Searcharxiv⌕ Search

arXiv subjects

Xiaowen Zhang

Publications and source records attributed to Xiaowen Zhang.

At least 19 recordsLinked to original sources

An Extensive Evaluation of Machine Learning and Deep Learning Approaches for Sentiment Analysis: Insights from Amazon Customer Reviews

The rapid growth of user-generated text on online platforms creates opportunities for understanding customer opinion and supporting data-driven decisions. This study compares five traditional machine learning classifiers (Random Forest, Logistic Regression, Bernoulli Naive Bayes, Linear Support Vector Classifier, and Multinomial Naive Bayes) and six deep learning architectures (LSTM, Bi-LSTM, GRU, RNN, Attention, and FastText) for three-class sentiment classification of 568,454 Amazon Fine Food Reviews. Machine learning models used Bag-of-Words (BoW) and TF-IDF features limited to 5,000 terms; deep learning models used learned embeddings over a 20,000-word vocabulary. Models were evaluated on the original imbalanced data and on a random-oversampled balanced set of 1,070,469 reviews, using accuracy, macro and per-class precision, recall, and F1-score, and training time. On the balanced set, Random Forest with BoW achieved 99\% accuracy and the Attention model achieved 98.7\%. Linear SVC reached 89\% accuracy in 1 min 15 s, and Logistic Regression reached 90\% in 3 min 37 s. On the imbalanced set, the Attention model achieved 97.8\% accuracy, and FastText achieved 95.3\% in 3 min 26 s, while the RNN reached 91.9\% with a neutral-class F1-score of 0.08. Deep learning models were more stable across class distributions, whereas the best classical model depended on balancing. We also report a user-facing application that applies the trained Random Forest and Attention models to new review text. Because the labels are derived from star ratings, the results measure agreement with rating-derived sentiment classes rather than annotated sentiment; the dataset is a single source, and the findings should be read with that limitation in mind.

cs.LG↗

Celia: Catalog Embeddings Learned from cosmologIcal simulAtions

Cosmological hydrodynamical simulations connect observed galaxies to cosmology and the physics of galaxy formation. Each simulation is expensive, and different codes model galaxy formation in different ways. We present Celia (Catalog Embeddings Learned from cosmologIcal simulAtions), a self-supervised foundation model that maps catalogs of subhalos and their galaxy properties to an embedding vector for the purposes of compression and organization of astrophysical and large-scale structure information. Celia is pretrained without parameter labels on the IllustrisTNG, SIMBA and Astrid suites of CAMELS and their N-body counterparts, by reconstructing masked subhalo properties and predicting the representations of a teacher network. Simple linear models fitted to the embeddings recover $Ω_{\mathrm m}$ with $R^2=0.999$, and $σ_8$ and both supernova-feedback amplitudes with $R^2\geq0.97$, in all three suites. The embeddings translate well to out-of-distribution simulations such as Swift-EAGLE, a code that Celia never saw. A linear model fitted on the three training codes predicts its $Ω_{\mathrm m}$ with a correlation coefficient $r=0.985$, and linear models fitted on 20 labeled simulations reach $r\geq0.94$ for $Ω_{\mathrm m}$, $σ_8$ and a supernova-feedback amplitude. We decompose the variation among the simulations of each suite into principal components of the embedding, and simple power laws of the simulation parameters describe the leading components. Celia gives simulated galaxy catalogs a shared coordinate system for simulation-based inference, emulation, the comparison of galaxy-formation models, the detection of physical degeneracies and the discovery of new astrophysical probes.

astro-ph.GA↗

Turán numbers of triangular pyramids

Let $\TP_k$ denote the triangular pyramid with $k$ layers. For a graph family $\mathcal F$, let $\ex(n,\mathcal F)$ be the maximum number of edges in an $n$-vertex graph containing no member of $\mathcal F$ as a subgraph. For a graph $H$, let $\cJ(H)$ be the family of subgraph-minimal graphs among all such graphs $H-I$, where $I\subseteq V(H)$ is an independent set. For every integer $k\ge4$ and all sufficiently large $n$, we prove that \[ \ex(n,\TP_k)=\max_{0\le m\le n}\{m(n-m)+\ex(m,\cJ(\TP_k))\}+O_k(n)=\left\lfloor n^2/4\right\rfloor+Θ_k(n^{4/3}). \] Moreover, for $k=4$, we characterize all extremal graphs and show that \[ \ex(n,\TP_4)=\max_{0\le m\le n}\bigl\{m(n-m)+\ex\big(m,\cJ(\TP_4)\big)\bigr\} \] for sufficiently large $n$.

math.CO↗

Forgetting is Not Always Bad: A Neuro-Inspired Memory Repair Mechanism for Poisoned LLM Agents

Large language model (LLM) agents are highly vulnerable to memory-poisoning attacks. Existing defenses primarily rely on external modules for either static memory isolation or continuous online auditing, resulting in low memory efficiency and high computational overhead. To address these limitations, inspired by neuroscientific mechanisms that weak reactivation can destabilize memories and induce decay in the absence of reinforcement, we propose an agent-intrinsic training-free memory repair framework, DREAM (Dynamic Reactivation for Engram Attenuation in Memory), that enables selective functional forgetting of poisoned memories. Specifically, DREAM implements a three-stage pipeline: perturbation-induced implicit reactivation to reveal the structural fragility of malicious memories, adaptive anomaly diagnosis based on multi-dimensional activation patterns to detect topological anomalies, and a dynamic memory repair module that selectively suppresses harmful memories while reinforcing benign ones. Extensive experiments across diverse attackers and real-world tasks demonstrate that DREAM reduces attack success rates by over 95\% against backdoor poisoning while preserving strong benign utility. DREAM also achieves competitive robustness against injection attacks. In terms of runtime and token consumption, DREAM achieves up to a 2.13$\times$ speedup and a 37.7\% reduction in token consumption compared with A-MemGuard. Furthermore, DREAM also achieves a task success rate of 92.96\% on poisoned multi-agent systems.

cs.AI↗

AI-assisted super-resolution cosmological simulations V: Cosmology-aware super-resolution

We present a cosmology-aware generative model for super-resolution (SR) of cosmological $N$-body simulations in particle phase space. The model jointly generates displacements and velocities, conditioning a generative adversarial network (GAN) based generator and discriminator on evolved low-resolution (LR) fields and five $Λ$CDM parameters $(Ω_m,Ω_b,h,n_s,σ_8)$. We train a single model on 100 matched low-high resolution pairs from the Quijote Latin-hypercube suite at $z = 0$, mapping $512^3$ to $1024^3$ particles in $1\,h^{-1}\,\mathrm{Gpc}$ boxes. On ten cosmologies excluded from SR training, the model reproduces the broad matter distribution and the cosmological variation of clustering and halo abundances. Over $0.1<k<3\,h\,\mathrm{Mpc}^{-1}$, the mean of the per-cosmology maximum absolute power-spectrum errors is $3.6\%$, while FoF halo abundances agree with HR to within $15\%$ in the analyzed mass bins below $10^{14.8}\,h^{-1}M_\odot$. Satellite abundances and occupations retain larger, mass-dependent discrepancies. Corrections measured at one near-fiducial cosmology reduce a shared component of these systematic errors when applied to the other nine test cosmologies. A power spectrum based inference test also shows that calibration reduces the mean shifts in all parameters. These results extend our SR model across cosmologies and provide 6D phase space realizations for subsequent studies of non-linear structure and field-level cosmological inference.

astro-ph.CO↗

Understanding Bugs in Modern Agentic Frameworks: A Study of Symptoms, Root Causes, and Triggering Conditions

Modern agentic frameworks such as CrewAI and AutoGen have evolved into complex, autonomous multi-agent systems, introducing reliability challenges that go beyond earlier pipeline-based LLM libraries. However, existing empirical studies focus on earlier LLM libraries or task-level bugs, leaving the unique complexities of these agentic frameworks unexplored. We present a comprehensive study of 409 fixed bugs across five representative agentic frameworks, proposing a five-layer architectural abstraction. Our taxonomy identifies previously unreported symptom categories---Unexpected Execution Sequence, User Configuration Ignored, and Incomplete/Incorrect Trace---and isolates agent-specific root causes including Model-Related Fault, Cognitive Context Mismanagement, and Orchestration Fault. Notably, the model integration layer is the most bug-prone yet receives disproportionately low test inclusion rate during bug fixing (47%), revealing a critical validation gap. Despite varying design paradigms, bug symptoms, root causes, and bug-prone components show substantial cross-framework consistency (JS similarity 0.62--0.88). Finally, we present the first systematic study of bug-triggering conditions, identifying error-prone factor combinations across element configurations, input patterns, and operations, and demonstrate their transferability across frameworks, providing a foundation for test oracle design and cross-framework benchmark.

cs.SE↗

Batched Speech Decisions Without Decoding: Single-Token Supervision Lets a Frozen LLM Hear Beyond the Transcript

Full-duplex voice agents make many small, closed decisions, which current systems answer by slow autoregressive decoding. We propose DuplexJev, which feeds ASR-encoder hidden states through a small connector into a frozen LLM and reads each question as a single-token distribution over its options. Nothing is decoded, and an 8-GPU node answers 80 decisions about eight utterances in about 0.1 s. With a last-layer connector, spoken QA stays close to reading the transcript (90% vs. 91%). DuplexJev also hears the speaker: gender and emotion accuracy both reach 90% (from 55% and 28%) with a cross-attention connector, whose spoken QA drops by only 1 point (83% to 82%). We train decisions with cross-entropy on the read-out answer token, instead of the usual transcript distillation, whose teacher never hears the voice, and keep distillation for content. Encoders and LLMs are interchangeable; we release weights, training recipe, a batched-inference pipeline for full-duplex serving and a bilingual spoken-QA set.

cs.AI↗

On Perfect Divisibility of Bull-Free Graphs Without Long Paths

A graph $G$ is {\em perfectly divisible} if, for every induced subgraph $H$ of $G$, $V(H)$ can be partitioned into $A$ and $B$ such that $H[A]$ is perfect and $ω(H[B])<ω(H)$. Chudnovsky and Sivaraman [J. Graph Theory \textbf{90} (2019) 54-60] proved that every ($P_5$, bull)-free graph is perfectly divisible, while Chen and Xu [Discrete Appl. Math. \textbf{372} (2025) 298-307] proved the same for ($P_7,C_5$, bull)-free graphs. We extend these results by proving that every ($P_8,C_5$, bull)-free graph is perfectly divisible and that, letting $F$ denote the Grötzsch graph, a ($P_6$, bull)-free graph is perfectly divisible if and only if it is $F$-free.

math.CO↗

Bidirectional Code Reuse in Software Redesign: An Action Research Study of Static Analyzers

Software redesign preserves functionality while improving quality attributes, but manual reuse of code and tests is costly and error-prone, especially in cross-repository redesigns. Focusing on static analyzers where cross-repository redesign needs often arise, we conduct a bidirectional study of the ongoing Soot/SootUp redesign using an action research methodology that combines empirical investigation with validated open-source contributions. Our study reveals: (1) non-linear migration that necessitates bidirectional reuse, (2) deferred reuse via TODOs, (3) neglected test porting, and (4) residual bug propagation during migrations. We identify tracking corresponding code and tests as the key challenge and address it by retrofitting clone detection to derive code mappings between original and redesigned projects. Guided by semantic reuse patterns derived from our study, we propose the Semantic Alignment Score (SAS), which incorporates semantic cues from preserved identifiers, API documentation, and comments. Evaluations on three redesigned project pairs (Soot/SootUp, FindBugs/SpotBugs, and ANTLR3/ANTLR4) show that SAS improves average F1 for code mapping detection by up to 0.34 on our manually labeled benchmark of 1,805 method pairs, while strong traditional detectors approach or surpass LLM detectors. The code mapping analysis also uncovered ongoing maintenance needs, leading to five issues and 10 pull requests, of which eight have been merged.

cs.SE↗

OpWeave: Flexible Operator Disaggregation for Heterogeneous LLM Serving

LLM serving systems increasingly disaggregate inference into finer-grained stages, with recent approaches separating attention from FFN or MoE execution during decode. This operator-level disaggregated serving (ODS) can improve hardware matching and enable independent scaling, particularly across heterogeneous devices. However, existing systems fix operator boundaries and lack a unified characterization of when disaggregation reduces serving cost. We present OpWeave, an end-to-end framework for heterogeneous ODS. OpWeave provides an analytical cost model that bounds the gains of homogeneous and heterogeneous ODS over colocated serving. It jointly optimizes operator partitioning and deployment configuration through a regularity-aware planner that keeps the search tractable even for hybrid-attention models. A vLLM-based runtime executes the synthesized plans with flexible operator stages across heterogeneous device groups. In our evaluation, OpWeave reduces serving cost by up to $1.78\times$ on homogeneous and $1.89\times$ on heterogeneous GPU clusters relative to the best feasible baseline, while meeting latency SLOs.

cs.DC↗

Structure, Coloring, and Perfect Divisibility of $(P_2\cup P_4, C_3)$-Free Graphs

Goedgebeur and Schaudt [J. Graph Theory 87 (2018), 188-207] conjectured that every $4$-vertex-critical $(P_7,C_3)$-free graph belongs to a family of seven explicitly defined graphs. In this paper, we establish a structural theorem for connected $(P_2\cup P_4,C_3)$-free graphs. As a consequence, we prove that the Mycielski-Grötzsch graph is the unique $4$-vertex-critical graph in this class, thereby confirming the conjecture of Goedgebeur and Schaudt for $(P_2\cup P_4,C_3)$-free graphs. Our structural theorem also yields a characterization of the chromatic number of these graphs and an $O(n^4)$-time algorithm for deciding whether an $n$-vertex $(P_2\cup P_4,C_3)$-free graph is $3$-colorable. We further study perfect divisibility in the larger class of $(P_2\cup P_4,\text{bull})$-free graphs. We prove that a $(P_2\cup P_4,\text{bull})$-free graph is perfectly divisible if and only if it is Mycielski-Grötzsch graph-free. This result generalizes the main theorem of Deng and Chang [Graphs Combin. 41 (2025), 63].

math.CO↗

The ASTRID Simulation at z=0: From Massive Black Holes to Large-scale Structure

We present the $z=0$ results for the cosmological simulation ASTRID. Hosting $2\times 5500^3\approx$ 0.33 trillion particles in a box of $370\, {\rm Mpc}$ per side, ASTRID is one of the largest cosmological hydrodynamic simulations evolved to $z=0$. ASTRID features a large population of massive black holes (MBHs), covering a wide mass range $4\times10^{4}\sim 2\times 10^{11}\ M_{\odot}$. The adopted dynamical friction model provides a relatively accurate description of MBH dynamics, making ASTRID a powerful tool to study MBH growth and mergers in a cosmological context. ASTRID successfully captures the co-evolution of MBHs and their host galaxies, producing $M_{\rm BH}-M_{\star}$ and $M_{\rm BH}-σ$ relations in good agreement with observations. Notably, ASTRID generates scatter in these relations that is more consistent with observations than previous simulations, indicating a more realistic MBH diversity. The galaxy stellar mass function at $z=0$ is generally consistent with observational constraints. When dust attenuation is applied, the galaxy luminosity function also agrees well with observations, and the bimodality in galaxy colors is reproduced as well. ASTRID hosts a large population of massive galaxy groups and clusters: 7 halos have $M_{\rm 200c}>10^{15}\ M_{\odot}$, and 9709 halos have $M_{\rm 200c}>10^{13}\ M_{\odot}$. We quantify the stellar mass content in these halos, and find that the correlations between the stellar and halo mass match well with observational constraints. Finally, we present the $z=0$ power spectra of MBH and galaxies, as well as their bias with respect to the matter power spectrum. We find that MBHs with $M_{\rm BH}\geq 10^{8}\ M_{\odot}$ and galaxies with $M_{\star}\geq 10^{10.5}\ M_{\odot}$ serve as good tracers of large-scale structure.

astro-ph.GA↗

Skill-Guided Continuation Distillation for GUI Agents

Improving GUI agents typically relies on behavior cloning on expert trajectories. However, as the current policy deviates from the expert policy, it inevitably encounters policy-induced off-trajectory states during closed-loop execution, i.e., states that fall outside the expert trajectories. Since expert trajectories provide no demonstrations for these unseen states, such states receive no effective supervision, leaving the policy unable to select the correct action. To close this supervision gap, we propose Skill-Guided Continuation Distillation (SGCD), an iterative self-improvement framework. SGCD first runs the plain policy without skill guidance for a few steps to reach realistic off-trajectory states. From these states, a skill-guided policy then completes the task and produces successful continuations, which are mixed with expert trajectories to supply supervision over policy-induced off-trajectory states. The skills are extracted from both successful and failed rollouts, consisting of Continuation Plans, Critical Targets, Failure Traps, and Success Criteria. On OSWorld-Verified, SGCD improves the success rate of three base models from the low-30\% range to over 50\%, demonstrating its effectiveness and generality.

cs.AI↗

WebChallenger: A Reliable and Efficient Generalist Web Agent

Autonomous web navigation remains challenging for LLM agents, and the strongest generalist systems rely on proprietary reasoning models whose inference cost is prohibitive for the repetitive tasks where such agents would be most useful. We argue this gap stems not from insufficient model capability but from agent architectures that fail to replicate three human cognitive advantages: selective attention to relevant page regions, persistent memory of website structure, and procedural fluency with common interaction patterns. We introduce WebChallenger, a web agent framework that addresses each gap through architecture design rather than model scale, built around PageMem: a structured page representation deterministically constructed from the DOM that exposes each page as a hierarchy of semantic sections with short summaries. On this shared substrate we build three mechanisms that mirror the three cognitive advantages: a divide-and-conquer observation pipeline that lets the agent skim section summaries and extract details only from task-relevant regions; a lightweight exploration and memory system that traverses each website once to build a reusable map of pages and element behaviors; and compound action workflows that collapse common multi-step interactions into single agent actions, handling partial state changes automatically. Because all three operate over PageMem, the framework generalizes across websites without site-specific adapters. Using off-the-shelf open-weight models without fine-tuning, our system achieves 56.3% on WebArena, 48.7% on VisualWebArena, 51.0% on Online-Mind2Web, and 70.9% on WorkArena, approaching frontier proprietary systems at a fraction of the cost. Our code is released at https://github.com/jayoohwang1/webchallenger

cs.CL↗

Three-Dimensional Atomic-Scale Structural Transformation in a SrTiO3 Grain Boundary

Grain boundaries (GBs) in complex oxides play critical roles in governing their functional properties, which are intrinsically linked to their three-dimensional (3D) atomic configurations and local chemical environments that can deviate markedly from those of the bulk. However, the 3D atomic structures of GBs remain poorly understood due to the projection limitations of conventional (S)TEM. Here, using multislice electron ptychography, we resolve the 3D atomic structure of a Σ13(510)/[001] tilt GB in SrTiO3 with simultaneous visualization of both cation and oxygen columns. Depth-resolved reconstruction reveals pronounced structural inhomogeneity along the GB, uncovering a transition from the canonical symmetric configuration (STR1) to an asymmetric configuration (STR2) that is hidden in conventional projection imaging. Quantitative analysis of atomic-column intensities demonstrates that these two GB configurations possess distinct local chemical and vacancy distributions. By further mapping the atomic displacement fields, we reveal that the transformation between STR1 and STR2 proceeds via local atomic shuffling at the GB core and collective shear displacement in the adjoining grains, mediated by the step and dislocation character of the junction, respectively. Moreover, analysis of oxygen octahedral rotations reveals a strong dependence on the local atomic structure with pronounced asymmetry around the STR2 region. These findings establish a direct link among the 3D atomic structure, local chemical composition, and lattice order parameters at the GB, underscoring the critical importance of depth-resolved characterization in understanding and engineering GB-mediated functionalities in complex oxides.

cond-mat.mtrl-sci↗

Physical design of cold neutron direct geometry inelastic spectrometer at China Spallation Neutron Source

The Cold-Neutron Inelastic Spectrometer (CNIS) is a direct-geometry, time-of-flight instrument designed for China Spallation Neutron Source (CSNS) and optimized to probe low-energy lattice and magnetic excitations. The instrument integrates a long flight path with bent supermirror guides and an elliptical-focusing geometry to suppress high-energy background while improving cold-neutron delivery to the sample. A flexible multi-disk chopper suite provides pulse shaping, band selection and monochromatization, enabling multi-$E_\textrm{i}$ operation. Modular features, including an interchangeable high-focusing guide insert, radial collimation and a vacuum ``airbox'' for simplified sample-environment integration, enhance signal-to-noise and operational versatility. Through combined flight-path and chopper optimization, CNIS achieves excellent routine-mode energy resolution and can reach approximately $\sim 1\%$ in a dedicated high-resolution configuration. CNIS is planned to commence user operation in 2029, offering a highly flexible platform for cold-neutron inelastic scattering studies.

physics.app-ph↗

HyMem: Hybrid Memory Architecture with Dynamic Retrieval Scheduling

Large language model (LLM) agents demonstrate strong performance in short-text contexts but often underperform in extended dialogues due to inefficient memory management. Existing approaches face a fundamental trade-off between efficiency and effectiveness: memory compression risks losing critical details required for complex reasoning, while retaining raw text introduces unnecessary computational overhead for simple queries. The crux lies in the limitations of monolithic memory representations and static retrieval mechanisms, which fail to emulate the flexible and proactive memory scheduling capabilities observed in humans, thus struggling to adapt to diverse problem scenarios. Inspired by the principle of cognitive economy, we propose HyMem, a hybrid memory architecture that enables dynamic on-demand scheduling through multi-granular memory representations. HyMem adopts a dual-granular storage scheme paired with a dynamic two-tier retrieval system: a lightweight module constructs summary-level context for efficient response generation, while an LLM-based deep module is selectively activated only for complex queries, augmented by a reflection mechanism for iterative reasoning refinement. Experiments show that HyMem achieves strong performance on both the LOCOMO and LongMemEval benchmarks, outperforming full-context while reducing computational cost by 92.6\%, establishing a state-of-the-art balance between efficiency and performance in long-term memory management.

cs.AI↗

FlowForge: A Staged Local Rollout Engine for Flow-Field Prediction

Deep learning surrogates for CFD flow-field prediction often rely on large, complex models, which can be slow and fragile when data are noisy or incomplete. We introduce FlowForge, a staged local rollout engine that predicts future flow fields by compiling a locality-preserving update schedule and executing it with a shared lightweight local predictor. Rather than producing the next frame in a single global pass, FlowForge rewrites spatial sites stage by stage so that each update conditions only on bounded local context exposed by earlier stages. This compile-execute design aligns inference with short-range physical dependence, keeps latency predictable, and limits error amplification from global mixing. Across PDEBench, CFDBench, and BubbleML, FlowForge matches or improves upon strong baselines in pointwise accuracy, delivers consistently better robustness to noise and missing observations, and maintains stable multi-step rollout behavior while reducing per-step latency.

cs.LG↗