state-of-the-art performance on WebWalkerQA (71.47\%) and GAIA (72.8\%) using open-weight models. Our automated generation pipeline achieves over 81\% tool synthesis success rate, while the Practice module improves performance on AIME 2024/2025 by +2.7\% and +5.4\% respectively. Moreover, our Agent RL training achieves 40\% speedup with steady performance improvement on 7B LLMs, enhancing coding/reasoning and searching capabilities respectively up to 35\% and 21\% on Maths and general/multi-hop QA benchmarks.

cs.AI↗

arXiv2025

Unusual strain relaxation and Dirac semimetallic behavior in epitaxial antiperovskite nitrides

Antiperovskite nitrides (X3AN) are the structural analogues to perovskite oxides, while their epitaxial growth and electronic properties remain largely unexplored. We report the successful synthesis of Ni3InN thin films on substrates with different lattice constants. First-principles phonon calculations confirm the dynamical stability of cubic phase Ni3InN, providing the basis for epitaxial synthesis. High-resolution scanning transmission electron microscopy reveals coherent (001)-oriented interfaces when Ni3InN is grown on LaAlO3 and SrTiO3, while an unexpected (011)-orientation forms on DyScO3, aligning with surface-energy predictions. Transport measurements highlight a strain-controlled Fermi-liquid behavior, correlated with variations in the Ni-3d bandwidth and hybridization. Band structure calculations reveal a dual character near the Fermi level: a high-mobility Dirac-like band and a Ni-3d manifold that drives strange-metal transport with a reduced slope compared to oxide perovskites. The formal Ni valence (+2/3) places Ni3InN in an overdoped correlated-metal regime, distinguishing from most perovskite oxides. This positions antiperovskite nitrides as a promising platform for investigating overdoped Fermi liquids and strange-metal behavior.

cond-mat.mtrl-sci↗

arXiv2025

CKM-Enabled Joint Spatial-Doppler Domain Clutter Suppression for Low-Altitude UAV ISAC

The rapid development of low-altitude economy has placed higher demands on the sensing of small-sized unmanned aerial vehicle (UAV) targets. However, the complex and dynamic low-altitude environment, like the urban and mountainous areas, makes clutter a significant factor affecting the sensing performance. Traditional clutter suppression methods based on Doppler difference or signal strength are inadequate for scenarios with dynamic clutter and slow-moving targets like low-altitude UAVs. In this paper, motivated by the concept of channel knowledge map (CKM), we propose a novel clutter suppression technique for orthogonal frequency division multiplexing (OFDM) integrated sensing and communication (ISAC) system, by leveraging a new type of CKM named clutter angle map (CLAM). CLAM is a site-specific database, containing location-specific primary clutter angles for the coverage area of the ISAC base station (BS). With CLAM, the sensing signal components corresponding to the clutter environment can be effectively removed before target detection and parameter estimation, which greatly enhances the sensing performance. Besides, to take into account the scenarios when the targets and clutters are in close directions so that pure CLAM-based spatial domain clutter suppression is no longer effective, we further propose a two-step CLAM-enabled joint spatial-Doppler domain clutter suppression algorithm. Simulation results demonstrate that the proposed technique effectively suppresses clutter and enhances target sensing performance, achieving accurate parameter estimation for sensing slow-moving low-altitude UAV targets.

eess.SP↗

arXiv2025

Learn the Ropes, Then Trust the Wins: Self-imitation with Progressive Exploration for Agentic Reinforcement Learning

Reinforcement learning (RL) is the dominant paradigm for sharpening strategic tool use capabilities of LLMs on long-horizon, sparsely-rewarded agent tasks, yet it faces a fundamental challenge of exploration-exploitation trade-off. Existing studies stimulate exploration through the lens of policy entropy, but such mechanical entropy maximization is prone to RL instability due to the multi-turn distribution shifting. In this paper, we target the progressive exploration-exploitation balance under the guidance of the agent's own experiences without succumbing to either entropy collapsing or runaway divergence. We propose SPEAR, a self-imitation learning (SIL) recipe for training agentic LLMs. It extends the vanilla SIL, where a replay buffer stores good experience for off-policy update, by gradually steering the policy entropy across stages. Specifically, the proposed curriculum scheduling harmonizes intrinsic reward shaping and self-imitation to 1) expedite exploration via frequent tool interactions at the beginning, and 2) strengthen exploitation of successful tactics upon convergence towards familiarity with the environment. We also combine bag-of-tricks of industrial RL optimizations for a strong baseline Dr.BoT to demonstrate our effectiveness. In ALFWorld and WebShop, SPEAR increases the success rates of GRPO/GiGPO/Dr.BoT by up to 16.1%/5.1%/8.6% and 20.7%/11.8%/13.9%, respectively. In AIME24 and AIME25, SPEAR boosts Dr.BoT by up to 3.8% and 6.1%, respectively. Such gains incur only 10%-25% extra theoretical complexity and negligible runtime overhead in practice, demonstrating the plug-and-play scalability of SPEAR.

cs.LG↗