arXiv subjects

Yuan Tian

Publications and source records attributed to Yuan Tian.

At least 19 recordsLinked to original sources

arXiv2026

A Fine-Grained Empirical Study of Upstream Security Events and Downstream Mitigation in Maven

Open-source vulnerability mitigation spans upstream libraries and the downstream projects that depend on them. Upstream, a vulnerability progresses through a public report when one exists, a fix, a fixed release, and NVD disclosure. Downstream, a project mitigates it by adopting the fixed release, and the timeliness of this adoption is measured relative to these events. Prior work studies the two sides separately, so it is unclear whether downstream exposure stems from delayed upstream release, delayed downstream adoption, or both. We address this gap by following 832 CVEs in the Maven ecosystem from the upstream fix to its adoption by each exposed downstream project. We classify upstream fixes as \textit{silent} or \textit{transparent}, recover the order of report, fix, release, and NVD disclosure, and link vulnerable artifacts to their downstream dependents. We then compare this order with downstream adoption and dependabot usage. We find that delay arises on both sides. Upstream, the fixed release comes before NVD disclosure for about 82\% of CVEs, for both silent and transparent fixes. For the other 17.9\%, downstream projects have no fix to adopt for days to more than four months after disclosure. Downstream, exposure is concentrated, with the top 10\% of CVEs accounting for 72.1\% of it. Of 1,677 observed adoptions, 85.3\% happen after NVD disclosure, rising to 88.3\% for silent fixes released before disclosure. Most adoptions come through routine dependency updates. Adoptions with dependabot version updates have a median release-to-adoption delay of 10.4 days, compared with 110.6 days without dependabot. Yet version updates were configured at adoption time for only 8.5\% of total adoptions. We translate these findings into implications for upstream maintainers, downstream developers, and dependency-automation tools.

cs.SE↗

arXiv2026

When Skills Tell Agents Not to Ask: A Large-Scale Study of Oversight Directives in Agent Skills

Agent skills, reusable instruction packages centred on a SKILL.md file, extend AI agents on demand, and a single line in a widely copied skill reaches every agent that loads it. Some lines are oversight-loosening directives: they tell the agent to proceed without approval, to bypass checks with --no-verify, or to keep working without checking in. Although such directives can remove a checkpoint from every agent that loads them, no prior work has examined them. We present the first large-scale study of oversight-loosening directives, analyzing 1.53 million English skills from GitSkills. We split skills into units (sentences, bullets, table cells, and code lines), develop a codebook that labels oversight directives by polarity, mechanism, and scope, label 2,097 units by hand, and train a detector to scale the analysis. Our results show that about one in seventeen skills (about 88,000 skills) contains a loosening directive. These directives most often loosen oversight through permission flags, and most of them (70.1%) apply unconditionally rather than only in specific contexts such as CI. Compared with skills that have the same amount of oversight language, skills that run shell code, install packages, or use sudo have 1.7-1.8 times the odds of containing a loosening directive, an association that holds across all our robustness checks, and they do not compensate with more tightening. Loosening spreads through a few upstream skills rather than through edits by their adopters: 2.4% of the skill families that carry loosening account for half of all copies of loosening-carrying skills, and 1.1% of distinct loosening units for half of all loosening occurrences. Open-source skill scanners catch at most 46% of loosening directives, with a precision below 0.4, while the LLM behind them, prompted with our codebook instead, catches 61% at a precision of 0.71.

cs.CR↗

arXiv2026

Planar energetics do not determine twin persistence in nanocrystalline aluminium

Planar energetics are the standard descriptors of twinnability in face-centred cubic metals and the usual validation targets for interatomic potentials. Here we show that they do not determine whether deformation twins persist in nanocrystalline aluminium. Two machine-learned potentials reproducing the same zero-load slip curves and closely spaced layered energy surfaces predict, from a common separated-fault precursor under identical loading, retained twin coverage of 64\% and 14\%. The divergence traces to forces within distorted partial cores acting along the trailing translation that would remove an outer fault. Planar energetics do set the margin: an outer fault more than halves the incremental barrier for middle-plane shear, yet twin completion and fault removal remain separated by only 10--21 mJ m\textsuperscript{-2}. Whether an assembled twin persists is decided instead by the orientation of local traction relative to the full-slip bisector and by the arrival order of partial-slip fronts from the grain-boundary network.

cond-mat.mtrl-sci↗

arXiv2026

Retention-constrained fine-tuning of a pre-trained aluminium MACE potential for non-equilibrium defect dynamics and plasticity

Adapting a pre-trained machine-learning interatomic potential for non-equilibrium defect dynamics and plasticity requires improving defect accuracy without degrading the near-equilibrium elastic and vibrational response. Here we achieve this balance for the aluminium MACE potential QP17 by integrating conditional readout uncertainty, iterative full-parameter updates and protocol-compatible relative-energy anchoring. Sensitivities over 144 linear readout weights guide acquisition, with within-batch covariance updates and evolving neural representations. After four active-learning rounds (32 acquired configurations and 400 gradient updates), force-uncertainty acquisition reduces mean development force error from 22.43 to 21.52 meV/Angstrom, outperforming random sampling (22.70), geometry-diversity selection (22.46) and uniform training across all 104 candidate labels (22.71) under the same update budget. All three force-uncertainty runs satisfy bulk-mechanical retention criteria: historical relative-energy error decreases from 0.375 to 0.207 meV/atom, while harmonic phonon root-mean-square drift remains below 0.05 THz. Matched objective ablations reveal that relative-energy anchoring suppresses harmonic drift. Out-of-sample evaluations across planar faults, twin boundaries and grain boundaries demonstrate transferable gains in defect energies and deviatoric stresses, even though virial stresses are never fitted. A closed-loop molecular dynamics and density functional theory (MD-DFT) pilot confirms that automated acceptance gating intercepts updates that violate historical elastic-group energy or vibrational tolerances. Together, these components establish a retention-constrained fine-tuning protocol for improving plastic-deformation observables while preserving the tested bulk mechanical response.

cond-mat.mtrl-sci↗

arXiv2026

Stationary Measures and Boundary Correlations for Multispecies Half-Line ASEP

We construct stationary measures of the multispecies ASEP on the positive half-line, with injection of highest-priority particles at the left boundary, using vertex models. Our results extend existing constructions of Martin and Aggarwal-Nicoletti-Petrov for the full-line ASEP to the half-line. For the two-species process, we compute the joint distribution of occupation colors at the first two sites. We also partially address the connection with the half-space ASEP speed process.

math.PR↗

arXiv2026

Kagome goldene with flat bands and Dirac nodal line fermions via line-graph epitaxy

The kagome lattice has emerged as a promising platform for investigating exotic quantum phases. However, achieving a single-atomic-layer kagome lattice in elemental materials remains a significant challenge. Here, we introduce line-graph epitaxy, a novel approach that enables the atomic-scale synthesis of goldene, a monolayer of elemental gold atoms arranged in a kagome lattice. Through scanning tunneling microscopy/spectroscopy (STM/STS), and density functional theory (DFT) calculations, we demonstrate the formation of kagome goldene, featuring a flat band with a van Hove singularity approximately 1.1 eV below the Fermi level, signaling strong electron correlation effects. Notably, the flat band is disrupted at the zigzag edges of goldene nanoflakes, revealing substantial edge effects. Furthermore, our calculations show that weak interlayer interactions between goldene and the underlying Au2Ge substrate generate dual Dirac nodal lines through a proximity effect. These findings offer not only a novel strategy for constructing elemental kagome lattices, but also a generalizable framework for fabricating and controlling line-graph materials. This research advances the exploration of quantum phases driven by strong correlations and the design of materials for next-generation quantum technologies.

cond-mat.mes-hall↗

arXiv2026

From phase analysis to phase tracking: a robust approach to atomic-scale strain mapping

Atomic-scale strain governs the electronic, ferroic and mechanical properties of materials, and atomic-resolution microscopy measures it most directly. Geometric phase analysis (GPA) is the standard tool, but it fails when a Bragg reflection moves outside its fixed Fourier aperture under large strain or rotation, or in misoriented grains. Here we introduce geometric phase tracking analysis (GPTA), whose Fourier aperture tracks the Bragg reflection at each pixel in real space. A narrow aperture can therefore reject noise without losing a displaced reflection. On 13,795 simulated and experimental atomic-resolution images under various imaging conditions, GPTA reduces GPA's failure rate from 27% to 11.5% with up to 335 times better accuracy. It also gives strain and dislocation maps across grains, resistance to dynamical-scattering extinction, and a confidence map, capabilities that GPA lacks. Tracking the aperture combines reciprocal-space dose efficiency with real-space spatial resolution and extends atomic-scale strain metrology to beam-sensitive materials and challenging imaging conditions.

cond-mat.mtrl-sci↗

arXiv2026

Attention-Aligned Reasoning for Large Language Models

Large Language Models (LLMs) tend to generate a long reasoning chain when solving complex tasks. However, as the reasoning chain extends, critical intermediate steps and the original prompt will be buried in the context, receiving insufficient attention and leading to errors. In this work, we present ATAR, a novel reasoning method that leverages the inherent reasoning structure to steer LLM attention. Our experiments show that ATAR outperforms SOTA methods across six benchmarks, achieving up to 15.39% absolute improvement. Furthermore, with ATAR, "non-reasoning" models achieve comparable or even better performance compared to reasoning models of the same size in most benchmarks. Finally, our ablation studies show that the attention alignment component contributes significantly, and that these improvements are persist under different attentionsteering backends.

cs.CL↗

arXiv2026

LG-Bench: A Graph-Structured Evaluation Benchmark for Life Sciences

Traditional evaluation benchmarks reduce inherently interconnected scientific knowledge in life sciences into flat lists of questions, disregarding the underlying topological structure of the knowledge. We introduce, the first graph-structured benchmark for life sciences, featuring over 10,000 high-quality multiple-choice questions across medicine, biology, and chemistry. Our approach constructs a weighted evaluation graph using bidirectional matching and semantic similarity algorithms, where nodes represent questions and edge weights capture their semantic relationships. Leveraging this graph topology, we design two novel evaluation metrics. The Global Coherence Score (GCS) measures a model's consistency within semantically related neighborhoods, while Knowledge Balance Score (KBS) analyzes how model errors are distributed across the graph to reveal conceptual blind spots. LG-Bench facilitates fine-grained comparison of LLMs by surfacing differences in conceptual coherence and patterns of knowledge organization across models. Our framework shifts the evaluation paradigm from flat accuracy metrics to structure-aware analysis, offering a new lens for diagnosing and improving LLM performance in the life sciences domain.

cs.AI↗

arXiv2026

RogueMerge: Robust and Unified Attacks against LLM Model Merging

Model merging composes specialized capabilities into a single LLM by aggregating task-specific LLMs sourced from unverified public platforms, exposing a critical supply-chain attack surface: Because any malicious behavior can be encoded into a task vector, and merging grants third-party vectors direct write access to model weights, an attacker-provided task vector can enable or amplify diverse downstream threats. Prior work only examines backdoor attacks against model merging for classifiers using static arithmetic heuristics, which fail to effectively achieve diverse attacks on generative LLMs for three reasons: (i) LLMs rely on autoregressive decoding, where even a minor parameter drift introduced by merging compounds across tokens and rapidly degrades the attack. (ii) Attackers have no knowledge of the victim's merging configurations, causing a static attack vector optimized in isolation to be easily diluted or neutralized. (iii) Practical threat induction must generalize to attack prompts unseen during optimization, which static vectors cannot adequately encode. In light of these gaps, we present RogueMerge, the first unified attack framework addressing all three challenges. To handle autoregressive generation, we replace static arithmetic with a joint optimization that explicitly enforces attack success after merging. To better simulate unknown merging settings, we formulate attack injection as a stochastic min-max optimization problem and solve it via meta-gradient descent. To generalize across heterogeneous attack prompts, we employ distributionally robust optimization and derive a tractable first-order Taylor approximation at LLM scale, with a provable error bound. Across four threats, six merging algorithms, and over 170 merged LLMs, RogueMerge consistently outperforms prior attacks. It also remains stable across diverse merging settings and resists standard defenses.

cs.CR↗

arXiv2026

Specialized machine learning force fields for materials dynamics

Machine learning interatomic potentials (MLIPs) are transforming atomistic simulations by accessing unprecedented length and time scales. While pretrained equivariant graph neural networks achieve robust zero-shot performance for near-equilibrium properties across broad chemical spaces, their translation to complex materials dynamics remains fundamentally challenged by out-of-distribution reactive states, representation biases, and computational scaling limits. In this Review, we examine how physics-driven specialization extends the applicability of MLIPs to complex dynamical systems. We systematically evaluate the structural trade-offs in MLIP design: the undersampling of highly strained configurations, the prohibitive computational overhead of high-order message-passing architectures, and the necessity of nonlocal interactions for open and field-coupled systems. Through four demanding application contexts - electrified interfaces, compositionally fluctuating open systems, multiphase evolution, and large-scale fracture - we establish a framework for observable-specific validation. Highlighting the complementary roles of universal foundation models and task-specific potentials, we emphasize that targeted adaptations must be rigorously benchmarked against intended observables. We conclude with a roadmap for developing physically consistent, hardware-aware force fields that seamlessly connect electronic-structure accuracy to macroscopic materials phenomena.

cond-mat.mtrl-sci↗

arXiv2026

Harness Evolution Hits a Ceiling: When Weight Training Should Begin

Improving a long-horizon LLM agent means evolving the harness around a frozen model or training its weights. We let a self-evolving harness make the system stronger first, then cross seed and evolved harnesses with base and trained weights to learn which gains the trained model keeps and which still need the runtime. We show that the right lever can be read off the agent's failure composition: labelling failed trajectories by the first signal that fires separates process failures (blocked calls, loops, exhausted step budgets) from content failures (a delivered plan that is poor). Harness evolution repairs the former, the behaviour it instils can be trained into the weights, and content failures are what weight training is for. On DeepPlanning, a self-evolving harness loop lifts the held-out score of Qwen3.5-4B from 0.16 to 0.30 and of Qwen3.5-9B from 0.32 to 0.44; for 4B, held-out delivery rises from 55% to 90% while content failures are left for the weights. LoRA adapters trained on evolved-harness trajectories internalise the gain: under the original harness they add +0.13 on held-out tasks for both sizes; on 4B they stack with the harness to more than double the held-out score, and on 9B the adapter alone matches the full evolution line, cutting content failures from a quarter of trajectories to one in twenty. A placebo adapter trained on answer-shuffled trajectories falls below the base model. The loop transfers to WebArena-Lite (+0.09 on 117 unseen tasks), where the gain lives in what the model sees and adapters do not add to it. The result is a diagnose-then-intervene rule applied twice: read the failure composition to choose between harness and weights, then read what the accepted edits changed to decide which gains to train in. Scores are four-rollout means against fresh anchors, same-night except where marked, across eight models from six families and two benchmarks.

cs.AI↗

arXiv2026

SafetyFlow: An Agent-Flow System for Automated LLM Safety Benchmarking

The rapid proliferation of large language models (LLMs) has intensified the requirement for reliable safety evaluation to uncover model vulnerabilities. To this end, numerous LLM safety evaluation benchmarks are proposed. However, existing benchmarks generally rely on labor-intensive manual curation, which causes excessive time and resource consumption. They also exhibit significant redundancy and limited difficulty. To alleviate these problems, we introduce SafetyFlow, the first agent-flow system designed to automate the construction of LLM safety benchmarks. SafetyFlow can automatically build a comprehensive safety benchmark in only four days without any human intervention by orchestrating seven specialized agents, significantly reducing time and resource cost. Equipped with versatile tools, the agents of SafetyFlow ensure process and cost controllability while integrating human expertise into the automatic pipeline. The final constructed dataset, SafetyFlowBench, contains 23,446 queries with low redundancy and strong discriminative power. Our contribution includes the first fully automated benchmarking pipeline and a comprehensive safety benchmark. We evaluate the safety of 49 advanced LLMs on our dataset and conduct extensive experiments to validate our efficacy and efficiency.

cs.CL↗

arXiv2026

Tight Privacy Audit in One Run

In this paper, we study the problem of privacy audit in one run and show that our method achieves tight audit results for various differentially private protocols. This includes obtaining tight results for auditing $(\varepsilon,δ)$-DP algorithms where all previous work fails to achieve in any parameter setups. We first formulate a framework for privacy audit \textit{in one run} with refinement compared with previous work. Then, based on modeling privacy by the $f$-DP formulation, we study the implications of our framework to obtain a theoretically justified lower bound for privacy audit. In the experiment, we compare with previous work and show that our audit method outperforms the rest in auditing various differentially private algorithms. We also provide experiments that give contrasting conclusions to previous work on the parameter settings for privacy audits in one run. Our implementation is at https://github.com/zihangxiang/tight-privacy-audit-in-one-run.git

cs.CR↗

arXiv2026

BurnRiSc: Toward Non-Invasive Burnout Screening in Open Source from Public Repository Signals

Burnout is a chronic occupational syndrome, and open source is close to a worst case for it: maintainers absorb unbounded demand with no manager to reallocate work and no organization to notice decline. The cost is not only personal. Burnout precedes withdrawal, and in projects sustained by a handful of maintainers, one departure can break infrastructure that thousands of downstream systems depend on. Yet the field has no way to see it coming: self-report inventories, the only existing measure, miss exactly the contributors most in need of detection and cannot be applied retroactively, so the field cannot even ask how common burnout is or what helps. We present BurnRiSc, a framework that operationalizes the Oldenburg Burnout Inventory's two dimensions, exhaustion and disengagement, as 14 behavioral and linguistic signals computed from GitHub activity and scored against each contributor's own history. The signals aggregate into two weighted dimension scores, with weights learned from labeled cases, and average into a monthly Burnout Risk Score (BRS). In a preliminary evaluation across 68 contributors in ten repositories (ten disclosed burnout cases, twelve comparable-volume collapses, and 46 comparison contributors), sustained BRS elevation precedes 6 of 10 disclosures by 6-15 months, 8 of 10 when adding peak BRS as a second criterion, and 10 of 10 over any prior time frame. We thus present BurnRiSc as evidence that burnout is screenable from public data.

cs.CY↗

arXiv2026

CARES: A Conversational AI System for Regulation-Grounded Safety Reporting in Construction Education

Construction safety reporting often relies on manual logs and static templates that provide limited feedback and leave daily activities disconnected from relevant regulations. This paper introduces CARES (Conversational AI Reporting for Enhanced Safety), a conversational AI system that integrates regulatory guidance into daily reporting to support construction safety education. CARES combines proactive multi-agent dialogue, retrieval-augmented generation (RAG), and automated report generation. The system guides users through reporting tasks, retrieves relevant regulatory passages using hybrid retrieval, and converts conversations into structured daily reports. Regulatory sources and the evolving report are displayed alongside the dialogue to support user review and correction. A preliminary evaluation involving 15 construction management students assessed retrieval quality, response faithfulness, and conversational relevance. CARES achieved an overall faithfulness score of 0.74 and an answer relevance score of 1.00, while initial retrieval ranking remained an area for improvement. These results provide preliminary evidence of the technical feasibility of regulation-grounded conversational reporting. The study highlights opportunities to integrate regulatory knowledge into routine documentation, with future work needed to evaluate effects on report quality, safety awareness, and learning outcomes.

cs.HC↗

arXiv2026

Surrogate-Assisted Genetic Programming with Phenotypic Characterisation in Dynamic Multi-Mode Project Scheduling

Dynamic multi-mode resource-constrained project scheduling requires decisions to be made under precedence constraints, limited resources, multiple execution modes, and uncertain activity durations. Genetic programming (GP) can automatically evolve heuristic rules for such problems, but its simulation-based fitness evaluation is computationally expensive. This study investigates phenotypic characterisation (PC) in surrogate-assisted GP to evolve higher-quality scheduling heuristics under a fixed budget of full simulation-based fitness evaluations. A key question is how GP individuals should be encoded into phenotypic characterisations to support effective fitness estimation. To answer this question, three PC encoding schemes with different levels of information richness are designed: priority-value encoding, which preserves raw rule outputs; rank encoding, which captures candidate ordering; and binary encoding, which represents final scheduling decisions. These encodings are combined with different distance metrics to measure behavioural similarity between GP individuals. The experimental results show that binary encoding with Euclidean distance provides the most effective and robust surrogate guidance. Further analyses show that surrogate estimation accuracy alone does not fully explain the performance differences. The PC representation also determines how effectively phenotypically redundant offspring are removed and how much behavioural diversity is retained after preselection. Ablation experiments further demonstrate that duplicate removal and surrogate preselection provide complementary benefits, with their combination producing the largest improvement. These findings highlight that effective surrogate-assisted GP depends not only on identifying promising offspring, but also on controlling redundancy and preserving useful diversity during evolutionary search.

cs.NE↗

arXiv2026

StreamWSR: Streamable and Lightweight Waveform-Domain Neural Speech Super-Resolution

This paper proposes StreamWSR, a Streamable neural Waveform-domain model for speech Super-Resolution (SR). By adopting a fully causal architecture with compact frame-level waveform representation, the proposed StreamWSR supports zero-look-ahead streaming inference while avoiding vocoder-based reconstruction and explicit phase prediction. Specifically, StreamWSR downsamples the input waveform into a compact frame-level representation using strided causal convolutions. Then, a lightweight causal long-short-term modeling backbone is employed to capture both local waveform structures and long-range historical dependencies under causal constraints. Finally, the modeled output is converted back to the waveform domain through a causal transposed-convolution and combined with the input waveform via a residual connection to generate the final high-resolution speech. Experimental results on 16 kHz speech SR show that StreamWSR achieves competitive or superior speech quality and intelligibility compared with representative waveform- and spectrum-based baselines, while maintaining a zero-look-ahead streaming advantage with only 9M parameters and 2G FLOPs.

eess.AS↗