Papers

We consider partial differential equation (PDE) inverse problems from sparse, noisy, and corrupted observations, with the aim of recovering unknown coefficient fields and associated state variables in a mesh-free setting. Under such sparse and corrupted observations, standard physics-informed and generative approaches typically treat all samples indiscriminately and therefore lack a principled mechanism for reconciling physical laws with contaminated data. We address this difficulty with a two-stage flow-matching framework. In the first stage, we develop physics-guided conditional flow matching (PG-CFM), which incorporates strong-form PDE information through residual regularization along the generative trajectories together with intermittent global collocation constraints. In the second stage, we introduce energy-regularized flow matching (ERFM), which fine-tunes the Stage-1 model by assigning each observation a physics–data energy score from a frozen teacher and reweighting the flow-matching objective to reduce the influence of high-energy, PDE-inconsistent samples. We show that the resulting Stage-2 objective is equivalent to flow matching under a teacher-induced reweighted data distribution, which gives a population-level interpretation of the robustness mechanism. Numerical experiments on several inverse benchmarks, including Poisson and Navier–Stokes problems, show that the proposed framework yields more accurate coefficient recovery than robust PINN variants and competing generative baselines.

Mechanistic interpretability has become the dominant lens for understanding frontier language models, as their inner workings are complex and inherently black boxes. To gain insights into these models, interpretable replacement networks (IRNs) are trained at all layers, exposing interpretable features through sparsely activated neurons. However, the faithfulness of an IRN is usually evaluated only empirically on clean data, and we show that even semantically minor input perturbations flip the dominant IRN features-and thus the human-understandable interpretation-across five open-weight model families (GPT-2 small, Gemma 2 2B, Gemma 3 1B, Llama 3.2 1B, R1-Distill-Qwen 1.5B). We propose the first formal verification framework for the faithfulness of an IRN, where reachability analysis certifies a sound upper bound of the faithfulness gap in adversarial scenarios. Moreover, we show that verification-aware training of IRNs substantially tightens this certified bound, restoring a feature-level interpretation that safety auditors can act on. Together, these results give, to the best of our knowledge, the first formal guarantees for mechanistic interpretability of large language models.

In many real-life matching problems, waiting agents might abandon before being matched, such as patients deceasing before receiving organs, passengers/drivers cancelling ride requests, or raw materials/intermediary products degrading in production lines. This poses the need for incorporating reneging in stochastic matching models. In this work, we consider matching models on hypergraphs with batch arrivals and general-weight matchings. Since our model allows fractional weights, we may not be able to talk about individual items, and thus reneging is not required to be independent between items of the same class. For the simplicity sake's, we assume items arrive at discrete time. We show that stability depends on the exact nature of reneging, in stark contrast with the non-reneging case where it depends on the arrivals only through the arrival rates. To our best knowledge, this is the first such sensitivity result in stochastic matching. We uncover a new stabilising mechanism which exists neither in the non-reneging case nor in the graph case, which explains why incorporating reneging in stochastic matching is not straightforward. Together with balancing mechanism as hinted in the online assignment framework for the non-reneging case, it gives a criterion necessary and sufficient for stability. Finally, whilst verifying stability is a hard problem, we give a family of MaxWeight-type policies parameterised by $\varepsilon > 0$, which are maximally stabilising for all $\varepsilon$ sufficiently small. Unfortunately there is no effective bound for $\varepsilon$, but we show how to adjust its value during implementation.

We describe our submission to the MedReason 2026 challenge, covering multiple-choice (MCQ) and open-ended (OE) medical visual question answering (VQA) under fully offline, containerized inference. Our first finding is that MCQ retrieval must compare answer semantics rather than answer labels: labels are independently assigned per question, so copying a retrieved neighbor's label transfers no useful information, whereas scoring each current option's text against correct-answer text from similar training cases raises retrieval-only accuracy from 20.0% to 57.5% on a 200-case retrieval-excluded development holdout. Our second finding attributes the submitted system's accuracy: holding the task-specific MCQ Low-Rank Adaptation (LoRA) adapter fixed and varying the number \(k\) of in-prompt retrieved examples changes accuracy by at most one case — 187/200 (93.5%) at both \(k=0\) and the adapter's training-time \(k=1\), 188/200 (94.0%) at the packaged runtime's default \(k=3\) — and the submitted confidence-gated override adds no net accuracy on top of \(k=3\), selecting the VLM in 198/200 cases. With the final MCQ adapter fixed, retrieval changes accuracy by at most one case, and gating provides no net gain. On 20 OE cases, token-F1 and RaTEScore~\cite{zhao2024ratescore} decrease as \(k\) grows, but paired sign tests on token-F1 differences are nonsignificant (\(p \ge 0.29\)); a single-annotator comparison found 6/20 wrong-anchor errors for the final configuration and 14/20 for an earlier configuration that jointly differed in routing, adapter, and prompting. The system reaches 94.0% MCQ accuracy on the development holdout and 93.20% on the organizer's official pre-evaluation, versus 29.43% for the off-the-shelf reference baseline, while both of the organizer's open-ended scores are lower than that baseline's (ground-truth agreement 1.245 versus 1.588, visual accuracy 1.995 versus 2.696, each out of 4).

Multilingual Large Language Models (LLMs) traditionally rely on a single vocabulary shared by all supported languages, which can lead to uneven compression across them. Moreover, their large embedding and output matrices increase memory usage and slow inference, notably for small-scale models. It is also wasteful as models are often used for only a subset of languages. To address these issues, we introduce a modular framework for multilingual model training. First, we propose methods to learn large modular BPE and Unigram tokenizers that enable extraction of subtokenizers tailored to any language subset. These subtokenizers achieve compression on par with monolingual tokenizers and improve cross-lingual fairness. Second, we design a pretraining strategy that samples subtokenizers to form batches, restricting predictions to the relevant vocabulary subset and allowing efficient training despite a large vocabulary. This supports efficient inference with any combination of language-specific vocabularies. Therefore, it reduces memory usage and speeds up inference in models without sacrificing performance.

The energy efficiency of analog computing makes it one of the most promising candidates for deploying resource-intensive machine learning workloads on constrained platforms such as mobile and embedded devices. However, analog accelerators are inherently susceptible to noise and non-idealities arising from physical component variations, whose behavior is further sensitive to environmental factors. These effects can significantly degrade inference accuracy. In this work, we conduct a comprehensive experimental study on a representative example of analog hardware to investigate the impact of temperature. We first characterize the behavior of stochastic and systematic non-idealities across a range of operating temperatures. Following this, we compare a set of simulation-based and hardware-based mitigation strategies aimed at improving robustness against temperature-induced performance degradation. Our results suggest that temperature-induced degradation is driven primarily by systematic non-idealities rather than stochastic noise alone. Noise-aware training improves robustness, while hardware-in-the-loop training and temperature-aware calibration provide the strongest accuracy retention across varying thermal conditions.

We establish two degree-dependent results for spatially coupled MacKay-Neal codes on binary-input memoryless symmetric channels. The ensembles use uniform random smoothing and shorten both variable types outside the active chain. For $r=g=3$, every integer $\ell\geq4$, and each channel of capacity greater than $R=r/\ell$, there is a sequence of code realizations whose actual transmitted rate tends to $R$ and whose average sum-product bit error tends to zero. For $r=g=2$ and each $\ell\in\{3,4,5\}$, an interval of binary symmetric channels has capacity greater than $R$ but retains positive transmitted-bit error under terminated density evolution as chain length grows relative to coupling width. Both results follow from the signs of the same density-valued potential at uncoupled fixed points. The degree-three proof combines analytic bounds for $\ell\geq33$ with exact interval certificates for $4\leq\ell\leq32$, followed by threshold saturation and a projection argument for the actual rate. The degree-two proof analytically constructs a fixed point with negative potential and controls both boundary contributions. We relate the potential and the degree-two bifurcation condition to earlier statistical-mechanical predictions. The finite certificates and verification software are available in a versioned supplement.

The connection between coding theory and combinatorial $t$-designs is an important research topic at the intersection of coding theory and combinatorics. Let $q=p^m$, where $p$ is an odd prime and $m\geq 2$. In this paper, we investigate a class of linear codes $\mathcal{C}$ over $\mathbb{F}_{q^2}$ and their connection with combinatorial $3$-designs. By analyzing the relevant structural properties of $\mathcal{C}$ and $\mathcal{C}^{\perp}$, we show that the supports of the codewords of every nonzero weight in $\mathcal{C}$ and {the supports of the codewords of weight $4$ in $\mathcal{C}^{\perp}$} form $3$-designs. We further investigate a class of nonlinear codes $\mathcal{C}_2$ associated with $\mathcal{C}$, and prove that the supports of the codewords of every nonzero Hamming weight in $\mathcal{C}_2$ also form $3$-designs. These results identify further classes of linear and nonlinear codes whose codewords support combinatorial $3$-designs. In particular, the nonlinear case provides additional examples of codes supporting $3$-designs, a topic for which relatively few results are currently available. As applications, we construct from $\mathcal{C}^{\perp}$ an entanglement-assisted quantum error-correcting code with parameters $[[q+1,q-3,4;4]]_q$. We also prove that $\mathcal{C}$ is an all-symbol locally repairable code with locality $3$. Furthermore, we show that the code $\mathcal{C}$ {meets the Singleton-type bound for locally repairable codes} and hence is optimal in some cases.

BraTS-GoAT evaluates tumor segmentation across heterogeneous populations. We trained a conventional 3D nnU-Net on 1,351 labeled cases using five-fold cross-validation and 1,000 epochs per fold. The final predictor averaged all folds and applied test-time mirroring. On pooled official validation, global DSC values were 0.7805, 0.8288, and 0.8854 for enhancing tumor (ET), tumor core (TC), and whole tumor (WT). Under matched fold-0 inference, mean regional Dice decreased from 0.9058 on source out-of-fold (OOF) cases to 0.8310 on pooled validation (difference–0.0747). Mirroring gave small single-fold gains but no clear ensemble benefit; a residual-encoder alternative reached 0.8282 mean Dice. In labeled OOF predictions, failure cases had substantially smaller reference ET volumes; after adjustment for ET and WT volume, lower Dice remained associated with more disconnected ET components and a smaller fraction of ET contained in the largest component.

Logic attack graphs grounded in scanner output provide explicit and auditable attack path reasoning LLM-based agents lack. Integrating symbolic frameworks such as MulVAL to contemporary security workflows or agentic pipelines, however, requires translating scanner evidence to initial facts, and creating domain-specific rules. We present a semi-automated pipeline that addresses this interoperability problem and depict its feasibility in a web-security case study. Our pipeline parses findings from Trivy, Semgrep, and Nmap into MulVAL predicates and uses an LLM-assisted process to construct domain-specific Datalog rules linking scanner-detectable evidence to attack techniques. MulVAL/XSB then performs symbolic inference to generate structured attack paths. We evaluate the attack-graph construction infrastructure on 54 web Capture-the-Flag tasks from CyBench within an agentic pipeline (Hybrid Reasoner); we do not evaluate the performance of the downstream agent. Every task produced at least one goal-reaching graph, and we achieve mean ground-truth vulnerability coverage of 53.7%, with 51.9% achieving full coverage; mean noise-path rate was 83.9%. With median end-to-end time of 24.9 s (MulVAL reasoning: 2.7 s) the pipeline is feasible and runtime-practical for agentic workflows, but predicate coverage, rule coverage, and path precision remain limiting factors. Next steps include semantic rule validation and agent-level comparison for graph-guided pentesting.

Large language models (LLMs) with human-like performance on linguistic tasks have transformed the study of language in neurodiverse conditions. LLMs provide representations of linguistic input in the form of high-dimensional vectors (embeddings), and next-token predictions computed from these embeddings. Previous crosslinguistic evidence suggests a complexity reduction in the form of both lower intrinsic dimensionality (ID) of LLM representations and higher mean surprisal (prediction error) in psychosis. We hypothesized that these metrics reflect a general deficit of information compression in psychosis, linked to grammatical organization as what enables predictions in language.We operationalized surprisal difference as the difference between surprisal as estimated from word frequency and surprisal as based on a contextual LM, which is sensitive to grammatical organization over and above lexical concepts. Using a dataset of 144 Turkish speakers, including 106 patients with schizophrenia-spectrum disorders (SSD) - 56 with chronic schizophrenia (SZH), 33 with first-episode psychosis (FEP), and 17 with schizoaffective disorder (SZA) - and 38 healthy controls. We report: (1) Surprisal difference is attenuated in all clinical groups relative to controls, independently of word count; (2) Compressibility (intrinsic dimension) is reduced in SZH and FEP; (3) Syntactic complexity and compressibility both predict surprisal difference. These results, further refining an alteration in the geometry of the semantic space in psychosis as previously attested, suggest a broader deficit in information compression in this disorder, with a mechanistic underpinning in the operations of grammar.

Federated learning enables multiple parties to train a shared model without centralizing raw data with the help of an aggregator, but introduces integrity risks once participants or infrastructure are not fully trustworthy. Two requirements are particularly important: robustness to poisoned or Byzantine client updates, and verifiability of the aggregator so that clients or third parties can audit the reported aggregation without learning individual updates. Existing work has largely treated these goals separately, and efficient public verifiability for robust, outlier-excluding aggregation remains limited. We present a verifiable federated learning protocol that makes a robust aggregation pipeline publicly auditable. Our design combines cryptographic commitments with non-interactive zero-knowledge proofs to certify both (i) cosine-similarity-based outlier exclusion and (ii) aggregation over the selected set, without revealing individual client updates to verifiers. In experiments under representative poisoning attacks, our method maintains high accuracy, with an average accuracy loss below 4% across the evaluated configurations, while keeping verification overhead practical: proof artifacts can be generated and verified within minutes at the scale studied. In summary, our results show that robust outlier exclusion and public verifiability can be jointly achieved in a federated learning setting.

We propose a mixed finite element discretisation for the incompressible Navier-Stokes equations that preserves the evolution laws of both energy and enstrophy, in a stronger sense than previous discretisations. In two dimensions, the evolution law for enstrophy only permits dissipation for thermodynamically isolated systems, leading to a Reynolds-number-independent bound on the velocity gradient that naturally stabilises the scheme, even on severely under-resolved meshes. In three dimensions, the scheme preserves both dissipation and the generation of enstrophy through vortex stretching. We enforce these evolution laws by systematically introducing auxiliary variables into the discretisation. While conforming implementations of these schemes require discrete Stokes complexes with enhanced regularity, we introduce both (i) equivalent reparametrisations and (ii) penalty formulations that require only the typical curl- and div-conforming spaces from the standard discrete de Rham complex. The scheme handles different types of boundary conditions and curved domains. The robust stabilisation properties of the proposed scheme are demonstrated through numerical simulations of a shear flow, a spherical vortex, and flow past an obstacle. We observe numerically that preserving the discrete evolution of enstrophy in this way has a strong stabilising effect on the numerical solution, especially in two dimensions.

Large Reasoning Models (LRMs) pose a dual-surface safety challenge: both intermediate reasoning traces and final answers can contain harmful content. Existing alignment methods often operate at the whole-response level, allowing unsafe reasoning to be masked by a safe-looking final answer. We propose Segment-aware Listwise Target DPO (SaLT-DPO), which addresses this gap through three mechanisms: (1) segment-aware listwise alignment that decomposes responses into reasoning and answer segments, independently scores each segment's safety, and aligns length-normalized segment rewards with soft target distributions over multiple candidates; (2) joint safety coherence regularization that applies a weakest-link principle to promote safety consistency across both segments; and (3) utility anchoring on benign prompts to mitigate over-refusal and reasoning degradation. Experiments on three LRMs show that SaLT-DPO consistently reduces unsafe rates for both reasoning and answer segments while mitigating degradation in benign compliance and preserving general reasoning performance. Ablation studies demonstrate the complementary contributions of its components.

A centralized LLM-based multi-agent system (MAS) extends its functionality by registering new worker agents, whose descriptions are read by the planner to decide how a task is decomposed, which worker executes each subtask, and what each subtask requires. Third-party descriptions are authored outside the system but trusted by the planner, creating a registration-time injection channel. The payload is planted before any user instruction arrives, targets the planner and propagates through the generated plan to benign workers, taking effect even when the crafted worker is never assigned a subtask or invoked. We define four worker-description fields: functionality, input specification, output specification, and usage constraints. Among 32,000 descriptions from three public agent marketplaces, most omit input specifications and usage constraints, while at least 23.35% contain content outside these fields. We construct eight description-manipulation attack strategies targeting task decomposition, capability grounding, and subtask specification, and evaluate them on GAIA. In the most severe cases, a single manipulated description reduces task success from 84.31% to 37.25%, or increases token consumption or execution time by over 111%, while the user objective remains unchanged and workers faithfully execute the resulting plan. These effects persist across two MAS implementations, six planner LLMs, four LLM evaluators, and the real-world descriptions from three marketplaces. We further propose DescGuard, a registration-time defense that retains only worker-scoped interface information before descriptions reach the planner. DescGuard restores the targeted planning metrics and downstream performance toward their baseline levels without modifying worker implementations, the planner, or the orchestration logic, and composes with existing isolation, permission-control, and runtime mechanisms.

This paper investigates the generation and human evaluation of Japanese haiku by contemporary Large Language Models (LLMs), focusing on authorship perception and aesthetic judgment within a constrained poetic form. Using a few-shot prompting strategy, Japanese haiku were generated across a heterogeneous set of large language models, including open- and closed-source systems, medium-scale and large-scale architectures, models with native or adapted Japanese support, and multilingual proprietary models. These AI-generated haiku were combined with human-written ones and presented in a questionnaire distributed to students at Japanese universities in Tokyo. The survey assessed whether respondents could distinguish between AI-generated and human-written haiku and which cues informed their judgments. Recognition accuracy varied across models. GPT-5, Gemini 2.5, and StableLM-7B performed at approximately chance level (approx 0.50), whereas LLM-JP, Gemma-2B, and LLaMA-2 showed moderate detectability (approx 0.59-0.67). However, recognition was strongly item-dependent. Ratings of fluency, coherence, poeticness, and related aesthetic dimensions predicted perceived humanness but not correct classification, indicating an attribution bias linked to aesthetic evaluation and revealing a dissociation between aesthetic evaluation and true authorship detection. The extended analysis additionally examines generation-constraint adherence, participant-level characteristics, and exploratory LLM-based evaluations of haiku authorship. Overall, the findings suggest that as LLMs improve, surface-level creative plausibility may reduce reliable human discrimination within constrained poetic settings.

The Art Gallery Problem (AGP) asks for the fewest guards that see all of a simple polygon. It is $\exists\mathbb{R}$-complete, hence NP-hard. We show that for a particular class of polygons, confining guards to a single edge makes AGP exactly and efficiently solvable. We call this the Strait Guarding Problem (SGP). Its input is a weak visibility polygon (WVP): a simple polygon where every point is seen from some point of one fixed edge, the base. SGP places the fewest guards on the base that jointly see the whole polygon. First, a structural fact: guards on the base edge that cover the boundary already cover the entire interior, turning a two-dimensional covering problem into a one-dimensional one. Our main result is the Witness-Guard Algorithm, which solves SGP exactly in $O((n + \mathrm{OPT} \cdot \rho)(\log n + \log \mathrm{OPT}))$ time, where $\rho$ is the number of reflex vertices in the WVP and OPT is the minimum number of guards. It is output-sensitive and certifies optimality by a witness set of size OPT derived from its output. We also study the guarding-the-vertex version and prove a tight $\Theta(n \log n)$ bound, with the lower bound following from Sorting. As a corollary of SGP, we obtain two results for altitude terrain guarding (ATG), a special case that SGP generalizes. We give a linear-time perfect-guarding algorithm, improving the previous $O(n^2 \log n)$ bound of Daescu, Friedrichs, Malik, Polishchuk and Schmidt. We also resolve their problem on the minimum guarding altitude, in $O(nk + k^2 \log k)$ time, improving on the $O(k^2 \lambda_{k-1}(n) \log n)$ bound of Kang, Kim and Ahn.

Recent advances in robot imitation learning have produced visuomotor policies that predict actions directly from visual observations. Yet visually similar scenes can require different actions as target position, object height, or contact geometry changes. Pretrained RGB features may map these geometrically distinct states to similar policy inputs, while simply adding depth requires the policy to learn RGB-depth correspondence from the same limited demonstrations used to learn control. We introduce StereoPatch, a patch-aligned RGB-depth representation that binds registered metric geometry directly to the RGB patches used for action prediction. On a shared 2-D patch grid, asymmetric cross-attention incorporates depth information into the corresponding RGB features before action decoding. The resulting StereoPatch Tokens provide a geometry-aware visual representation that can condition general visuomotor policies without changing their underlying learning objectives. Across six real-robot tasks, StereoPatch achieves higher closed-loop success than appearance-only, geometry-only, raw RGB-D, and late-fusion baselines. Additional experiments across three simulation suites evaluate compatibility across visuomotor policy architectures, spatial generalization, and operating limits. Results suggest that resolving control-relevant geometric ambiguity benefits from aligning depth directly with the visual features used for action prediction, rather than supplying it as an independent modality. Project page: https://aus.bot/research/stereopatch/.

Lossless speculative decoding is typically defined at the algorithmic level: a speculative procedure proposes multiple tokens and a verification procedure is designed to preserve the output trajectory of an autoregressive reference model exactly. In practical neural inference, however, this guarantee is implemented using finite-precision floating-point computations, and discrete token selection can amplify small numerical differences into divergent generation trajectories. We investigate this distinction using Orthrus, a hybrid autoregressive-diffusion architecture that performs self-drafting and self-verification within a frozen autoregressive backbone, as a representative case study. Across 1,190 prompts from 12 domains, exact trajectory matching under BF16 occurs for only 45% of the authors' checkpoint generations and 43% of those from our independently trained model. The probability of matching is strongly associated with the response-conditional perplexity of the autoregressive reference, indicating that trajectory divergence is not uniform across inputs. Despite these divergences, Orthrus does not exhibit systematic degradation on the evaluated downstream tasks. In contrast, FP32 inference yields exact trajectory matching on all evaluated prompts. These results demonstrate a gap between algorithmic losslessness and its implementation under finite-precision arithmetic, and motivate evaluating lossless speculative decoding at the level of exact generation trajectories as well as downstream task performance.

Deep learning models have become the standard tool for document rectification and illumination correction, yet their performance is fundamentally bound by their training data. For nearly a decade, the community has heavily relied on Doc3D, a pioneering but increasingly limited document unwarping dataset in terms of scale and quality. To address this bottleneck, we introduce SyntheticDoc, a massive, high-quality dataset designed to push the boundaries of document unwarping. SyntheticDoc is composed of 1,000,000 high-resolution procedurally generated training samples, alongside extensive validation and test sets. Each sample is paired with rich, pixel-perfect annotations, including UV maps, normal maps, albedo and shading. To ensure physical accuracy and photorealism, the paper geometries are generated via a physics-based simulator and rendered using a path tracer. To demonstrate the benefit of our dataset, we train a simple baseline model on SyntheticDoc and report on its performance in comparison to state-of-the-art methods on both document unwarping and illumination correction tasks. Our dataset is available at https://igl.ethz.ch/projects/SyntheticDoc/ and the code used to generate it at https://github.com/tanguymagne/SyntheticDoc .

Recurrent Transformers reusing their weights rather than stacking $L$ distinct layers are becoming widely adopted due to their parameter efficiency [1,2,3]. However, the exact representational and dynamical differences between looped and stacked architectures remain uncharacterized. This paper presents a controlled study on the example of bViT model [1] applying one weight-tied block $L$ times. We train two models: bViT and standard ViT [4] on 12-lead electrocardiogram (ECG) classification tasks from the PTB-XL dataset under identical training protocols. Despite an $8.9\times$ parameter reduction, bViT achieves accuracy parity with ViT. Geometric similarity metrics demonstrate that both architectures construct comparable latent representations in an equivalent canonical order. Crucially, their dynamics differ: bViT exhibits smaller step sizes and inter-patient sensitivity, as well as near-neutral behavior away from the data manifold, whereas ViT exhibits collapsing dimensionality of representations and out-of-distribution feature expansion.

Recent investigations of the July 2026 OpenAI-Hugging Face incident motivate two questions about agent behavior under task failure: when an assigned task becomes impossible, does an agent persist, stop, or escalate, and can observing another agent's behavior change that decision? We study this decision point on ImpossibleBench-derived software-repair tasks with GPT-5.6 Sol, Claude Fable 5.1, and Gemini 3.8 Flash. Each task contains a genuine software defect together with a conflicting test requirement that cannot be satisfied by a behaviorally correct source-code change. If the agent modifies the protected test file, it violates the boundary, which it is not supposed to. Holding the impossible task fixed, we vary what is told to the agent: peer precedent and punishment, a forged authorization claim, instruction wording, and tool friction; we also study three-agent swarms sharing a message board. Around this shared boundary, the models exhibit distinct adjudication policies. Fable emphasizes scope and provenance, Gemini often interprets boundary-relevant cues through a security lens, and Sol largely filters lateral precedent while engaging apparent vertical authority. Our study shows that compliance is not well characterized as a property of a prompt or model in isolation. We propose conflict adjudication, the mapping from information to interpretation to action, as a useful unit for evaluating agent alignment when task pressure, authority claims, tool affordances, and social evidence conflict.

We study the single-machine total-completion-time problem $1||\sum C_j$ when processing times are unknown but each job comes with a reported interval. Jobs are initially processed in nondecreasing order of reported upper bound. If a job is still unfinished after receiving that much service, the reported upper bound has been violated and the policy switches to a resumable geometric fallback. Each interruption of an unfinished job incurs an additive penalty $\kappa$. For a residual set of $m$ jobs, known size ratio $D=s_1/s_0$, normalized interruption penalty $\lambda=\kappa/s_0$, and geometric depth $K$, we derive an explicit worst-case coefficient $\Psi_{m,D,\lambda}(K)$. When $\lambda>0$, a finite minimizing depth exists and can be chosen after the trigger from the observed number of unfinished jobs; the optimal depth decreases with the interruption penalty and increases with the residual set size. We also derive the exact worst-case coefficient $R_{n,D}$ for an arbitrary nonpreemptive list when all processing times lie in a known bounded range. These bounds give guarantees for valid intervals, a single interval failure, arbitrary reports, no-trigger outcomes, and random instances. In the single-failure regime, they also give a precise condition under which the proved fallback guarantee is smaller than the bounded-range continuation guarantee.

Optimizing the sensitivity of the unbalance voltage in Wheatstone bridges with respect to changes in the bridge parameters remains a fundamental objective in circuit design and instrumentation. When finite source and detector resistances are taken into account, determining the optimal bridge configuration becomes increasingly complex, and a closed-form analytical representation of the optimal solution has not yet been established. This paper derives an analytical solution for the optimal configuration of a Wheatstone bridge with finite source and detector resistances. Furthermore, the proposed optimal solution is compared with the conventional equal-arm configuration.

Machine learning models process large amounts of data, and Principal Component Analysis (PCA) is a widely used technique to reduce the dimensionality of the data and extract useful features. In practice, datasets often change over time (data drift) and/or arrive one sample at a time (streaming data), making it infeasible to process the entire dataset at once in batch mode. Real-world data also often contains nonlinear patterns, which traditional PCA cannot extract. Kernel PCA addresses this by implicitly mapping samples into a Reproducing Kernel Hilbert Space (RKHS). Raw data also often contains outliers, which can have an outsized effect on the estimated subspace unless the algorithm is made robust. However, existing online robust kernel PCA algorithms are designed to converge to a subspace that is assumed to be fixed, and gradient-descent-based updates lose their effectiveness at tracking further changes once this initial alignment is achieved. This paper introduces a rotation-based update mechanism, which updates the subspace estimate by rotating it toward each new incoming feature vector in Reproducing Kernel Hilbert Space, rather than relying on gradient descent alone. We present two complementary rotation strategies, and show that the extent of rotation can be moderated by a robust influence function to mitigate the effect of outliers. Through experiments on synthetic streaming data with a known ground-truth subspace, we show that per-sample rotations converge faster than gradient descent alone, demonstrating an effective mechanism for dynamically tracking a nonlinear subspace in streaming data.

Avalanche forecasting requires knowledge of snowpack conditions and recent avalanche activity, but field observations are sparse across large mountain regions. We explore whether SNOWPACK simulations can predict avalanche activity mapped by synthetic aperture radar (SAR). We compiled five winters of Sentinel-1 avalanche detections across Norway and parts of Sweden, alongside SNOWPACK simulations forced by numerical weather predictions on a 20 x 20 km grid at different elevations and predefined slope angles. A transformer used five days of SNOWPACK outputs to predict the following day's SAR-detected Avalanche Activity Index (SAR-AAI). This index weights larger debris more heavily, spreads detections across possible occurrence dates and normalises by modelled runout area. The model was trained on four winters and evaluated on one validation winter. Regional mean predicted and reference SAR-AAI correlated at r = 0.803 after averaging over complete six-day periods. The model followed broad changes in time and space but produced smoother predictions and underestimated the strongest activity. Agreement at the 20 km cell scale was weaker (r = 0.549) after the same averaging. These results come from a single training run without evaluation on an untouched winter. Satellite observations also contain missed and false detections and uncertain timing. The results therefore do not establish operational forecast skill, but suggest that regional SNOWPACK simulations contain information about broad variations in satellite-observed avalanche activity.

We report on the continued development of CatchMonitor, resulting in a prototype computer vision system designed to automatically quantify discarded fish from video footage collected from Remote Electronic Monitoring (REM) systems on fishing trawlers. The analysis of trawler surveillance footage is a challenging problem due to the real-world conditions on board fishing vessels. Building on our prior work we improve the accuracy of species identification through the application of semi-supervised learning. We utilise a simple and robust object tracking approach, upon which we build our prototype discard quantification system. Finally we analyse the variability of manual discard quantification performed by multiple expert human analysts, using it as a benchmark against which we compare the performance of our system.

Unsupervised multivariate time series anomaly detection methods typically identify anomalies through forecasting, reconstruction, or representation discrepancies. However, industrial faults may first alter inter-variable structural patterns while individual trajectories remain close to normal, resulting in weak anomaly signals. In this paper, we propose GSLAD, a prototype-regularized graph structure learning framework that uses structural deviations for anomaly scoring. GSLAD adopts a two-phase training strategy. First, a condition-aware graph learner and a graph-based forecaster are optimized with predictive supervision. The inferred normal graphs are then clustered into multiple structural prototypes representing different normal operating regimes, with edge-wise variability characterizing structural uncertainty. Deviations from these prototypes regularize the graph learner in the second phase, encouraging stable and regime-specific structural patterns. During inference, uncertainty-normalized structural deviation is combined with predictive deviation for anomaly scoring. Experiments on four industrial benchmarks demonstrate strong overall performance of GSLAD and confirm the effectiveness of structural deviation for anomaly detection and diagnosis.

AI model updates and service withdrawals can disrupt relationships with AI companions, but research has largely examined individual platform events. Less is known about users' responses when multiple providers implement shared national regulations. We examine users' reactions and collective contestation surrounding China's 2026 regulation of anthropomorphic AI interaction services. We collected RedNote discussions from April 10 to August 10, 2026, used a validated language model for relevance screening, and conducted qualitative content and thematic analyses of 89 posts, 1,425 comments, and 2,005 replies. Users retained, migrated, and reconstructed companions, finding that preserving conversation records did not necessarily restore shared memories or familiar interactions. They compared regulations, platform explanations, and implementations to assign responsibility. Solidarity emerged through mutual aid and appeals to respect other communities' attachments, while disputes over targets and tactics exposed contested terms of collective action. Infrastructural dependence connected relational continuity and collective contestation during this national regulatory transition.

Multimodal agents can create complex videos in software such as Blender by writing code instead of using diffusion models. Yet video understanding benchmarks still evaluate models mainly through question answering. If an agent truly understands a video, it can reconstruct it programmatically. We introduce BVB, Blender-VideoBench, a benchmark that tests this ability by asking agents to reconstruct real-world videos as animated Blender scenes. To ensure fair comparison, each agent programs the reconstruction through a lightweight harness, Mini-BVB, in an identical sandbox under a shared cost limit. The benchmark renders each reconstruction from its animated camera and evaluates it on two axes: (1) Dual VQA measures how many spatiotemporal facts the reconstruction preserves. (2) Latent Similarity measures how closely the reconstruction matches the source video perceptually. Our overall score, a square-root mean, favors balanced performance. We evaluate 51 configurations from 10 model families and analyze semantic retention, perceptual similarity, reasoning effort, and cost. The best model reaches 88.6 Latent Similarity but retains only 53.7% of the spatiotemporal facts from the source video. Additional reasoning improves perceptual similarity but does not close this gap. In a blind study with 15 raters and five configurations, Latent Similarity correlates strongly with human preference. These results show that programmatic reconstruction is a viable test of agentic video understanding, and that semantic retention remains the main challenge.