arXiv:2607.13707v1 Announce Type: new
Abstract: Studies of bias in LLM-as-judge systems typically build synthetic corpora by prompting an LLM to generate a hallucinated answer to pair with a factual one, then presenting both to a judge. We report a case in which this generation step silently failed, and use it to argue that the failure mode is structural rather than incidental. In a multilingual (Turkish/English) faithfulness-judgment corpus, a decoding-budget parameter shared between judging and generation calls truncated one producer's hallucinated answers to a few words. The resulting items produced a large, statistically robust effect: a 32-point cross-lingual collapse in one judge's selection accuracy, replicated from N=50 to N=500, explained by a three-layer mechanistic account, and confirmed by a controlled producer-swap experiment, none of which was real. The effect vanished to ceiling once the shared parameter was corrected, and only manual reading of the raw generations, not any aggregate statistical check, exposed the fault. A second measured bias (Markdown-formatting preference) was not fabricated but distorted by the same fault, its magnitude and in one case its sign shifting with stimulus length, a mode aggregate metrics cannot distinguish from the first. We frame the underlying vulnerability using the test oracle problem: corpora whose negative examples are LLM-generated carry no mechanical way to verify item integrity, while corpora built by deterministic perturbation of a gold answer carry an item-level oracle for free. A positive control supports this claim directly: an analogous fault injected into a minimal perturbation-based corpus is caught with 100% accuracy by a zero-cost, zero-human gold-to-negative string comparison. We close with a validation protocol, derived from our own case, for analysts working in the oracle-less regime that we argue describes most contemporary multilingual LLM-as-judge corpora.
Science Journals
arXiv:2607.13407v1 Announce Type: new
Abstract: Semantic operator systems extend data processing with natural-language interfaces, supporting operations such as semantic filtering, mapping, and joining. Existing systems commonly execute these operators through interpretation-based execution: for each row, record, or candidate pair, an LLM is invoked to interpret the semantic intent and produce an output. Although expressive, this places expensive LLM calls inside the data-processing loop, causing high latency, monetary cost, and limited scalability.
We propose compilation-based execution of semantic operators. Instead of using an LLM as a runtime interpreter for every data item, we invoke it once during compilation to translate a semantic operator specification into deterministic executable code. The generated code serves as a compiled physical operator that approximates the behavior of the original LLM-based operator and runs locally over the dataset without per-row or per-pair LLM calls.
We instantiate this approach for semantic filter, semantic map, and semantic join operators, compare it with LLM-based interpreted execution, and integrate it into an existing semantic operator system. Preliminary results show that compilation-based execution substantially reduces execution time and LLM calls while preserving much of the output quality. More broadly, we argue that semantic operator systems should treat LLMs not only as runtime executors, but also as semantic compilers that generate efficient executable plans, opening new research opportunities at the intersection of database query processing, program synthesis, and LLM-powered data systems.
arXiv:2607.13622v1 Announce Type: new
Abstract: Miniature hybrid underwater gliders have attracted increasing attention for long-endurance ocean observation and confined-space inspection. Large-range wing reconfiguration offers a promising yet largely unexplored approach for simultaneously enhancing maneuverability and shape adaptability in constrained underwater environments. However, such morphing introduces substantial challenges in mechanical integration, dynamic modeling, and hydrodynamic characterization. This paper presents FoDeGlider, a miniature hybrid underwater glider equipped with two independently actuated wings capable of large-range folding and deflection. To capture configuration-dependent variations in mass distribution, center-of-geometry location, and hydrodynamic loading, a multibody dynamics model is developed by treating wing configuration as a structural variable. A composite rigid body algorithm (CRBA)-based projection formulates the composite inertia, wrench transformations, and component-level hydrodynamics into a unified Fossen-form dynamic model applicable to arbitrary wing configurations. A sequential parameter-identification framework is further proposed to estimate fuselage and wing hydrodynamic coefficients, resulting in an open benchmark dataset for model identification and validation. Extensive experiments are conducted, the results of which demonstrate accurate dynamic modeling and parameter identification across diverse morphing configurations. Gate traversal experiments further validate FoDeGlider's ability to actively reconfigure its morphology during locomotion, enabling enhanced navigation in confined underwater environments.
arXiv:2607.13257v1 Announce Type: new
Abstract: Industry 5.0 emphasises human-centric industrial system design, placing additional demands on modelling tools. Multi-level modelling (MLM) can directly represent three or more abstraction levels, but this comes at the cost of more complex semantic constraints that model correctness depends on. Large Language Models (LLMs) have been increasingly studied in model-driven engineering, but this evidence rests entirely on two-level modelling tasks, and whether it generalises to MLM, whose semantics differ in kind, remains untested. This paper presents the first empirical study of this question. We have three commercial LLMs (GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro) generate multi-level models for the MULTI Warehouse Challenge in the SLICER language under six prompting strategies, yielding 90 generated models compared against a manually validated reference using fourteen metrics. Syntactic correctness is within reach, but semantic correctness is only partially achieved, with Instantiation/Specialisation Correctness ranging from 52% to 79%. Models reproduce content stated explicitly in the task text, but rarely complete structure and constraints the text implies without stating. Prompting strategies trade off precision against completeness, and self-checking functions mainly as a rule checker rather than reliably improving alignment with the reference design. Among the three LLMs, Claude shows the most balanced profile. These results clarify the boundaries of current LLM capability for MLM and inform the design of human-centred, AI-assisted modelling workflows for Industry 5.0.
arXiv:2607.13261v1 Announce Type: new
Abstract: Aromatic organic solutes in water exhibit a delicate balance between hydrophobic solvation and directional O-H$\cdots \pi$ hydrogen bonds, yet widely used force fields and state-of-the-art density functional approaches struggle to provide a consistent picture of these pivotal interactions. We introduce a data-efficient upfitting strategy to train a machine learning interatomic potential (MLIP) based on the graph atomic cluster expansion for aqueous aromatic molecules with CCSD(T) accuracy for condensed phase simulations, using only finite molecular clusters. We apply our method to aqueous toluene (C$_6$H$_5$CH$_3$). The resulting CCSD(T)-quality MLIP reproduces coupled cluster energies and forces in bulk and reveals that commonly employed methods do not capture the crucial balance between hydrophilic and hydrophobic solvation, distorting the interactions of aromatic molecules with their environment. Representative biomolecular force fields substantially understructure the hydrophobic solvation shell and misorient interfacial water, while overestimating $\pi$-contacts, yielding an inconsistent solvation balance. Even hybrid DFT and MP2 overestimate barriers to breaking of water-$\pi$ hydrogen bonds. Our workflow provides a practical, general route to CCSD(T)-quality condensed-phase simulations of aqueous solutions, and thus constructed interaction potentials now open the door to consistent, highly accurate benchmark studies of $\pi$-contacts and hydrophobic effects in biomolecular contexts such as solvation of proteins and DNA in aqueous environments.
arXiv:2607.13433v1 Announce Type: new
Abstract: Automated essay scoring (AES) research has largely focused on cross-prompt generalization, where essays from unseen prompts are scored while the scoring criteria are typically held constant. In practice, however, educators may revise or even introduce new rubrics in their scoring task, to evaluate different aspects of essays. We study cross-rubric generalization: training on essays labeled under one set of rubrics and evaluating on previously unseen rubrics, which target different aspects of the essay. We use a Large Language Model (LLM) fine-tuning framework with two components: rubric-agnostic intermediate representations, called traits, and target-essay supervision under seen rubrics during training. On an AES dataset augmented with multiple rubric-defined labels of student critical thinking skills, we find that traits improve macro F1 by 5.0% over a baseline without traits in the hardest setting, where both target rubrics and target essays are unseen during training. We further find that increasing target-essay supervision improves performance, with our best fine-tuned open-source Llama-based model outperforming GPT-5-mini prompting by 2.1% macro F1 and trailing GPT-5 by 1.9%. These results show that trait-based intermediate structure and controlled supervision improve generalization to unseen rubrics.
arXiv:2607.13449v1 Announce Type: new
Abstract: 6-DoF pose estimation is a critical task in autonomous rendezvous and proximity operations. In the case of an unknown target, this task becomes challenging as it shall be paired with the reconstruction of the target shape model. In this article, we propose a novel framework for single-shot shape and pose estimation of unknown spacecraft objects. Given a single image, we first reconstruct a 3D shape model of the target, then estimate the relative six-degrees-of-freedom pose by learning dense 2D-3D correspondences. The image features are extracted using a frozen DINOv3 vision transformer, while the geometric features are computed from the reconstructed point cloud using a trainable dynamic graph convolutional neural network encoder. A dual-stream transformer matcher refines descriptors through alternating self- and cross-attention, producing soft correspondences that are passed to a Perspective-$n$-Point solver for pose recovery. We evaluate the method on the SPE3R dataset and consider FoundationPose as a representative baseline for current state-of-the-art capabilities. Results show reliable pose estimates achieving 0.157 degrees mean pointing error using only a single image and reconstructed geometry, demonstrating strong generalization to unseen spacecraft.
arXiv:2607.13403v1 Announce Type: new
Abstract: Efficient task allocation for large-scale Heterogeneous Multi-Robot Systems (HMRS) is critical, yet dealing with complex temporal logic tasks in partially known environment (PKE) remains a computational bottleneck. Existing approaches often struggle to balance exploring uncertain regions and exploiting known resources, while also suffering from exponential computational complexity. To address these issues, this paper presents a robust planning framework that simultaneously handles high-level logical constraints and environmental uncertainty without sacrificing scalability. We formulate the problem as a min-max regret optimization, proposing a Region-Binding Atomic Proposition (RbAP) to capture resource uncertainty within the automaton structure. To solve this, we propose the Extended Planning Decision Tree (E-PDT) equipped with a novel Regret-based Branch-and-Bound (BnB) strategy. Unlike traditional methods that rely on prior probabilities or worst-case analysis, our approach dynamically prunes suboptimal policies, effectively balancing the need for information gathering (exploration) and task completion (exploitation). Theoretical analysis confirms the feasibility and completeness of our approach. Extensive numerical and physical experiments demonstrate that the proposed framework achieves near-linear scalability with respect to the number of robots and types, significantly outperforming MILP-based baselines in both solution quality and computational efficiency.
arXiv:2607.13656v1 Announce Type: new
Abstract: Image-based indoor scene relighting remains challenging due to the complex interplay between cluttered geometry and local illumination, requiring precise modeling of light position, color, and intensity. Existing data-driven methods implicitly learn this relationship via paired multi-illumination datasets. Nevertheless, this data is costly and fails to scale, which is essential for accurate light-source-level control. Conversely, inverse-rendering methods reduce the data dependency by incorporating physical priors; however, they lack the robustness of intrinsic estimation in challenging conditions.
In this paper, we present FreeLit, a paired-free framework for controllable indoor relighting that explicitly manipulates light-source location, color, and intensity. Instead of relying on paired supervision, we construct a physics-guided illumination prior from intrinsic scene properties, generating a structured lightmap along with a pseudo-relit image to guide diffusion-based synthesis. To address instability in intrinsic estimation, especially in low-light scenes, we introduce a relighting-guided intrinsic stabilization strategy that enforces illumination-invariant reflectance through structure-aware distillation and consistency constraints. Furthermore, we propose controllability-oriented evaluation metrics to quantify alignment with user-specified illumination color and intensity. Experimental results demonstrate that FreeLit achieves stable, physically consistent, and controllable relighting, with improved robustness in low-light indoor scenes, without requiring paired supervision.
arXiv:2607.12785v2 Announce Type: replace
Abstract: Robot-assisted minimally invasive surgery (MIS) critically depends on reliable endoscopic perception for navigation and safety. However, conventional endoscopes provide only a limited field of view, leaving large portions of the surrounding anatomy unobserved. Recent neural rendering approaches, such as Neural Radiance Fields and 3D Gaussian Splatting, enable novel view synthesis from endoscopic videos, but their reliance on sparse observations often leads to severe artifacts when extrapolating beyond the training trajectory. In this work, we propose ExtraGS, a framework for enhancing endoscopic view extrapolation through diffusion-guided 3D Gaussian Splatting. Starting from an initial reconstruction, we introduce an uncertainty-guided virtual camera sampling strategy to actively explore blind spots and maximize information gain. The rendered views from these sampled locations are refined using a diffusion model to recover plausible anatomical structures, producing pseudo-observations that guide further optimization. To prevent the generated content from degrading reliable regions, we adopt a confidence-weighted fine-tuning strategy when incorporating these pseudo-observations. Extensive experiments on multiple public endoscopic datasets demonstrate that ExtraGS significantly reduces extrapolation artifacts and achieves state-of-the-art performance in endoscopic novel view synthesis.
arXiv:2607.13780v1 Announce Type: new
Abstract: We present space-time lensing by synthetic accelerated interfaces in dispersive media. Frequency-dependent group velocities map spectral components of a pulse onto distinct trajectories in the longitudinal space-time plane, while an accelerated interface reshapes their frequencies through time-dependent Doppler shifts. We develop an inverse-design procedure for the interface trajectory that focuses a quasi-monochromatic wave packet at a prescribed space-time event for an arbitrary dispersion relation, then extend the construction to polychromatic pulses. In its time-reversed form, the operation collimates a broadband pulse into a narrow-band wave. Moreover, cascading the forward and reverse operations enables a broadband pulse to propagate through an otherwise dispersive channel in a spectrally compressed state before being refocused. Full-wave simulations validate the focusing construction and demonstrate dispersion suppression.
arXiv:2607.13696v1 Announce Type: new
Abstract: Robots operating in human-robot collaboration must communicate not only their intended actions but also uncertainty arising from incomplete or ambiguous perception. This work introduces a mathematical framework for expressing perceptual uncertainty through robotic manipulator motion. Drawing on Laban Movement Analysis, robot behavior is organized in a Commitment-Vigilance state space that maps uncertainty-related states - confidence, curiosity, hesitance, fear, and inactivity - to distinct Laban Effort signatures. Five motion primitives - approach, pause, retreat, exploration, and oscillation - are then parameterized using eleven kinematic and geometric descriptors, including acceleration, pause and retreat characteristics, gaze angles, tilt, and shiver amplitude. A video-based human-subject study evaluated recognition of four expressive trajectories and the influence of individual descriptors on perceived intensity. Participants reliably identified the intended behavioral states, while several descriptors significantly modulated expressiveness. The results establish a perceptually grounded basis for encoding robot uncertainty in motion and support future autonomous trajectory generation using parametric movement representations for collaborative tasks in shared environments. Code, videos, questionnaire and appendices are available at "https://bit.ly/github-aou".
arXiv:2607.14027v1 Announce Type: new
Abstract: We propose SPECS, a genetic algorithm for automated analog circuit synthesis with joint topology and sizing optimization. SPECS is inspired by NeuroEvolution of Augmenting Topologies (NEAT), an evolutionary algorithm originally developed to synthesize neural networks. By reformulating the genome representation and adapting the genetic operators to the analog circuit domain, we successfully transfer the core principles of NEAT to analog circuit synthesis. Circuit-specific wiring constraints are incorporated to ensure valid and physically meaningful designs throughout the evolutionary process, and speciation is used to preserve innovation while maintaining population diversity. We evaluate the proposed method on a set of computational circuit synthesis tasks consisting of square, cube, square root, and cube root functions. Experimental results demonstrate that SPECS outperforms benchmark methods across all tasks in both solution quality and reliability. The synthesized circuits and their schematics are available in the supplementary repository.
arXiv:2607.14035v1 Announce Type: new
Abstract: Generative Engine Optimization (GEO) seeks to increase content's presence, likelihood of citation, or influence in answers produced by generative engines. Since the foundational GEO paper, the field has expanded rapidly, but terminology, metrics, and evidence standards remain heterogeneous. This critical survey reviews 45 studies selected under a November 2023-July 2026 publication window, including one earlier preprint published at EMNLP after the window opened, plus relevant RAG and evaluation work. We argue that GEO is not a single ranking task but a stochastic, partially observable pipeline spanning search activation, crawling and indexing, retrieval, reranking and context allocation, citation, prominence, factual absorption, fidelity, and user behavior. The foundational paper's widely cited gains are valid within its experimental setting but conditional on a source already being present in a fixed context; they establish neither organic discoverability nor durable traffic effects. Reviewed work indicates that topical relevance and context position are the most reproducible levers, generic heuristics transfer poorly, competition can erode individual gains, and citation-oriented rewrites can impair retrieval. Commercial audits further reveal low source overlap, substantial run-to-run variability, and persistent fidelity gaps. We contribute a multistage formal model, a visibility vector separating discoverability, citation, absorption, and economic outcomes, an evidence hierarchy, and a reproducible protocol based on repeated measurements, paraphrases, controls, human validation, and multi-actor interference. Within this corpus, the evidence is narrow: already-retrieved content can causally alter its citation or use, but no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability or downstream behavior.
arXiv:2607.13836v1 Announce Type: new
Abstract: Using a few concrete examples this paper has demonstrated a few observations on using the partial-structure R1 (pR1) to assemble small-molecule crystal structures. (1) Assembling can start with orienting and/or placing large fragments or small fragments containing heavy atom(s). (2) It is handy to attach a fragment to a partial model as only the orientation needs to be optimized. (3) In a structure containing heavy atoms it is impossible to directly calculate the orientations of a light-atom-only fragment. Instead, it is necessary to use residual reflection intensities in which the contribution of the partial model containing all heavy atoms is deducted away. These observations indicate that pR1 is a useful tool for solving some small-molecule crystal structures.
arXiv:2607.13895v1 Announce Type: cross
Abstract: For an oriented graph $G$, denote by $fvs(G)$ the minimum number of vertices whose deletion from $G$ makes it acyclic. We show that an oriented graph $G$ on $n$ vertices and $m$ arcs satisfies $fvs(G) \le \frac{2n+m+h}{9}$ where $h$ denotes the number of connected components of $G$ that belong to a special class of oriented graphs. This result has three consequences. First, when $G$ is planar, we obtain that $fvs(G) \le \frac{2n+m}{9}$. In particular, this implies that $fvs(G) \le \frac{5n-6}{9}$ for any planar oriented graph $G$, improving the best known upper bound of $\frac{3n}{5}$~[Borodin, Discrete Mathematics, 1979]. Then, applying this inequality to the planar digraphs without directed triangles, we get that $fvs(G) \le \frac{6n-8}{13}$, which improves the current best bound of $\frac{n}{2}$~[Li and Mohar, SIAM Journal on Discrete Mathematics, 2017]. Finally, when $G$ has maximum degree 6, we have $fvs(G) \le \frac{4n}{7}$ and this bound is tight, answering a conjecture of Ai, Gutin, Liu, Yeo and Zhou~[arXiv:2512.01676, 2025].
arXiv:2607.13243v1 Announce Type: cross
Abstract: We consider the problem of parameter inference for ordinary differential equation (ODE) models with structural non-identifiability. Such models arise in a wide range of scientific fields, including control theory, systems biology, and public health. Structural non-identifiability occurs when distinct parameter values provide identical model outputs, resulting in lower-dimensional manifolds of observationally equivalent solutions in the parameter space. This poses challenges for Bayesian inference and Markov chain Monte Carlo (MCMC) methods, often leading to poor mixing and slow convergence. We develop two MCMC methods that use information from structural identifiability analysis. The first, Identifiability-Aware Geometric MCMC, constructs proposals that move within and between non-identifiable manifolds. The second, Identifiability-Aware Pseudo-Marginal MCMC, performs inference on the space of identifiable parameter combinations and reconstructs full parameter values. We show that both methods target the correct posterior distribution and are ergodic under standard conditions. Numerical examples demonstrate improved sampling efficiency and convergence compared with standard MCMC methods.
arXiv:2607.13409v1 Announce Type: new
Abstract: Life-limiting congenital anomalies require accurate prenatal diagnosis for appropriate clinical decision-making. Prenatal ultrasound (US) examinations involve multiple anatomical planes, and diagnosis depends on identifying anatomical planes and selecting diagnostically relevant planes for each anomaly. Existing automated methods either rely on plane-level annotations or aggregate heterogeneous images without explicitly modeling these diagnostic capabilities. We propose AnomExpert, a prototype-driven framework for prenatal US anomaly diagnosis using only case-level supervision. AnomExpert introduces learnable plane prototypes to organize unordered images into latent representations corresponding to anatomical planes without requiring plane annotations. A disease-aware sparse selection mechanism further selects diagnostically relevant planes for each anomaly. Experiments on a multi-center dataset of 3,654 cases show that AnomExpert consistently outperforms nine representative multi-instance learning methods. Using a ViT-small backbone, it achieves 86.9% accuracy and 84.2% F1-score while maintaining parameter efficiency. These findings indicate that modeling anatomical plane identification and disease-specific plane selection improves weakly supervised multi-plane prenatal US anomaly classification. The code is available at https://github.com/TIanCat/AnomExpert.
arXiv:2607.13528v1 Announce Type: new
Abstract: This thesis develops the mathematical theory of intuitionistic dynamic logics - extensions of intuitionistic propositional logic with modalities and fixed point operators. Such systems provide formal tools for reasoning about change, such as encountered in mathematical systems evolving over time or in the knowledge state of an agent after an information update.
We investigate five intuitionistic dynamic logics: intuitionistic master modality, intuitionistic common knowledge logic, intuitionistic linear temporal logic, bi-intuitionistic modal logic and bi-intuitionistic linear temporal logic. On the proof theoretic side we develop sound and complete Hilbert-style axiomatizations as well as non-wellfounded and cyclic sequent calculi. On the semantic side we study these logics over various classes of dynamic models, which are birelational Kripke models satisfying confluence and frame conditions. We establish expressivity results, the finite model property, decidability, as well as complexity bounds.
The main contributions are threefold. First, we develop analytic cyclic sequent calculi for intuitionistic master modality and common knowledge logic, where completeness is obtained by a robust proof search argument. Second, we obtain the finite model property and decidability for bi-intuitionistic modal logic via an intricate combinatorial analysis of dynamic models. Third, we develop a sound and complete axiomatization for intuitionistic linear temporal logic featuring the temporal operators next, eventually and henceforth, thereby providing a positive answer to the long-standing open question concerning the existence of a finite axiomatization.
arXiv:2607.13516v1 Announce Type: new
Abstract: High-quality maps are fundamental for robotics tasks such as navigation and planning. Although modern graph-based LiDAR SLAM systems achieve good trajectory accuracies, a low trajectory error alone does not guarantee geometrically consistent maps, particularly at revisit locations where missed loop closures and residual drift can produce local misalignments. In this work, we address the problem of jointly improving global trajectory estimation and local map quality in 3D LiDAR SLAM. We first propose a framework to efficiently estimate geometry-dependent information matrices for ICP, enabling principled weighting of odometry constraints in a pose graph. We then introduce a hierarchical loop-closure module that decouples place recognition from geometric registration, together with a retroactive loop-closure module that exploits the optimized pose graph to recover missed loop closures. We also propose an evaluation protocol to measure map consistency at revisit locations. We evaluate our SLAM system on several datasets against state-of-the-art LiDAR SLAM systems. Experimental results demonstrate global trajectory accuracies on par with or better than existing methods while consistently improving local geometric map consistency at revisit locations. These results suggest that coupling uncertainty-aware odometry with geometry-guided loop-closure refinement leads to more accurate trajectories and higher-quality maps.
arXiv:2607.13537v1 Announce Type: new
Abstract: Inverter-based resources and IEC 61850 process-bus measurements introduce new protection challenges, including nontraditional fault behavior and measurement-domain cyber-physical attacks. This paper evaluates DL-Xformer, an attention-based Transformer classifier for multi-class fault and cyberattack diagnosis, side-by-side with Dynamic State Estimation-Based Protection (DSE-EBP) on identical high-fidelity electromagnetic-transient (EMT) streaming measurements from an IBR-rich power grid. The evaluation uses an 18-class taxonomy covering normal operation, 11 physical faults, and six measurement-domain attacks, including CT/PT ratio manipulation and GPS spoofing, sampled at 4.8 kHz from synchronized upstream and downstream merging units. DSE-EBP detects all streaming anomalies in 0.417-1.660 ms, with a mean detection time of 0.756 ms, while DL-Xformer classifies the same events in 2.50-50.42 ms, with a mean classification time of 13.46 ms. The longest delay occurs in a deliberate stress case where a CT ratio attack is introduced while residual oscillations from a preceding DLG fault have not fully settled; the event-window accuracy drops to 76.1 %, but the stable final classification remains correct. Measurement-level feature attribution shows that the DL-Xformer decision is driven by physically meaningful current and voltage channels at the attacked measurement location. Together, the two methods motivate a layered protection architecture for next-generation inverter-dominated smart grids.
arXiv:2607.13541v1 Announce Type: new
Abstract: To overcome data scarcity and privacy constraints in data collection, it has become standard practice across academia and industry to augment real training data with text-to-image (T2I)-generated synthetic data, a paradigm we term Real-Synthetic Mix-Training (RSMT). While substituting synthetic data for sensitive real samples is widely regarded as a means to mitigate privacy exposure of the substituted data, the risk to the remaining real samples that actively participate in training has remained largely unexamined.
This work reveals, for the first time, that RSMT can substantially amplify privacy leakage of these real training samples. We establish a theoretical framework, RSMT Memorization Amplification, proving that incorporating synthetic data displaces real samples toward peripheral regions of the mixed feature space, in turn forcing the model to memorize them more aggressively. Guided by this foundation, we propose RSMixLeak to systematically assess this risk through membership inference attacks (MIAs). RSMixLeak comprises two variants depending on the adversary's capability. The non-adversarial variant audits a benign RSMT pipeline with an honest T2I provider, establishing a lower bound on the leakage induced by the intrinsic gap between real and T2I-generated data. The adversarial variant considers an adversary who controls the T2I model or contributes crafted data to the T2I provider, and deliberately enlarges this distributional gap on a target class via either high-level semantic attribute binding or imperceptible pixel-level coating, further amplifying leakage on real training data while improving downstream model utility. Motivated by these findings, we further propose a lightweight leakage propensity indicator computable from real data alone that reliably identifies high-risk datasets unsuitable for entering RSMT, as a self-assessable mitigation.
arXiv:2607.13544v1 Announce Type: new
Abstract: During wind farm operation, Supervisory Control and Data Acquisition (SCADA) systems record numerous anomalies, transients, and specific operational modes, leading to large datasets. However, for a wide range of applications, only measurements corresponding to normal operation are required and, therefore, the SCADA data must be filtered. For this purpose, several methods have been proposed to automate and replace manual filtering conducted by experts via visual inspection of the data. In this paper, we compare the filtering accuracy of multiple clustering algorithms against manual filtering, introducing evaluation metrics that are suitable for unlabeled data and robust across potential applications. Based on the results, we provide recommendations for generalizing model calibration to different datasets and discuss potential use cases for each model. The models are applied to the SCADA data of three turbines of an existing offshore wind farm, using 10-minute statistics across multiple data channels. In addition to the anomalies and operational modes typically recorded, the dataset presents a large number of non-evident outliers due to several field tests. Overall, the results highlight the importance of extending the analysis beyond the power curve, both in feature selection and in the design of evaluation metrics. In most cases, cluster-based methods are able to detect both evident and subtle outliers, achieving higher accuracy than manual filtering. However, the accuracy and the amount of data retained vary considerably depending on the model, and expert involvement remains necessary, though to a reduced extent compared to manual filtering.
arXiv:2607.13517v1 Announce Type: new
Abstract: Vertex splitting replaces a vertex (v) by two nonadjacent vertices whose neighborhoods together equal (N(v)). A split is \emph{exclusive} if these neighborhoods are disjoint and \emph{shallow} if no newly created vertex is split again. For a graph property (\Pi), \textsc{(\Pi)-Vertex Splitting} asks whether at most (k) splits can transform a graph (G) into one satisfying (\Pi).
We continue the systematic study of this operation and settle several open problems. First, we prove that \textsc{Cograph Vertex Splitting} is \textsf{NP}-complete, even on graphs of girth at least 5, resolving a question of Firbas and Sorge (ISAAC 2024). More generally, \textsc{(P_t)-free Vertex Splitting} is \textsf{NP}-complete for every fixed (t\geq 4). We also prove that \textsc{Chordal Vertex Splitting} and \textsc{Unit-Interval Vertex Splitting} are \textsf{NP}-complete, resolving two questions of Abu-Khzam, Chakraborty, Isenmann, and Oijid (IWOCA 2026). Our hardness results extend to the exclusive and shallow variants. Assuming the Exponential Time Hypothesis, none of these problems admits an algorithm running in (2^{o(k)}n^{O(1)}) time; moreover, except for the unit-interval cases, none admits an algorithm running in (2^{o(n)}) time.
arXiv:2607.13411v1 Announce Type: new
Abstract: Clinical AI models can expose patients to harm when adversarial vulnerabilities go undetected, yet formal security auditing requires statistical expertise, specialized tools, and significant time. We present an open evaluation task, built on METR Task Standard v0.3.0, that tests whether frontier AI agents can autonomously implement a structured clinical AI security audit. Given a pre-trained clinical prediction model, a patient dataset, and written instructions, each agent must implement four attacks from pseudocode, compute a Security Posture Score covering FGSM robustness, membership inference resistance, expected calibration error, and boundary attack resistance, and write a structured JSON report in a Docker container using only a bash interface and no scaffolding code. Six variants span the Wisconsin Diagnostic Breast Cancer and MIMIC-IV ICU mortality datasets across three model architectures with increasing defense strength, with reference scores from 55.60 to 90.41. We ran 54 evaluations across three frontier models, with three runs per variant. Claude Sonnet 4.6 and GPT-4.1 completed all 18 runs and received perfect evaluator scores. GPT-4o completed 61 percent of runs and used about five times the per-run token count of Claude, although provider tokenization differs. Total API costs were 8 US dollars for GPT-4.1, 12 US dollars for Claude Sonnet 4.6, and 27 US dollars for GPT-4o. GPT-4o failures involved premature session termination, an aggregation error, and an empty submission file. The task, scoring infrastructure, and Wisconsin Breast Cancer assets are publicly released; MIMIC-IV variants require separate PhysioNet access.