Computer Science (arXiv)

Controlling an autonomous vehicle at the limits of handling is a challenging task. Due to external influences, such as road conditions or weather, a vehicle can easily become unstable. Since most control algorithms assume stable vehicle behavior, they might fail in these situations. Especially when operating expensive vehicles without a safety driver on board, as in autonomous racing, this poses a significant challenge. To enable safe operation at the vehicle's dynamic limits, we present a comprehensive stability control system that safeguards motion control algorithms in autonomous driving. The proposed system consists of an electronic stability control (ESC), a slip control (SC), and a countersteer system (CS), which collectively adapt steering and brake commands from the motion controller to maintain vehicle stability. We validate our approach through both simulation and experiments on a real-world, full-scale vehicle. The results show that the stability control system maintains vehicle stability in critical situations and extends the operational feasible region. To simplify integration, we provide an open-source implementation at github.com/TUMFTM/tam-stability-control.

We demonstrate that RL finetuning an LLM using debate, a two-player adversarial game between a generator and a critic adjudicated by a weaker LLM judge, reduces reward hacking compared to a reinforcement learning from AI feedback (RLAIF) baseline. Reward hacking is a central obstacle in RLAIF: as training progresses, the policy learns to exploit systematic errors in its AI judge, degrading task performance, a problem that worsens precisely when the judge is weaker than the policy, the setting most relevant to overseeing increasingly capable AI systems. We study mathematics tasks, where final-answer correctness is verifiable, allowing us to measure reward hacking dynamics. We train a Gemini~2.5 Flash-class policy with a frozen, weaker Gemini~2.5 Flash Lite judge, comparing a single-player RLAIF baseline against debate. While the baseline quickly hacks the judge, debate maintains judge performance throughout training, leading to a higher peak validation accuracy (45\% performance gap recovered) that persists through many RL steps. Additional experiments show that: 1) further weakening the judge leads to faster hacking, but this can be compensated by adding an additional debate round; 2) debate incentives override prompted misalignment; 3) RL using an LLM judge has a smaller train/validation reward gap than RL from verifiable rewards; 4) learning to critique to convince the judge using ground truth labels is possible but slow. Taken together, our results are a positive update on the feasibility of debate, while highlighting that balancing multi-agent training is critical: without player constraints, adversarial training risks defaulting to critic judge-hacking. We show that critique word limits (effective up to 150 words) successfully balance the game and avoid judge hacking, though this introduces a trade-off by restricting critic expressive clarity.

Anomaly detectors are hardest to deploy exactly where training data is scarcest: a newly commissioned production line has a handful of verified "golden" samples and no machine-learning engineer on the factory floor. We present a training-free human-in-the-loop framework in which a domain expert corrects a PatchCore detector by direct memory bank editing: no retraining, no gradients, no original training data. A false-positive correction inserts the reviewed image's normal patches through a self-calibrating novelty gate admitting only those beyond the median pool-normal nearest-neighbour distance. From a bank built on only ten golden samples, operator corrections close a median 66% of the gap to an uncorrected fully trained bank (mean 80%, raised by three categories that overshoot parity), significantly improving 12 of 15 MVTec AD categories and harming none: ten samples plus corrections outperform hundreds of samples without them. On already-trained banks the headroom is smaller and concentrated where the bank undersamples normal appearance (gated: toothbrush +0.10, metal nut +0.09, zipper +0.05, screw +0.05), and no category except grid is significantly harmed. Evaluation uses a held-out protocol (20 splits per category, Holm-corrected Wilcoxon), because corrected images entering the bank inflate naive evaluation toward AUROC 1.0 by memorisation. Passive and active querying are statistically indistinguishable; a matched-label-budget control attributes gains to deployment-time label production at 43% of exhaustive-review cost; a defect-memory extension fails decisively. Feedback is simulated from ground truth; live expert trials, where mislabelling is costliest on small banks, remain future work.

Fuzzy private set intersection (PSI) enables two parties to identify approximately matching elements between their input sets, where two elements are considered a match if their distance is at most a threshold $δ$ under a given metric. Although substantial progress has been made, existing constructions for general Minkowski distances either rely on strong two-sided geometric separation assumptions or incur substantial overhead under one-sided assumptions. In this work, we present the first concretely efficient fuzzy PSI protocols for general $L_{p\in[1,\infty]}$ distances under one-sided assumptions, relying solely on lightweight symmetric-key primitives. Our constructions support both sender-sided and receiver-sided settings. We further study sparser input distributions and present more efficient protocols tailored to this case. To reduce the overhead scaling with $δ$, we non-trivially incorporate prefix trie techniques into our protocols, achieving $O(\logδ)$ complexity for general $L_{p\in[1,\infty]}$ distances for the first time, improving upon $O((\logδ)^d)$ or $O(δ)$ complexities of prior works. Extensive experiments, across a wide range of parameter settings, show that our protocols significantly outperform prior works under the same assumptions. Specifically, against van Baarsen and Pu (EUROCRYPT'24), our protocols achieve up to $239\times$ faster computation and up to $20\times$ lower communication. Against Dang et al. (CCS'25), we achieve up to $518\times$ speedup and up to $63\times$ communication reduction. Against Bui et al. (ASIACRYPT'25), we achieve up to $4818\times$ faster computation and up to $282\times$ lower communication.

We study a dynamic coalition-formation process in the tradition of Konishi and Ray (2003): players repeatedly form and dissolve binding agreements, evaluate states by discounted long-term expected payoffs, and hold self-confirming beliefs about the process. States and payoff sharing follow Heitzig and Kornek (2018): a state is a hierarchy of nested agreements; agreements are formed by merging existing top-level coalitions, and are terminated together with all agreements containing them; and the members of a new agreement share the surplus it generates, measured against the state without that agreement. All payoff assumptions are structural. We prove that every grand state ever reached is absorbing, and that every absorbing state is grand, for every discount factor. A grand state is actually reached, almost surely, in three cases: small discount factors; three players; and, for any number of players and all discount factors, whenever every player prefers every grand state to every non-grand state in static payoffs, as when distributional stakes are smaller than each player's share of the efficiency gain. Otherwise the process can fail only by cycling for ever among non-grand states. We give exact necessary conditions on such a cycle, and show that for a fixed candidate cycle they reduce to a finite system of linear inequalities in the static payoffs, so the question is decidable. Solving it yields a counterexample: with four players and discount factor one half, under either termination rule, there is an equilibrium that cycles for ever, so the grand coalition need not form. The example survives a far-sighted variant of the sharing rule under which merging raises every player's discounted long-term payoff, not only the static one; there the merge is blocked purely by a better move available to a subgroup. Whether arrival can fail as the discount factor tends to one remains open.

Intelligent channel state information (CSI) feedback is essential for realizing the high capacity and spectral efficiency goals of future 6G systems, yet existing deep learning solutions face a trade-off between model generalization and scenario-specific performance. Large neural networks generalize well but incur high computational and tuning costs, while small models excel in particular environments but require repetitive costly end-to-end training for each base station (BS). To address these challenges, we introduce a model repository-based deployment framework in which a centralized AI data center maintains a catalog of scene-specific CSI models. The repository is enhanced with a Learnware-based framework, where each model is associated with a specification including semantic part (network architecture parameters) and statistical part (codeboo-fingerprint embeddings of training-data distributions). A BS submits only its local statistical specifications to retrieve the most relevant pre-trained model, enhancing data privacy by avoiding raw CSI transmission and drastically reducing retrieval latency and communication overhead. We further develop a data-driven search strategy that matches codebook fingerprints to model performance, achieving over 90% selection accuracy. In simulations, our scheme yields 18.8% and 57.7% performance improvements over the General Model in LOS and NLOS scenarios, respectively while reducing local fine-tuning by up to 1000 samples and 100 epochs. This Learnware-based approach minimizes redundant training, maximizes model reuse, and supports rapid,privacy-enhancing deployment of CSI feedback models.

Australian women, gender-diverse individuals, and culturally and linguistically diverse (CALD) communities are often more susceptible to phishing and other forms of cybercrimes due to factors such as language barriers, limited access to cybersecurity education, and social isolation. These communities encounter substantial obstacles both entering and progressing in the cybersecurity field. In Australia, the Higher Education sector still leans heavily on a largely uniform cybersecurity curriculum, focusing heavily on technical proficiency, overlooking the vital impact of intersectionality and user-centered thinking for boosting student engagement and learning. Without gender inclusivity and proper consideration of intersectionality forms such as CALD, the workforce is deprived of the varied perspectives necessary to tackle today's intricate cybersecurity issues. In this study, we conducted semi-structured interviews with 15 experienced academics teaching and coordinating cyber security programs from a diverse range of Australian universities, covering all states, to explore their perspectives on: i) current strategies for addressing the women, gender-diverse and CALD perspective in cyber security education in the Australian HE sector; ii) barriers to incorporate women, gender-diverse and CALD perspective in cybersecurity curriculums in higher education; iii) future work and support that is needed. Our research highlights a lack of systematic methods for integrating intersectional perspectives into cybersecurity curriculums. In particular, we identified four key barriers and four areas where support and future efforts are needed to address this issue. Our findings offer vital insights that can substantially guide curriculum development in cybersecurity education.

Memory is a key capability of LLM agents. Persistent memory extends this across sessions---enabling recall, revision, and personalization. Yet its multi-stage pipeline (ingestion, retrieval, filtering, generation) makes failures difficult to localize: end-to-end evaluation reveals that an error occurred, but not which stage caused it. Existing evaluations often report aggregate performance without paired statistical comparisons, slice-level non-regression checks, or stage-level diagnostic traces. We propose D$^2$ACCI (Diagnostic-Driven Artifact-based Closed-loop Controlled Iteration), a dual-loop protocol whose outer diagnostic gate promotes, feature-flags, or rejects memory interventions based on paired evidence, protected-slice monitoring, and trace-level localizability. We further introduce DCR, a graded observability metric that measures whether failures remain localizable, and D$^2$ACCI-Eval, a reusable artifact for gate replay. We instantiate the protocol in MemStack and evaluate on three public benchmarks, achieving 93.59% on LoCoMo, 90.93% on LongMemEval, and 57.20% on PersonaMem-V2. Five paired ablations show that supplement extraction, session-memory retrieval, and Forget Guard yield statistically significant gains (+1.9 to +3.7pp, all p $\le$ .003). In contrast, BM25/RRF is retained as a monitored feature flag---a distinction invisible to aggregate-only evaluation. A diagnostic audit shows enriched traces substantially improve root-cause agreement over result-only relabeling. Diagnostic artifacts reach 98--100% DCR@3 versus 0% for results-only logs. These results establish that robust memory-system iteration demands traceable, statistically grounded, and regression-aware evidence---exactly the gap D$^2$ACCI fills.

The ubiquity of the video game industry and its large user base have transformed video games into complex social and economic ecosystems. Unfortunately, this growing popularity also attracts cybercriminals who deliberately exploit game-specific mechanisms to target players. Despite this growing threat, cybercrime in the gaming ecosystem has received little systematic attention in prior research. In this work, we present an empirical study of cybercrime affecting video game players, combining qualitative and observational analyses to characterize gaming-related attacks, identify common attack vectors and motivations, and examine player responses. Our study is based on an online survey with 57 international participants, semi-structured interviews with two confirmed victims of gaming-related cybercrime, and an analysis of 2,574 publicly available posts reporting cybercrime incidents across multiple online gaming platforms. Our findings indicate that the theft of digital items is a prevalent motivation for attacks. We further observe that gaming-related features and services, such as item trading, team voting, and tournaments, create incentives for players to engage in risky interactions. In addition, our results highlight the targeted exploitation of weaknesses in customer support processes and reveal that certain security mechanisms provide only a false sense of protection.

Short-duration heavy-rainfall warning determines whether 1 h rainfall will exceed a threshold within a target-station neighborhood over the next few hours. Multitemporal infrared and water-vapor observations from the Fengyun-4A Advanced Geostationary Radiation Imager (FY-4A AGRI) capture cloud-top cooling, moisture evolution, and cloud expansion before substantial surface rainfall develops. However, most deep-learning nowcasting methods convert these signals into local warnings by post-processing gridded precipitation predictions, preventing station-neighborhood event targets from directly supervising the satellite-to-station learning pathway. We propose MAGPIE-Net, which embeds a geographically adaptive, differentiable grid-to-station mapping in a pathway combining convection-initiation features, multiscale encoding, and auxiliary gridded precipitation diagnosis. Station-neighborhood event losses thereby constrain the satellite representation and its mapping to irregular station locations for 0-3 h event prediction. In independent 2023 warm-season tests over central and eastern China, critical success index (CSI) values under the primary 40 km/20 mm h-1 definition were 0.371, 0.304, and 0.238 at 0-1, 1-2, and 2-3 h. Across episodes, MAGPIE-Net achieved a detection rate of 65.1% and a mean lead time of 64.6 min, compared with 23.6% and 18.3 min for the best gridded-output baseline, and remained superior for smaller neighborhoods and the 50 mm h-1 threshold. During the critical early-warning stage, when antecedent 1 h rainfall within 40 km remained below 1 mm, MAGPIE-Net detected 51.9% of episodes with a mean lead time of 38.5 min. These results show that event-oriented satellite-to-station modeling converts multitemporal geostationary cloud and moisture observations into local heavy-rainfall warnings more effectively than gridded-precipitation modeling.

Decentralised partially observable Markov decision processes (DecPOMDPs) provide a general framework for modelling multi-agent decision making under uncertainty. However, DecPOMDPs are known to suffer from exponential complexity in the number of agents. One way to combat this intractability in agent numbers is to look at partitions of agents that exhibit a form of symmetry among agents, allowing for a compact encoding by counting. However, a challenge arises as the policy space explodes, even though the model complexity and evaluation cost reduce to a polynomial dependence. In this paper, we redirect our focus from counting agents to counting policies, which actually enables tractability in agent numbers for so called policy-counted DecPOMDPs. Further, we present policy-counted dynamic programming using the compact representation to solve policy-counted DecPOMDPs efficiently.

Although text-to-image diffusion models exhibit remarkable generative power, concept erasure techniques are essential for preventing harmful content. Existing adversarial probes evaluate these methods by testing whether erased concepts can still be recovered. However, existing erasure and probe methods remain largely text-centric, focusing on whether the text-to-image mapping is severed while overlooking whether the corresponding visual knowledge remains. To investigate this question from a visual perspective, we leverage diffusion inversion to probe whether a generative trajectory can reconstruct visual instances of an erased concept. Under a null-text condition, standard inversion avoids the textual pathway but amplifies approximation errors, hindering faithful trajectory recovery. To address this challenge, we introduce TINA+, a diffusion-consistent Text-free INversion Attack equipped with optimization-based inversion. We also find that unconstrained diffusion inversion may discover spurious trajectories, even allowing a randomly initialized diffusion model to reconstruct the target concept. Such trajectories may falsely indicate residual visual knowledge. TINA+ therefore introduces Diffusion-Consistent Trajectory Regularization to suppress this failure mode. By penalizing trajectories that fall far below the expected marginal energy evolution of diffusion, TINA+ suppresses spurious inversion paths while preserving its ability to recover erased concepts. Experiments across twelve erasure methods, four concept-erasure tasks, and different model architectures demonstrate that TINA+ reliably probes residual visual knowledge through diffusion-consistent visual trajectories. These results provide stronger evidence that current methods often obscure concepts by severing text-image links rather than eliminating the underlying visual knowledge.

Take three frontier mixture-of-experts models (Alibaba, OpenAI, NVIDIA; 3.6-4.0B active parameters each) and fine-tune them to reason in a low-resource language. On accuracy benchmarks almost nothing happens, and the benchmark itself is noise at this scale: changing only the random seed moves the score by 7.7 points, more than every data and recipe effect we measured. That null is our first result. The real changes live where accuracy cannot see. Base models never think in Greek: 0 of 1,000 reasoning traces, even when the question is Greek, so the model answers correctly while reasoning in a form its user cannot read, audit, or correct. After supervised fine-tuning (SFT), every released checkpoint reasons in the language of the question on ~98% of items, one family at 3x fewer tokens, with judged grammaticality improving on all four models and general ability within a few points of each base: nothing was forgotten, and fluency was gained. We propose six behavioural dimensions that make such changes measurable, each gated to reject any metric that correlates with output length, and we report how our own instruments lied: six failures, each caught by a control. What SFT cannot do is fix its own defects: a quarter of answers skip the requested format, answers leak into the reasoning channel, and an explicit "think in English" is obeyed under half the time. Reinforcement learning with verifiable rewards, pre-registered before training, fixes the first two outright (fallback 24% to 2.5%, leak 3.5% to 0.0%, both against a flat random-reward control) and moves the third (+9.1pp), while the Greek reasoning habit survives an accuracy-only gradient untouched. We release five checkpoints. The instruments, the controls and the pre-registration travel to any low-resource language; Greek is the case that let us measure them.

OWL 2 DL ontologies, grounded in the description logic $\mathcal{SROIQ}$, express large knowledge bases in biomedicine and the Semantic Web. Neuro-symbolic (NeSy) learners over description logics either embed the ontology in a continuous space, abandoning classical entailment, or restrict to the Horn fragment $\mathcal{EL}^{++}$, which has a single canonical model. We present Baobab, which compiles a $\mathcal{SROIQ}$ ontology with a finite ABox into a Sentential Decision Diagram (SDD): it saturates a propositional core under a consequence-based calculus and instantiates the remaining $\mathcal{SROIQ}$ features (nominals, number restrictions, and the role axioms) over the active domain. The SDD's evidence-conditioned weighted model count then trains a perception network to recognize real images under partial ABox supervision: on an ontology that exercises every distinctive $\mathcal{SROIQ}$ feature, a CNN learns to read MNIST digits coupled by a successor relation and recovers latent ontology concepts that an independent perception leaves at chance. When the supervision admits several ontology-consistent completions, an independent perception collapses onto one, a reasoning shortcut: we show that a mixture indexed by the query's justifications can represent the calibrated posterior no independent perception can, and that seeding it from the circuit's enumerated completions attains the Bayes-optimal posterior on a real-image MNIST task where single-WMC and learned mixtures (the BEARS-ensemble hypothesis class) do not: to our knowledge the first to characterize and mitigate reasoning shortcuts in a non-Horn description logic. Soundness of the compiler and the representation result are machine-checked in Lean 4. Code is available at https://github.com/bio-ontology-research-group/baobab.

This study investigates cooperative control of connected and automated vehicles (CAVs) at partially observable highway bottlenecks in mixed traffic, aiming to mitigate congestion without relying on complete global traffic states or online trial-and-error. We propose a physics-informed world model-based offline multi-agent reinforcement learning framework that reconstructs a physically interpretable global traffic state from local CAV observation-action histories, with coupled macroscopic-microscopic traffic dynamics providing physics-based supervision. A probabilistic ensemble world model learns traffic-state transitions and system rewards, while model disagreement quantifies epistemic uncertainty. Multi-step imagined rollouts with pessimistic rewards and uncertainty-driven truncation are then used for offline policy learning. Experiments in a SUMO-based on-ramp bottleneck using approximately $1\times10^6$ offline transitions show that physics supervision improves state reconstruction and world-model prediction accuracy.

Ethereum's consensus safety relies on independent consensus client implementations agreeing on every state transition. When they diverge due to implementation errors, the network can fork, finality can stall, and severe attacks are possible. To prevent such consensus divergences, Ethereum provides a Python reference implementation (consensus-spec), which acts as a specification, and a hand-crafted official test suite (spectests). However, as an executable implementation, Ethereum's specification defines validity implicitly through runtime behavior. As a result, it lacks a systematic way to ensure that all validity conditions are thoroughly evaluated. We present SpecTrum, a framework that addresses this problem in three stages. First, we introduce Consensus-SpecTec, a mechanized specification of the Ethereum consensus algorithm, which makes validity conditions explicit as if-premises. Second, we define premise coverage, a metric that measures which if-premises are evaluated to true and false across spectests. Third, we develop a specification-based test generator that extracts constraints on premises not evaluated to false by spectests and generates inputs to evaluate them. Applying SpecTrum to five major Ethereum consensus clients, we identify 27 cross-client divergence cases, 22 of which cannot be found without the premises inserted in our mechanization. All 27 cases reproduce across fork versions, and extending the mechanized specification to a new fork takes modest effort proportional to the specification difference.

In bus routing, the task is to plan a bus route in a network with several agents, each of whom wants to travel from a starting point to a destination. A bus route should account for several factors, including agents' cost for reaching the bus stops, their travel time, or the energy consumption of the buses. We study the complexity of several variants of this problem, focusing on how the objective function and the models for agents' walking costs influence the problem complexity. After observing that even the simplest agent cost model leads to hardness on general networks, we consider networks with tree structure. Our main findings are as follows. First, allowing agent-specific cost models leads to hardness even on extremely limited trees such as stars. Second, consistent agent models (where agents differ only in their starting points and destinations) make the problem easier in some cases. Finally, allowing agents to choose between using the bus and walking directly can make the problem considerably harder. Most of our hardness results show not only classical NP-hardness but also parameterized intractability for the natural parameter $k$, the number of bus stops.

Diversity is a fundamental criterion for evaluating generative artificial intelligence (AI) systems, yet its measurement remains inherently ambiguous. Existing approaches typically represent generated samples in an embedding space, compute pairwise distances or similarities, and aggregate them into a single scalar score. Such scalar summaries are convenient, but they often encode different inductive biases and may yield contradictory rankings of the same sample sets. In this paper, we argue that diversity evaluation for AI-generated content is intrinsically under-specified when reduced to a single number. We first review representative diversity metrics, and then diagnose their limitations from two complementary perspectives: an axiomatic analysis showing that no representative scalar metric satisfies all desirable properties simultaneously, and an empirical analysis showing that high-dimensional representation spaces can induce concentrated, modality-dependent distance distributions. To address these issues, we propose diversity profiles: curve-valued, condition-aware summaries that evaluate a parameterized diversity family across a range of thresholds, scales, exponents, or orders under a specified representation and distance or kernel function. Diversity profiles reveal whether a comparison is robust across resolutions or instead depends on an arbitrary parameter choice. We instantiate profiles for several representative metric families and demonstrate their practical use in generative AI evaluation. Overall, diversity profiles provide a more transparent and resolution-aware framework for comparing the diversity of AI-generated content.

The widespread adoption of Internet of Things (IoT) devices has expanded the digital attack surface, making firmware analysis critical for modern software security. A key security concern stems from the frequent reuse of third-party software components, a practice that often introduces known vulnerabilities into firmware images. Whether a given image actually exposes such a flaw is an open question, and public proof-of concept exploits make answering it urgent. Directed Greybox Fuzzing (DGF), a technique that enables targeted exploration of specific binary locations, offers a promising solution for detecting such vulnerabilities. However, DGF has reached firmware only at function granularity, too coarse to aim at the vulnerable block itself. This article presents BULLSEYE, the first DGF framework to schedule closed-source Linux-based firmware fuzzing by basic-block-level distance to user-specified targets. Our methodology combines static and dynamic analysis to enable DGF in the constrained firmware domain, focusing on vulnerabilities in reused third-party components. We introduce novel DGF heuristics that address limitations of traditional approaches. We compare BULLSEYE against four greybox-fuzzing baselines sharing its execution back-end, including reimplementations of AFLGO and WINDRANGER, and against GREENHOUSE, a state-of-the-art firmware re-hosting framework. On 40 vulnerability sites across 32 firmware images, BULLSEYE reproduces every target within budget, against 35 for the strongest of the four baselines, and reduces Time-to-Exposure by a geometric mean of 9.5x to 72.5x over them; against GREENHOUSE, on the 18 targets its pipeline supports, BULLSEYE is faster by a geometric mean of 9.8x.

Bridge structures are regularly inspected for structural damage such as cracks and corrosion in order to ensure public safety and reduce maintenance costs. Much research has been done on automating this process using computer vision methods, which are often evaluated and compared using metrics such as intersection over union, mean average precision, etc. However, predicting the actual effectiveness of an inspection method within the field of structural engineering from these metrics remains challenging. To enable the systematic use of these increasingly popular methods in engineering practice, evaluating the performance of these methods in a way that is compatible with standard engineering approaches is therefore an urgent necessity. We present a new statistical evaluation framework to allow the comparison of computer vision methods with conventional visual inspection for crack detection in steel bridges. The framework is based on probability of detection curves and can account for the influence of image resolution. We apply this evaluation method to the real-world ``Cracks in Steel Bridges'' dataset, which contains annotated images of cracks in bridge structures. The quantification of the probability of detection and its uncertainty enables a practical assessment of the effect of automated methods for damage detection in structural reliability analyses. In turn, this enables the wide-spread use of automated (AI-based) damage detection in safety critical applications. This evaluation method provides evidence that the proposed computer vision approach approach is robust for the crack detection task and can have a high added value as an addition to conventional visual inspection methods.

Analog gauges remain common in industrial environments where manual inspection is costly or hazardous. The engineering application addressed here is direct numerical reading of single-target analog-gauge images, while the artificial-intelligence contribution is a systematic evaluation of specialization, transfer, robustness and reliability for a general-purpose vision-language model (VLM) without an explicit pointer-segmentation and geometric-reading pipeline. The Qwen2.5-VL-7B-Instruct model is evaluated using zero-shot prompting, in-context learning (ICL) and parameter-efficient fine-tuning with Quantized Low-Rank Adaptation (QLoRA) on a public synthetic dataset, a video-derived Pressure Gauge dataset and a proprietary industrial dataset. All fine-tuning experiments use a fixed 20-epoch protocol with the final epoch used for analysis; separate models with and without supplied gauge ranges remove prompt-setting confounds. The primary metric is range-normalized mean percentage error (MPE). The best fine-tuned MPE values are 2.39% on the synthetic dataset, with a 95% bootstrap confidence interval (CI) of 1.43-3.90%; 2.61% on the Pressure Gauge dataset, with a CI of 1.66-3.80%; and 4.43% on the proprietary industrial dataset, with a CI of 2.31-7.14%. Leave-one-dataset-out experiments reveal substantial transfer degradation on held-out synthetic and proprietary data, while robustness tests identify Gaussian blur as the strongest tested corruption. Reliability analysis shows that high-confidence errors remain possible, motivating abstention and independent validation in safety-critical use. These results support QLoRA-specialized VLMs for direct single-gauge reading but not yet a deployment-ready plant-monitoring pipeline.

Vision-Language models (VLMs) achieve outstanding performance largely due to the amount of training data available on the internet. At the same time, data holders (e.g., artists) urgently need to determine whether their data has been used for model training without authorization, which concerns both intellectual property rights and personal privacy. Data auditing, particularly through membership inference (MI), has attracted attention as a direct tool. This work proposes MemCatalyst, a set of data poisoning tools, aiming to amplify the data auditing performance on VLMs. MemCatalyst employs two strategies: Poisoning Text (PT) and Poisoning Image (PI). MemCatalyst forces VLMs to over-learn specific inconsistencies between image features and textual semantics during training, thereby increasing their susceptibility to membership information auditing. Crucially, the transferability of poisoned samples across different VLM architectures is demonstrated to be effective in the black-box setting. Extensive evaluations using five state-of-the-art data audits on two prominent VLMs demonstrate that MemCatalyst markedly enhances MI AUC scores with a minimal budget of poisoned samples, while maintaining a negligible impact on model performance.

The quantum approximate optimization algorithm (QAOA) targets NP-hard maximum-likelihood (ML) detection in multiple-input multiple-output (MIMO) systems. Existing $M$-ary quadrature amplitude modulation (M-QAM) detectors design angles by expected Ising energy: online per instance, warm-started, or ramped, while train-once designs remain B/QPSK-only or block-local, leaving M-QAM without a size-scalable benchmark. Their infinite-size spin-glass theory assumes independent disorder, matching the retained covariances at B/QPSK but not M-QAM's correlated couplings and fields. We develop a correlated infinite-size multi-species spin-glass framework whose covariance-matched evaluators make that energy an offline objective with a size-scalable benchmark. In addition, the ML rate, the exponential rate of sampling the ML string, is for the first time exploited for QAOA angle design in MIMO detection. The energy evaluator is $q$-free at $O(p\,4^p)$ cost while the ML rate transfers angles from a fixed $q_{\rm ref}$-qubit reference. Tests reach 4096-QAM, 128 antennas, $p=30$ and per-symbol SNR 0-45 dB. In simulations, ML rates fall as a power law $r_0\,p^{-α}$, with larger exponents for the sampling design, which tracks exact ML at $5\times5$ 16-QAM (0-20 dB) and $3\times3$ 64-QAM (8-28 dB) while its bit-error rate (BER) advantage widens with SNR to two orders of magnitude. The approach points toward near-optimum decoding on deeper noiseless fault-tolerant quantum (FTQ) circuits.

Context: Software systems that depend on commercial large language model APIs must migrate to successor versions when vendors deprecate older models. Migration decisions typically rely on aggregate benchmark scores, which compress heterogeneous item-level behaviour into a single net figure. Objective: We measure what that compression conceals. Method: On three pairwise upgrades in the GPT-5.4 to GPT-5.6 Sol product sequence, we query 900 public benchmark items (graduate-level knowledge, olympiad mathematics, instruction following) 50 times per item per model, classify each item as reliably improved, reliably regressed, practically equivalent, or inconclusive under false-discovery-rate control and a practical-significance threshold, and calibrate the results against a label-permutation null. Results: Across all nine migration-benchmark cells, reliable improvements and reliable regressions coexist. Edges with aggregate gains of up to 7.3 percentage points contain up to 8.3% reliably regressed items; edges with aggregate losses contain up to 10.7% reliably improved items. On the instruction-following benchmark, the gap between strict and loose scoring widens by 3.9 percentage points on the latest migration: a 3.9-point regression under strict scoring shrinks to 0.04 points under loose scoring. Conclusion: Migration decisions based on aggregate scores alone miss substantial bidirectional item-level change. The complete response-level archive and per-item scoring outputs are released.

Long-horizon agents increasingly operate across many steps, tools, and observa- tions. In this setting, the relevant oversight question is not only whether each action is locally valid, but whether the evolving trajectory still corresponds to the task the user authorized. Drift can accumulate quietly: an agent may call the right tool with plausible arguments at every step, while its prefix moves toward a broader role, an adjacent objective, or evidence the user never supplied. Existing monitors mostly check local compliance, deliver final-trace verdicts, or score generic risk; they do not directly estimate this prefix-level relation. We introduce ontological trust, a task-conditioned property of trajectory prefixes, and instantiate it as RGE, an online monitor that decomposes trust along Role, Goal, and Evidence. RGE uses LLMs only to derive structured task and step representations; trust-state updates, projec- tions, and intervention decisions are deterministic, so the output is a replayable and auditable trust trajectory rather than a single end-to-end judge verdict. We construct a cross-domain trajectory corpus from OSWorld, FinanceBench, and EICU-AC, covering benign executions, prefix-paired drift, and pseudo-consistency failures. On this corpus, RGE outperforms adapted rule-, judge-, and shield-style baselines on prefix-paired drift detection. With the two larger estimator models, it exceeds 93% Drift F1 on every benchmark while keeping benign coverage at or above 95.8%. Pseudo-consistency is harder: detection depends on whether task completion is externally visible, a structural limit we characterize empirically.