Papers

Diagnostic meta-analyses commonly assess small-study effects using the Deeks test, which tests for a study-size trend in the log diagnostic odds ratio. Under the hierarchical summary receiver operating characteristic (HSROC) model, this quantity reflects both latent accuracy and threshold unless the curve is symmetric. A threshold trend can therefore generate a log diagnostic odds ratio trend without a latent accuracy trend, or conceal an existing one. We develop a likelihood-ratio test of the latent accuracy trend under a common-shape binomial HSROC model. Its null hypothesis states that the summary curve is unchanged across study sizes. A log diagnostic odds ratio test using the same binomial fit provides a comparison with a different null hypothesis. In simulations, the proposed test remained near its nominal level under its null and had higher power than Deeks with a latent accuracy trend alone. In a colorectal cancer review of faecal immunochemical tests, it detected a negative latent accuracy trend obscured on the log diagnostic odds ratio scale by an opposing threshold contribution. All three tests detected size trends for impedance plethysmography in deep-vein thrombosis, where the fitted curve was nearly symmetric. The proposed test distinguishes changes in the summary curve from changes in operating position.

Large language models (LLMs) can influence people's beliefs, yet little is known about whether and how they can manipulate each other. To investigate this, we simulate conversations between two agents: a target LLM that role-plays a human persona based on demographic and psychological attributes, and an influencer LLM that aims to make the target's beliefs more extreme. We examine radicalization along two pathways: resonance, where the influencer reinforces a target's pre-existing belief, and persuasion, where the influencer promotes a belief the target initially considers unimportant. Across affective and behavioral metrics, we find that both mechanisms radicalize the target. However, resonance produces consistently stronger effects than persuasion. Different influence tactics, such as using sycophancy and unverified claims, produce different levels of radicalization, but not consistently across metrics. We further show that resonance propagates to related beliefs, suggesting interconnected belief structures within AI agents. These findings indicate that AI agents are susceptible to radicalization, particularly when messages align with their existing beliefs, raising concerns about the vulnerability of personalized AI agents and multi-agent AI ecosystems.

We prove a dimension-free Poincaré inequality for isotropic unconditional log-concave probability measures, establishing the Kannan-Lovász-Simonovits (KLS) conjecture in the unconditional setting. The proof combines a new application of Dunkl operator theory combined with an appropriate transformation of the log-concave measure. Specifically, the proof follows by a direct spectral analysis of a new Dunkl-Langevin operator adapted to the transformed measure. The core ideas of the proof were generated by AI generative tools through interactions with the authors over several weeks. The authors refined and formalized the argument.

We study the generation of agentic workflows that jointly optimize multiple objectives, such as accuracy, cost, latency, robustness, and consistency. Existing methods for workflow generation typically optimize accuracy alone or a weighted sum of objectives, so each trained generator commits to one fixed trade-off and must be retrained from scratch when preferences change. To alleviate this, we propose MoFlow, which generates workflows optimized across varied preferences. Specifically, MoFlow formulates workflow generation as a multi-objective Markov decision process and solves it by leveraging Convex-Hull Monte Carlo Tree Search with optimistic set-valued backups, where every node stores a set of reachable trade-offs rather than one weighted score. A single search thus approximately covers the Pareto front, from which MoFlow can return a workflow for any preference by lookup without retraining. We evaluate MoFlow against six strong baselines on six benchmarks spanning mathematics, code, and question answering. Since the baselines are single-scalar optimizers by design, an apples-to-apples comparison is difficult. We instead adopt an evaluation setup that favors the baselines, in that they are rerun for each testing preference, which MoFlow never sees. Even under this stringent setup, MoFlow achieves the highest average hypervolume.

In this article I revisit the frequency interpretation of probability of von Mises, and argue that it contains the key elements of any coherent interpretation of scientific probability - notwithstanding the heavy criticisms von Mises collected from mathematicians and philosophers alike. I propose a modified definition that is not based on von Mises' "infinite collectives" but retains an essential ingredient of his interpretation, namely that probability can only quantitatively be defined for events that are, or can be, repeated in similar conditions and that exhibit frequency stabilisation. New is that the mentioned "conditions" should be "partitioned". Thus, I will partition probabilistic systems into object and environment, or in object, initiating, and probing subsystem, and show that such partitioning solves a series of classic problems. Other consistent interpretations can be unified with frequentism, e.g. Bayesianism à la Jaynes. I explore in detail the tantalising subjective or operationalist touch of probability, and argue that it can be put on objective grounds. One further upshot of the model is that several quantum puzzles can be seen as deriving from an imprecise interpretation of probability.

Claims concerning a dynamical equivalence between general relativity and two reformulations of it, in which gravity is encoded in torsion and non-metricity respectively, have attracted some attention. If taken to be true, this 'Geometric Trinity' suggests that it is impossible to determine, even in principle, whether one inhabits a spacetime solution possessing curvature, torsion or non-metricity (insofar as the probes of geometry are paths of test particles). This claim is in direct tension with the long tradition of studying gravity as a gauge theory, in which distinct geometric attributes couple to distinct properties of matter (such as energy-momentum to curvature and spin angular momentum to torsion). In this short paper, I compare these two conflicting frameworks for geometrizing spacetime and describe the precise source, nature and implications of their disagreement. The equivalence claim holds only under a substantive restriction on the matter sector, namely that matter does not couple to an independent affine connection. Because the relevant couplings remain beyond current empirical reach, the choice between the frameworks turns on a trade-off among ontological, epistemic, and methodological forms of parsimony.

Large language model (LLM) agents are vulnerable to safety risks such as injected malicious instructions or misleading information, motivating runtime defenses that prevent unsafe action in execution across diverse risks while preserving benign-task utility. Existing system-level defenses either focus on risk detection rather than timely prevention or rely on predefined rules with limited flexibility across diverse risks. We propose a risk-aware harness that integrates LLM-based monitoring for flexible risk detection and structures monitor-guided execution around three core modules: trigger, monitor, and feedback, enabling targeted safety interventions while limiting disruption to benign task execution. To adapt the harness to different risks and deployment settings, we introduce HARDE, a two-stage harness optimization framework that first performs isolated probing of each module to derive an optimization guide, then uses this guide to iteratively optimize the harness based on safety and utility feedback. Experiments across three attack benchmarks show that HARDE improves runtime safety while preserving utility, outperforming manually designed harnesses and naive optimization baselines. Our analysis shows that effective runtime defense benefits from complementary safety mechanisms, attack-aware harness optimization, and harness designs matched to monitor capabilities. Our code is available at https://github.com/Liuz233/HARDE.

Recent advances in the classification of certain classes of Yang-Baxter operators through a holistic categorical approach has some of us wondering about ways to identify solutions. In particular the classification of charge conserving solutions in all dimensions inspires the question: \emph{what fraction of solutions are close to charge conserving solutions, under any reasonable notion of equivalence?} Of course reasonable and close to are open to interpretation. Here we suggest some precise notions of both, and prove that in dimension $2$, 10 out of 11 local families of solutions are close to charge conserving, and the one that misses is a singleton.

In personalized AI applications, such as conversational assistants and recommender systems, users interact not with models in isolation but with broader systems that access, infer, and reuse user information across components and over time. While such use of user information is integral to personalization, it also raises important privacy questions. In this paper, we argue that individual model- or component-level analyses may not capture all privacy risks arising in such systems, motivating a system-level perspective on privacy. We distinguish and analyze four interconnected privacy-risk channels in personalized AI, and subsequently propose four requirements for system-level privacy evaluation, covering interaction trajectories, internal information flows, indirect leakage, and the privacy-utility trade-off. We argue for their systematic incorporation into privacy audits of personalized AI.

We present AREX-2, an effort to advance the self-improving capability of LLM agents, which we define as the ability to iteratively refine a solution at test time. This ability rests on two complementary capabilities: reflection, which produces a solution better than the current one, and long-horizon execution, which keeps the iteration effective over many rounds. We hypothesize that both capabilities are domain-agnostic, and can therefore be learned in scenarios that are well suited for supervision. Accordingly, we synthesize long-horizon improvement trajectories from machine learning and algorithmic programming tasks, two domains that offer verifiable feedback and reward sustained iteration. Trained on this data, our agent, built on Qwen3.8-27B, achieves strong results on MLE-bench Lite (81.8) and Frontier-CS (70.7), transfers to deep research with 84.0 on BrowseComp, 52.6 on HLE, 92.2 on GAIA, and 93.8 on DeepSearchQA, and keeps improving as its budget of rounds grows. These results show that long-horizon reflective data is an effective route toward self-improving agents.

Simulation-augmented generation (SAGE) is a recent technical proposal in which models simulate individuals' viewpoints at inference time in order to provide more representative answers to contentious user queries. A core challenge for SAGE is making inference-time simulation efficient without sacrificing representation quality. We introduce the first formalization of this problem, based upon an axiom from proportional clustering known as metric proportional justified representation+ (mPJR+) which is the strongest proportionality axiom known to always be satisfiable by centroid-based clustering. We prove that to proportionally represent the viewpoints of a population of $n_H$ humans on a given prompt, we need only create simulations of $n \ll n_H$ individuals, and at inference time, need only dynamically route to $k \ll n$ of those simulations based upon the prompt. This twofold reduction still yields approximate proportional representation guarantees for the entire population. Empirically, across two domains-political questions and personal advice-our proposed routing algorithm achieves higher mPJR+ satisfaction rates than $k$-means-based or random selection baselines.

For a graph $H$, let $f(n,e,H)$ be the least number of colors in an edge-coloring of some $n$-vertex graph with at least $e$ edges in which every copy of $H$ is rainbow. Burr, Erdős, Graham, and Sós conjectured that $f(n,\lfloor n^2/4\rfloor+1,C_{2k+1})=(1/8+o(1))n^2$ for every fixed $k\ge3$, and Bucić, Chen, and Ma recently proved this for all $k\ge4$. We prove the remaining case $k=3$: \[ f\left(n,\left\lfloor n^2/4\right\rfloor+1,C_7\right) =\left(\frac18+o(1)\right)n^2. \] The lower bound rests on a weighted palette inequality, which we prove with an exact rational certificate on five sampled vertices. Its main ingredients are a fractional matching of compatible triangular edges and private resources attached to nontriangular edges. A stable form of the inequality, combined with regularity, triangle removal, and a direct argument for graphs close to bipartite, transfers the bound to arbitrary edge-colorings. We also describe a Lean 4 formalization of the conjecture for every fixed $k\ge3$, which combines the new seven-cycle proof with a formalization of the Bucić-Chen-Ma argument for $k\ge4$.

Vision-language models (VLMs) can contradict themselves across views of the same spatial relation and fail to respond when that relation changes. Addressing these failures requires supervision that captures error magnitude and geometric dependencies across observations, both of which remain implicit in training on individual answers or ordinal preferences. Therefore, we introduce GaugeVLM, which makes this structure explicit through controlled object and camera interventions in explicit 3D scenes, producing linked observations with measured differences between spatial relations and shared truths across views. To translate this structure into learning signals, its core objective, GaugeDPO, converts measured errors into preference margins, directly supervises correct canonical rankings across views, and links intervention-induced answer-odds contrasts to measured relation changes with view-specific scales. Our analysis bounds canonical prediction error and establishes that the cross-view and intervention constraints can be jointly satisfied. Empirically, GaugeVLM improves all 10 established spatial metrics over supervised fine-tuning across three VLM backbones, with the main 7B model gaining 15.0 and 18.9 percentage points on MSMU distance and QSpatial+, respectively. These gains also extend to autonomous driving and embodied reasoning, demonstrating the robust generalization across domains.

We refute Aigner's majorization conjecture for star forests in adaptive quantitative group testing with two defective items. Each test counts the defectives in a chosen subset; the possible defective pairs form the edges of a star forest. Aigner's necessary majorization condition is sufficient through five tests but is not sufficient at any budget from six onward. Nevertheless, for every fixed number of stars, it becomes sufficient once the test budget is large enough. We prove that majorization is sufficient for all forests with at most twenty-nine stars, and that this bound is sharp. We also construct counterexamples that pass arbitrarily many successive compatibility checks based on majorization. The five-test converse, counterexamples, eventual sufficiency theorem and unbounded-delay construction have analytic proofs. The sharp finite support bounds also use exact integer-inequality certificates.

Recent progress in network renormalization has identified various ways to transform network representations consistently across resolution levels. In particular, the renormalization flow describing how random graph models transform under node aggregation has identified a fixed point corresponding to an aggregation-invariant model that preserves the form of the connection probability across arbitrary scales. While this probabilistic multiscale approach has successfully been used to model real-world networks, construct latent-space node embeddings, and reconstruct finer-grained versions of aggregate network data in a principled resolution-invariant way, it has so far been restricted to binary graphs. Here we reformulate multiscale network renormalization for weighted random graphs. Via a generating-function approach, we identify a weighted multiscale model whose full probability law, encoding both link creation and weight assignment, has an invariant compound-Poisson form under arbitrary node aggregation. The renormalization rules that transform parameters across aggregation levels allow the ensemble to be calibrated at one level and applied at another, enabling consistent modeling and reconstruction of weighted networks across arbitrary scales.

Vision-language models may rewrite anomalous text in images into linguistically plausible expressions, compromising OCR transcription faithfulness. Sequence-level task rewards and local teacher guidance are complementary, but guidance from the same teacher may not remain equally effective as the student improves. Offline analysis shows that supervision from a fixed teacher becomes progressively less favorable as the student improves, both across training checkpoints and across response groups with different task rewards. Motivated by this observation, we introduce GAD-RL, which adaptively regulates teacher supervision during joint post-training according to the student's current task performance and local distributions. A frozen teacher conditions on reference transcriptions and student-generated prefixes. GAD-RL disables distillation for response groups containing an output with task reward at least 0.95 and continuously attenuates distillation strength as group-mean reward increases. It also weights forward KL by the student's probability of the teacher's Top-1 token, moderating local auxiliary updates when student support for that candidate is low. On Qwen3.5-2B, GAD-RL achieves 59.92% Micro Recall on CHAOS-Bench, surpassing GRPO and GRPO+OPD (fixed-weight) by 8.45 and 4.43 percentage points, respectively, while achieving an Overall score of 91.18 on OmniDocBench v1.6.

Extended reality (XR) systems are increasingly used in healthcare applications ranging from surgical planning to remote rehabilitation and mental health support. However, the rich streams of sensor, biometric, behavioral, and environmental data that enable these applications also introduce substantial privacy and security risks. Adversaries may exploit insecure communication, sensor side channels, application-layer vulnerabilities, or data-processing pipelines to infer sensitive information or disrupt clinical workflows. Despite growing interest in XR security and privacy, the healthcare-specific literature remains fragmented. In this Systematization of Knowledge (SoK), we review 65 peer-reviewed studies published between 2017 and 2024 across XR, security, privacy, and healthcare venues. We develop a unified threat taxonomy spanning device, user, network, and cloud layers and introduce XR-PRISM, a quantitative Privacy and Risk Impact Scoring Metric for systematically characterizing security and privacy risks. Our analysis identifies several gaps in the literature: more than 70% of proposed countermeasures lack standardized risk evaluation, fewer than 15% of studied attacks require high attack prerequisites, and reproducibility is limited by the scarcity of publicly released artifacts and datasets. Based on these findings, we outline a research roadmap emphasizing shared benchmark datasets, stronger artifact-release practices, improved cloud-layer protections, and more comprehensive detection, mitigation, and recovery mechanisms. This SoK provides a structured and data-driven foundation for understanding existing risks and guiding the development of more secure, privacy-preserving, and usable XR healthcare systems.

A comparison between adaptive and fixed-time traffic signals can change when only their timing settings change. We examine this sensitivity in 240 simulation runs on one OpenStreetMap-derived urban network, using eight configurations, three arrival rates and ten paired demand seeds. The experiment varies the maximum green of presence-based actuation, retimes the fixed baseline, and includes a discharge-based phase-termination rule. The endpoint is finite-window time spent per offered trip, including entry waiting and penalties for removed or abandoned trips. At the highest tested load, actuation with a maximum green of twice the planned split has a mean endpoint 57.4 per cent above the original fixed plan. Reducing that multiplier to 1.25 puts it 70.7 per cent below the same plan. Against the best of three tested fixed plans, discharge-based termination reduces the mean endpoint by 14.8, 15.8 and 18.3 per cent across the three loads. Both timing sweeps favour their shortest tested setting, so neither identifies an optimum. This exploratory study uses synthetic demand, negligible start-up lost time and physical removal of persistently stationary vehicles. The results support testing timing sensitivity before attributing performance to a controller family; they do not establish a setting for real-world deployment.

Public understanding of how people use LLM-based conversational AI assistants comes primarily from aggregate platform reports by OpenAI and Anthropic, which apply fixed taxonomies and inferred demographics to hundreds of millions of users and release only summary statistics that outside researchers cannot re-analyze. We provide a complementary, conversation-level view: complete ChatGPT exports comprising 202,590 conversations from 1,252 users across India, Nigeria, Brazil, and Pakistan, paired with self-reported age and gender and spanning December 2022 to February 2026. To our knowledge this is the first conversation-level, demographically grounded comparison of ChatGPT use across multiple non-Western countries. We ask what these users use ChatGPT for (purpose), what they talk about (topics), and how they engage with it (mode of interaction), using the platform's own classifiers, unsupervised topic discovery, and a thematic analysis of expressive conversations. Personal use accounts for 55-64% of conversations in every country and coursework is about as common as work, so workplace productivity describes a minority of use. Unsupervised topic discovery surfaces country-specific uses that the OpenAI taxonomy folds into generic categories: health and wellness in India and Brazil, Urdu-English translation in Pakistan, current affairs in Nigeria, religious questions in Nigeria and Pakistan, and self-reflection in Brazil. Over three years, the share of conversations that seek information declined only modestly and the share that delegate a task did not grow, while conversations in which users express themselves rose from a few percent to roughly a fifth or more in every country. The same product is thus attached to different local needs in each country, and understanding what adoption means requires conversation-level, country-sensitive measurement alongside global aggregates.

Self-supervised learning (SSL) removes the need for annotations and makes models that are capable across more domains than supervised learning. The autoencoder SSL framework learns by reconstructing its own input after information loss through a bottleneck or noise injection. Masked autoencoders (MAE) are the most successful instantiation of this framework: they encode a random subset of patches, then decode the masked-out patches. In this work, we introduce key modifications to improve MAEs. Our method augments an image in two different ways, then masks and encodes each view separately. It then exchanges the global representations (CLS tokens) between views before decoding the masked patches. By design, our Masked Swingers encourages learning a view-agnostic summary of the image to facilitate efficient transfer. We perform extensive experiments, and find Masked Swingers outperforms MAE by +3-5% on ImageNet-1K kNN and provides large gains on fine-grained tasks, e.g., relative gains of +45% on instance retrieval, +22% on animal re-ID, and +76% on Omniglot character recognition. To boot, Swingers reduces error -64% relative to MAE on three new state-probing datasets, opening the door to world modeling. Welcome to our Swingers party.

This review summarizes the changing observational and theoretical landscape that has led to rapid developments in our understanding of the orbital history of the Milky Way's most massive satellite galaxies, the LMC and SMC. The determination of high precision center-of-mass proper motions for the Clouds and their environs, coupled with new insights into the structure of dark matter halos from cosmological simulations, indicates that the Clouds are recent additions to the Milky Way's satellite system, where the LMC was sufficiently massive at infall to induce distortions in the Milky Way's dark matter halo, and bring in satellites of its own. To gain insights into the orbital history of the Clouds it is, therefore, necessary to integrate orbits in a time evolving Milky Way and LMC potential. The proper motion measurements further indicate that the LMC and SMC have had a recent strong encounter with each other. This leaves open the impact of this encounter on their internal structures, and the degree of correspondence between their stellar, HI and dark matter centers. The choice of kinematic center in turn affects the center of mass proper motions; this is the biggest observational unknown, even in the high precision Gaia era. Given the expected mass loss of the SMC, the distortions it induces in the LMC halo, and the recent LMC-SMC close encounter, understanding the orbits of objects about the LMC (including the SMC) and interpreting observational data of the Clouds requires full N-body simulations that are constrained by their present-day structure and kinematics.

Alcohol intoxication is a leading contributor to fatal and severe-injured e-scooterist crashes. Current countermeasures, such as temporal restrictions or pre-ride cognitive screening, cannot continuously assess an e-scooterist's physical motor control or impairment in real time. We conducted a controlled experiment in which 25 participants rode an instrumented e-scooter through a test track while sober and at two targeted blood alcohol concentration levels (0.05% and 0.08%). The e-scooter was instrumented with a six-axis inertial measurement unit (IMU), and throttle and brake lever position sensors, all sampled at 100 Hz. Two complementary signal features were computed: normalised permutation entropy, which quantifies temporal complexity, and standard deviation, which quantifies signal amplitude. Repeated measures correlation identified seven kinematic features (all IMU and throttle signals) whose entropy decreased (p < 0.001) while standard deviation increased (p < 0.01) with increasing intoxication, indicating that intoxicated riders shift from continuous, low-amplitude micro-corrections to fewer, high-amplitude reactive corrections. An entropy based multi-class logistic regression classifier, evaluated through leave-one-participant-out cross-validation, achieved 85% overall accuracy and a weighted one-vs-rest area under the receiver operating characteristic curve (AuROC) of 0.94, with a sober-vs-high AuROC of 1.00. Steering rate and lateral acceleration were the most important predictive features, indicating that alcohol induces a distinct collapse in lateral equilibrium during riding. Ultimately, these results demonstrate that onboard kinematic sensing combined with entropy-based signal analysis can reliably distinguish sober from intoxicated e-scooter riding, providing a foundation for automatic intoxication detection systems that preserve mobility for sober riders.

Persistent memory helps LLM agents carry information across long interactions, but correct memory can still be used incorrectly when queries change or memory states evolve. Existing work mainly studies memory content errors or evaluates fixed test cases, leaving memory-use failures hard to discover systematically. We formulate this issue as a fuzzing problem and categorize such failures into query-related and memory-state failures. We then develop U-Fuzz, which starts from memory checkpoints as test seeds, mutates queries or memory states under explicit mutation obligations, validates each mutant, and uses observed memory behavior to guide iterative testing while keeping failure labels outside the search. We evaluate U-Fuzz across several memory systems against diverse fuzzing baselines, and further test an output-only setting with API-based LLMs where memory retrieval is hidden. Across these settings, U-Fuzz consistently uncovers more confirmed memory-use failures, showing that its search remains effective across different memory architectures and even when only final responses are observable.

The performance ceiling of an LLM team is constrained not only by individual model capabilities, but also by inter-member error resonance and predictive differences. Although heterogeneous teaming is often observed to be effective in practice, existing approaches lack complementarity metrics that are computable, interpretable, and optimizable, leaving team composition to rely on heuristics. We propose a heterogeneity-driven team selection framework that performs offline profiling to characterize individual capability along with two complementary signals: one captures decorrelation in error patterns to reduce co-failures, while the other measures divergence in predictive behavior to capture strategy diversity. We formulate team selection as a standardized quality–complementarity combinatorial objective and apply an efficient greedy search to select a small team from a candidate pool. Experiments across multiple benchmarks demonstrate that our framework consistently outperforms quality-only baselines under controlled candidate pools and team sizes, establishing reusable selection principles for multi-LLM systems.

In this paper, an algorithm for constructing orbifolds of Calabi-Yau manifolds of the Berglund-Hubsch-Krawitz (BHK) type and their mirror pairs for loop-type polynomials is proposed and applied. A complete enumeration of loop polynomials satisfying the Calabi-Yau condition was carried out, yielding 216 unique polynomials. For each polynomial, the Euler characteristics, the numbers of generations and singlets were computed, and the twisted sector states were analyzed using Roan's generalized combinatorial construction. It was established that Roan pairs are absent for all studied configurations, indicating a zero contribution of the twisted sectors to the difference in Hodge numbers upon resolving the orbifold singularities. The primary physical result of this work is the discovery of two configurations with diagonal exponents $\{74, 3, 3, 5, 3\}$ and $\{74, 3, 5, 3, 3\}$, for which the number of quark and lepton generations is precisely three ($N_{\text{gen}} = 3$), which is of key interest for constructing phenomenologically realistic models of particle physics.

The scaled relative graph (SRG) represents the joint gain and phase properties of an input–output operator in the complex plane. This paper characterizes the SRG closure of causal, stable, square discrete-time linear time-invariant (LTI) systems on one-sided square-summable sequences. The Beltrami–Klein image of this closure equals the convex hull of the numerical ranges of the transformed frequency-response matrices. For real-coefficient single-input single-output systems, this yields the hyperbolic convex hull of the discrete-time Nyquist locus. The proof addresses the Hardy-space restriction imposed by one-sided inputs, which excludes persistent sinusoids, by replacing them with explicitly constructed finite-duration sinusoids. This construction shows how points in frequency-wise SRGs arise as limits of operator SRG points generated by admissible inputs. The characterization enables model-based computation of the closure from a state-space realization using frequency-response evaluations, numerical ranges, and planar convex hulls, without solving linear matrix inequalities.

Morphological characteristics such as spiculation and lobulation play an important role in assessing pulmonary nodules on computed tomography (CT), particularly in relation to malignancy risk. This study examines whether learning radiologist-annotated morphological features together with malignancy risk from lesion-centred 3D CT volumes improves classification performance. The Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset was used, comprising 3,918 reader-level nodule annotations from 742 patients after excluding indeterminate malignancy ratings. Patient-level splitting was used for training, validation, and testing, with 112 patients and 628 reader annotations in the held-out test set. A single-task 3D convolutional neural network was compared with a multi-task model predicting malignancy risk, spiculation, and lobulation. The single-task model achieved a balanced accuracy of 0.548 and receiver operating characteristic area under the curve (ROC-AUC) of 0.552, while the multi-task model achieved 0.539 and 0.558, respectively. Patient-level bootstrap analysis showed an ROC-AUC difference of 0.005 (95% confidence interval (CI): -0.087 to 0.090) and a balanced-accuracy difference of -0.009 (95% CI: -0.067 to 0.043). The auxiliary tasks were strongly imbalanced and showed limited predictive performance. Overall, including morphological features did not clearly improve malignancy-risk classification, showing the importance of class balance, label formulation, and reader-level annotation structure in multi-task pulmonary CT analysis.

Advancing beyond traditional static scoring models, LLM-powered agentic recommender systems (LLM-ARS) instantiate users and items as autonomous agents, whose semantic states are dynamically refined through a recurrent process known as collaborative reflection. While this mechanism improves recommendation quality, it simultaneously introduces a systemic vulnerability: adversarial evidence injected into a single agent can be rationalised into a legitimate preference narrative, written back into memory, and propagated to other agents through interaction contexts. We term the local rationalisation process reflection laundering, and its system-wide escalation through collaborative reflection collaborative-reflection hijacking. Existing attacks on recommender systems, whether based on interaction-level data poisoning or text-level adversarial perturbations, assume static pipelines and thus cannot exploit this recurrent, multi-agent amplification pathway. To bridge this gap, we first conduct a controlled vulnerability analysis that establishes two exploitable properties underlying collaborative-reflection hijacking: reflective persistence and cross-agent propagation. Then building on these findings, we propose VirusCascade, the first black-box targeted promotion attack that jointly shapes semantic and structural attack surfaces: the former ensures the target item is naturally rationalised as satisfying broad user preferences, the latter positions it for system-wide propagation. Extensive experiments on four real-world datasets across diverse LLM-ARS architectures demonstrate that VirusCascade consistently achieves state-of-the-art targeted exposure under evaluated stealth constraints, reaching a mean E@20 of 0.384 and surpassing the strongest baseline by an absolute margin of +0.185.

Coding agents are increasingly asked to build software rather than patch it, yet benchmarks for from-scratch repository construction are mostly limited to a single language and depend on manually curated tasks. We introduce Zero2Repo, a benchmark in which an agent receives a product requirements document, an interface contract, and an empty workspace, and must deliver a complete repository in the project's native ecosystem. Tasks are produced by a language-agnostic authoring pipeline that converts real, version-pinned open-source projects into behavioral specifications, reproducible environments, and hidden acceptance tests. Each task is validated by execution: a reference implementation derived from the upstream project must pass, and adversarial validation must show that the tests reject incorrect implementations. Evaluation runs production coding agents in isolated containers, withholds the acceptance tests until an explicit submission, and assigns a binary reward only when every test passes, with no LLM judge. The pipeline and harness make no language-specific assumptions and apply to mainstream programming ecosystems; the current release contains Python, TypeScript, Go, and C++ tasks. Even on 11 tasks drawn from repositories that frontier models have very likely seen during training, the strongest agent solves only 10, and every failing submission passes 90-99% of the hidden tests; for the two strongest agents, 67-100% of failed tests trace to a single omission or a low-frequency rule stated in the specification rather than to a missing subsystem, so each failure is a concrete target for improvement.

In the nearly unstable heavy-tailed regime of Jaisson and Rosenbaum, where the intensity of a linear Hawkes process converges to a rough square-root Volterra process, we ask what a record of length $T$ reveals about the relative distance to criticality $1-\rho$ and the exogenous rate $\mu$. For the first, the information is governed by $\mathcal I_T=T\mu(1-\rho)$, not by the number of events; $\mathcal I_T$ stays bounded although the event count diverges like $T^{2\gamma}$, so no estimator of the relative distance to criticality is uniformly locally consistent. For the second, the Fisher information diverges because the rough limit has an atom at zero: the asymptotic information geometry is singular in the exogenous direction and, unlike the light-tailed case, has no Feller-type threshold. We prove that the joint maximum likelihood estimator of $(\mu,\rho)$, whose likelihood is jointly concave, recovers $\mu$ consistently at the random rate $\{Q_T/\log\log Q_T\}^{1/2}$ given by the observed information $Q_T$, while its estimate of the relative margin is only tight: the exogenous rate is learned although the endogenous parameter it is coupled to is not. For empty-start records $Q_T$ diverges because the rough limit started at zero spends positive time at zero; on windows observed long after the start, still with the branching ratio unknown, learnability holds with probability tending to one as the window grows, through the ergodicity of the rough limit and the atom of its stationary law. We also prove finite-dimensional convergence of the stationary intensity to the stationary rough Volterra process.