Papers

Enterprise AI assistants must produce responses that are verifiable and traceable to source evidence. However, retrieval augmented generation (RAG) over heterogeneous enterprise data can suffer from citation drift, unsupported content, and weak source traceability. We present EvidenT (T = Trust + Transparency + Traceability), a lightweight pipeline that verifies extracted evidence against retrieved documents before answer generation, without model retraining. EvidenT combines structured passage extraction with deterministic lexical alignment to filter unsupported content, correct citation drift, and preserve source-span traceability. On approximately 500 real enterprise queries, EvidenT improves gold-source hit rate by an average of 29% over prompting baselines, produces no citations to nonretrieved urls, and achieves near-saturated answer-to-source lexical coverage.

Bangla, the seventh most spoken language in the world, exhibits significant regional dialectal diversity, with dialects such as Barishal, Chattogram, Sylhet, Noakhali, and Mymensingh differing in lexical, morphological, and syntactic characteristics. These variations pose substantial challenges for Named Entity Recognition (NER), limiting the generalization of models trained on Standard Bangla or a single regional dialect. This paper presents a cross-dialect Bangla NER framework using the publicly available ANCHOLIK-NER dataset, comprising 17,405 annotated sentences and 101,817 tokens across five major Bangla regional dialects. A Leave-One-Dialect-Out Cross-Validation (LODOCV) strategy is adopted, training models on four dialects and evaluating on the remaining unseen dialect. Eight pretrained transformer-based models, including BanglaBERT, MuRIL, XLM-RoBERTa, and Multilingual-E5, are evaluated under identical experimental settings. Multilingual-E5 Large achieves the highest F1-score in every fold, peaking at 97.26% on Mymensingh and reaching its lowest, 82.38%, on Chattogram, the most challenging target dialect. To improve interpretability, Local Interpretable Model-agnostic Explanations (LIME) are applied to word-level predictions, revealing that the model's decisions are driven primarily by the surface form of the target entity word itself rather than by surrounding sentence context. These findings establish a benchmark for cross-dialect Bangla NER and demonstrate the effectiveness of transformer-based transfer learning for low-resource regional dialects, while highlighting the surface-form dependence of current models as a direction for future work.

Determining the microscopic Hamiltonian of a quantum magnet from inelastic neutron scattering measurements remains a central challenge in condensed matter physics. While the dynamical spin structure factor contains, in principle, sufficient information about the underlying interactions, extracting Hamiltonian parameters from experimental spectra constitutes a highly nontrivial inverse problem due to the large parameter space and the presence of experimental noise. Here we demonstrate that neural networks can efficiently solve this inverse problem for a broad class of quantum magnets. As a benchmark, we consider an eight-parameter family of honeycomb lattice Heisenberg spin models in the fully polarized phase, where the dynamical spin structure factor can be computed exactly within linear spin-wave theory. Using a large synthetic dataset of neutron scattering spectra, we train neural networks to infer the underlying Hamiltonian parameters directly from the dynamical response. We compare the performance of three different architectures—fully connected neural networks (FCNNs), one-dimensional convolutional neural networks (CNN1Ds), and two-dimensional convolutional neural networks (CNN2Ds)—and find that all achieve high predictive accuracy. Remarkably, the trained models remain robust in the presence of substantial experimental uncertainty, reliably recovering the Hamiltonian parameters even when the input spectra are contaminated by random noise with amplitudes reaching $10\%$ of the signal intensity. Our results establish machine learning as a powerful framework for quantitative Hamiltonian reconstruction from neutron scattering data and provide a practical route toward automated characterization of quantum magnetic materials.

Strategic classification studies how a decision maker should choose a classifier when the agents being classified can adjust their features in response to it. In existing models, the classifier is the only instrument available to the decision maker, and therefore a feature that is predictive but easy to fake can only be down-weighted or discarded. However, in many settings the decision maker can also verify: lenders verify income, admissions offices check documents, and tax authorities audit returns. In this paper, we propose a model of strategic classification in which the firm jointly designs a linear classifier and an audit profile, which assigns to each fakeable feature a probability of detection and a penalty when caught. We show that under linear costs, the classifier affects the audit problem only through the distribution of gaming rents it induces, so that the joint design problem decomposes into the choice of a score rule and an audit allocation problem. We use this decomposition to characterize the optimal audit allocation, to identify when the allocation problem is tractable and when it is NP-hard (namely, when agents can game through overlapping features under an inspection cap), and to bound the regret of a firm that has to learn the rents by auditing. We further show that audit intensity is a quantity to be tuned rather than maximized: welfare is single-peaked in it, and a firm and a social planner disagree on the mix of detection and penalty that delivers a given level of deterrence. Notably, two populations with identical costs and causal structure can game in one domain and improve in the other, a difference that a cost-only model cannot account for. Together, our findings highlight that whether a feature is “gameable” depends on the institution's verification policy as much as on the feature itself.

We investigate the nonequilibrium phase-ordering kinetics of a binary mixture of uniaxial ellipsoidal particles interacting via the anisotropic Gay-Berne potential using large-scale molecular dynamics simulations. Following a temperature quench from the isotropic and homogeneous phase into the coexistence region, the system exhibits two distinct ordering processes occurring on different time scales. At first, rapid orientational ordering leads to the formation and coarsening of nematic domains, characterized by dynamical scaling, generalized Porod-law behavior, and a growth law $\ell(t) \sim t^{1/2}$ consistent with nonconserved order-parameter dynamics. Later, the system undergoes binary phase separation driven by compositional fluctuations. This compositional ordering process shows a transition from diffusive growth $\ell(t) \sim t^{1/3}$ to hydrodynamic regimes, with finite-size scaling analysis indicating asymptotic viscous growth $\ell(t) \sim t$. Despite the presence of global orientational order, the phase-separation kinetics remains effectively isotropic. These results demonstrate a clear separation between orientational and compositional ordering dynamics, governed by distinct conservation laws, and provide a unified framework for understanding phase ordering in anisotropic particle systems.

Stellar magnetic activity alters thousands of spectral lines, limiting high-precision radial-velocity measurements, abundance analyses, and planetary characterization. We present a preliminary classification of 6403 atomic lines in synthetic visible spectra from a sequence of NLTE semi-empirical dG2 atmospheric models with increasing chromospheric heating. Principal-component analysis shows that the dominant component traces overall response amplitude, while the second distinguishes early from late responders. DBSCAN applied in the full nine-dimensional response space identifies a dense stable core and an activity- sensitive non-core group comprising about 11% of the lines. The stable core provides candidate lines for activity-insensitive measurements, while sensitive lines provide candidate activity diagnostics. Sensitive transitions tend to have low lower-level energies, although the populations overlap, making lower-level energy a statistical discriminator rather than a line-by-line predictor. A spectral sensitivity map shows that small average variations can hide nonlinear or compensating responses, demonstrating the value of the complete line-response trajectory for classification.

Dashboards, feeds, and leaderboards determine whom agents believe they are playing against. We study how such information architectures shape strategic behavior, and how a designer who controls them can use visibility as a lever. We propose a behavioral model in which a transparency parameter governs how selectively agents are exposed to opponents of nearby reasoning depth, and embed it in a Cognitive Hierarchy environment, which provides a tractable language for tying agents' beliefs about the sophistication of others to what they see. We show that because the exposure operator is an exponential tilt of the type distribution, the effect of transparency on aggregate behavior has a closed form: the derivative of the aggregate action with respect to $p$ is proportional to the population-average variance of perceived sophistication. This identity signs the effect in games of strategic complements (opacity raises aggregate effort) and in coordination environments (opacity lowers the mean action), and an ordinal version holds for any response rule that is monotone in the belief. Notably, under a Poisson prior, opacity acts as a sophistication multiplier: every reasoner plays as if the population were $\mathrm{Poisson}(\tau/p)$. We use this structure to characterize the disagreement-versus-level tradeoff faced by a coordination-minded designer, to identify the reasoning depths that respond most to visibility, and to show that opacity imposes a “sophistication tax” under which the deepest reasoners can lose to moderate ones. Simulations in which agents see only a noisy leaderboard of published actions reproduce both the reversal and the tax. Together, our findings highlight that who an agent sees, and not only how deeply it reasons, shapes strategic behavior, and point to visibility as a design lever.

The interplay between optical spin and crystal symmetry provides a powerful mechanism to control light-matter interaction, enabling advances in imaging, holography, sensing, signal processing, and more. Here we reveal that this interplay can also be exploited as a programmable mechanism for selecting and shaping nonlinear frequency-conversion pathways. Using a z-cut lithium niobate thin film, we experimentally demonstrate spin-selective switching between complementary three-wave mixing pathways, where co- and cross-circular-polarization excitations activate sum- and difference-frequency generation, respectively. We further generalize this concept to establish a quantitative spin-to-frequency mapping, that encodes the pump spin composition into the generated frequencies, and enables the reverse retrieval of the pump spin state. Finally, we leverage this concept also to demonstrate mapping of spin patterns to a combination of frequency-dependent real space images, and Fourier-space diffraction patterns. These results pave the way for all-optical programable frequency-conversion and spatiotemporal nonlinear signal manipulation in free-space and integrated photonic systems.

International law provides the normative framework through which states coordinate action, regulate armed conflict, and protect human rights, yet its texts remain without token-level named entity recognition (NER) resources. We introduce IntLawNER, a NER dataset and benchmark for codified sources of international law, covering 2,987 gold-annotated sentences and 8,094 entity spans from International Court of Justice (ICJ) decisions, UN Security Council resolutions, and European Court of Human Rights (ECtHR) judgments, annotated with seven institution-specific entity types. We construct IntLawNER with a cost-effective hybrid algorithmic-agentic pipeline that reduces 468k source sentences to a compact annotation set through candidate retrieval, LLM-based vetting, and human review, with 89.6% of gold spans accepted unchanged from the silver layer. However, the silver-to-gold analysis reveals that human-machine aggregate agreement metrics can be misleading in domain-specific NER: Cohen's kappa=0.964 on boundary-matched spans masks a macro-F1 of 0.753 when missing entities, boundary errors, and label corrections are included. The benchmark shows that zero-shot span-based GLiNER collapses on entity types dependent on institutional function rather than surface form (0.243 micro-F1), while fine-tuned transformers struggle on rare labels. Carefully selected few-shot examples that demonstrate label contrasts improve every LLM over zero-shot prompting, with Claude Opus 4.6 reaching the best score of 0.873 micro-F1. We release IntLawNER as a benchmark and reusable resource for extracting references in international legal texts.

We establish several sufficient conditions ensuring that a perturbed family of operators in a given Hilbert-Schmidt frame for a separable Hilbert space remains a Hilbert-Schmidt frame. The perturbation criteria are formulated through Nagy-type and Pollard-Hilding-type inequalities, yielding explicit estimates for the resulting frame bounds. Several examples are presented to illustrate the applicability of the results.

We study a one-dimensional diffusive particle subject to stochastic resetting to its initial position, in the presence of an imperfect, localized target that can absorb (kill) the particle, modeled by a delta-function killing rate. The central question is how stochastic resetting competes with drift-induced transience and the target's finite reactivity. Exact Laplace-space expressions are derived for the non-normalized propagator and survival probability, for arbitrary diffusion coefficient $D$, resetting rate $r$, and killing strength $k$, first in the absence of drift and then under a constant drift. The long-time behavior separates into distinct regimes. Without resetting, unbiased diffusion exhibits the recurrent algebraic survival law $S(t)\sim t^{-1/2}$, whereas any nonzero drift renders the motion transient with respect to the target and leaves a nonzero ultimate survival probability. By contrast, any $r>0$ and $k>0$ repeatedly renews encounters with the target and restores eventual absorption, yielding an exponential survival law $S(t)\sim A e^{-\theta t}$. The decay rate $\theta$ is given by the dominant (rightmost) pole of the Laplace transform, and the surviving density converges after normalization to an explicit quasi-stationary profile. The driftless and perfectly absorbing limits recover standard resetting results. These formulas provide a unified description of the crossover between recurrence-controlled, transience-controlled, and renewal-controlled survival.

In this letter, we investigate the interaction between stationary point-like external sources within the framework of Pseudo-Quantum Electrodynamics (PQED). Besides ordinary electric charges, we consider configurations involving Dirac points and generalized topological sources. For the configurations analyzed, we compute the corresponding interaction energies, forces, torques, and electromagnetic field configurations. Furthermore, we compare the results obtained in PQED with those arising in ordinary planar Maxwell electrodynamics, showing that the nonlocal character of PQED gives rise to physical effects with no counterpart in the conventional local planar theory. In particular, our results show that topological sources induce intrinsically anisotropic and non-central interactions, leading to the emergence of nontrivial torques between the interacting objects. The directional character of these interactions suggests possible applications in the effective description of topological defects and anisotropic phenomena in planar condensed matter systems.

The Precision Time Protocol (PTP) provides the time and phase synchronization required by disaggregated Open Radio Access Networks (O-RAN). Yet, in current open fronthaul deployments, PTP traffic lacks mandatory authentication and integrity protection, leaving synchronization vulnerable to spoofing, replay, and delay manipulation attacks that can degrade radio access performance. Existing protections are poorly suited to this setting: they either add excessive latency, do not support multicast dissemination efficiently, or fail to contain key exposure under partially trusted RUs. This paper analyzes the security risks of unprotected O-RAN PTP and develops a threat model for open fronthaul deployments. We then introduce PRTESLA-C, a lightweight synchronization protection mechanism that combines per-round delayed key disclosure with ASCON-based message authentication. PRTESLA-C uses an apply-then-verify-and-correct paradigm: timing samples are applied immediately to preserve real-time control, verified after key disclosure, and removed from persistent synchronization state if authentication fails. This design maintains sub-microsecond synchronization accuracy, provides strong protection against spoofing and replay, and bounds the impact of delay manipulation with minimal computational and latency overhead.

Models of opinion dynamics on networks, provide a framework to study how a network of agents aggregates dispersed opinions into a consensus. However, existing models assume that agents truthfully report their opinions, and do not account for environments in which a platform aggregates agents' reports across multiple networks, and agents can strategically report different opinions on each. In this paper, we propose a model of strategic opinion manipulation on multiplex networks, in which a platform merges L network layers with attention weights, and each agent, holding a private opinion, chooses (potentially different) reports on each layer, at a layer-specific misreporting cost, so as to pull the merged consensus toward their own opinion. We show that this game has a unique Nash equilibrium in closed form, that the resulting consensus is the truthful consensus under a centrality tilted toward a manipulability index of each agent, and that the resulting distortion is the (centrality-weighted) covariance of agents' manipulability and opinions. We further show that agents' reports on every layer are more extreme than their opinions, so that the polarization observed by the platform overestimates the true polarization. Notably, we highlight that merging layers is a double-edged sword: as manipulability depends on the square of an agent's influence on each layer, spreading attention across layers dilutes manipulation, while heterogeneous misreporting costs across layers and shifts in agents' centralities can amplify it. Finally, we characterize the platform's optimal choice of attention weights, in closed form when the layers share a stationary distribution, and through an exact gradient and a marginal test otherwise. Together, our findings shed light on when aggregating opinions across networks is (not) robust to strategic manipulation, and point out potential interventions to alleviate it.

We study frames of unilateral iterations generated by finitely many vectors under an invertible bounded operator $T$. Suppose that the spectrum of $T$ lies in $\{\rho e^{it}:0<\rho\le1,\ |t|\le c\}$, where $c<\pi$, and that its essential spectrum lies in $\{e^{it}:|t|\le c_{\mathrm e}\}$, where $0\le c_{\mathrm e}\le c$. Real powers are defined through the principal logarithm. Our structural result identifies fractional-orbit synthesis, up to an invertible map and a compact perturbation, with synthesis of projected vector-valued exponentials on an arc containing the essential spectrum. Semi-Fredholm properties and Fredholm index therefore transfer. For a uniformly separated temporal set $\Lambda$, this reduction gives a frame when its one-sided lower uniform density satisfies $D_+^-(\Lambda)>c_{\mathrm e}/\pi$ and its logarithmic block density satisfies $L(\Lambda)>c/\pi$. Completeness at the latter threshold requires only an invertible operator with spectrum in the corresponding angular sector and a complete integer multiorbit. At $L(\Lambda)>2c/\pi$, the complement property is equivalent to a finite condition in channel-cyclic subspaces, with real phase-retrieval consequences. Generalized divided differences admit the same compact reduction and the analogous frame theorem with multiplicities counted. For arbitrary one-point-per-cell sampling with cell length $r$, the sufficient raw-sampling conditions are $rc<\pi$ and $2rc_{\mathrm e}<\pi$. Thus, the full range $rc<\pi$ holds for raw samples when the essential spectrum is contained in $\{1\}$; over the unrestricted class, the sharp universal range is $rc<\pi/2$. Divided-difference preconditioning restores $rc<\pi$ for all operators considered and converts exact collisions into logarithmic-derivative data.

Cosine similarity is widely used to analyze transformer representations, implicitly assuming that similarity reflects task-relevant structure. We study when this assumption fails in dialogue-conditioned large language models. Across three 7-8B chat-tuned models, ambient cosine similarity substantially underestimates linearly decodable persona structure on the same hidden states; numerically, linear probe AUC is in the 0.73-0.97 range while cosine kNN is in the 0.56-0.77 range on a 30-class task. A low-dimensional supervised subspace recovers much of this gap, whereas a matched-rank PCA subspace does not and in some cases degrades performance. This mismatch is regime-dependent: it is absent in single-sentence sentiment classification (SST-5), and a matched-cardinality control rules out attribute cardinality as a confound. The gap does not systematically increase across dialogue turns, and the task-aligned subspace remains stable over time. However, two of three models violate a pre-registered within-subspace separability invariance criterion (|Delta AUC| <= 0.03), and one model violates a pre-registered turn-invariance criterion (|Delta L| <= 0.05). These results show that cosine similarity can fail to reflect task-aligned structure in dialogue representations even when that structure is linearly accessible.

The performance of learned robot visuomotor policies depends heavily on the size and quality of their training data, yet collecting high-quality demonstrations remains costly for robots in the real world. Although large-scale robot and human datasets are increasingly available, embodiment gaps and mismatched action spaces make them difficult to leverage directly. Cross-embodiment transfer, reusing experience from other embodiments to improve learning on a target embodiment, is therefore crucial for scaling robot learning beyond per-robot data collection. In this work, we find that efficient transfer can be achieved by learning from what is shared across embodiments, the visual dynamics of how the world responds to motion, and by effectively exploiting the scarce target-embodiment data at test time. The proposed framework, called Latent Policy Steering (LPS), implements an embodiment-agnostic pretraining phase, which trains an image-based World Model (WM) with optical flow across diverse embodiments. The resulting WM is finetuned on the target embodiment with robot actions. It then steers the base policy toward better actions by searching in the WM's latent space for plans that stay close to the finetuning data. LPS is a policy-agnostic framework: it can flexibly accommodate different policies without having to retrain them. In Robomimic and real-world evaluations, LPS improves the average performance of Diffusion Policy relatively by 16% and 62%, and Pi0.5 by 8% and 14%, with only 50 demonstrations on an unseen target embodiment.

In this paper, we consider diffusion processes on a graph. On the $i$-th edge, each process is governed by an operator of the form $D_{v_i}D^+_{u_i}$, and the behavior at the vertices is determined by the gluing conditions. The paper characterizes the weak convergence of a family of such processes to a limiting process in terms of the behavior of the scale and speed functions $u_i^n$ and $v_i^n$ and the parameters of the gluing conditions. This may be viewed as a generalization of Freidlin and Wentzell's result on the weak convergence of a family of processes on a line.

Recent advances in generative artificial intelligence have enabled the synthesis of complex human motion with unprecedented fidelity. However, current text-to-motion systems rely strictly on linguistic semantics: if an input reads "I put my hands up", the model searches for a pose with raised hands, and every non-semantic property of the text is discarded as noise. In this work, we treat that discarded structure as the signal. We present an embodied visualization instrument that amplifies not what a text means, but how it is built. Our method first quantizes dance kinematics into a motion codebook of 256 stylistic "regions" using Principal Component Analysis and K-Means clustering, and orders those regions along the dominant axis of movement. We then map the raw byte representation of any input text directly onto this codebook, producing a deterministic sequence of regions that we call the text's "choreographic genome". A precomputed plausibility graph and a set of physics smoothing routines turn this genome into fluid, full-body movement, so that the dancing body becomes a display surface for the byte-level structure that semantic systems ignore. Through a series of artistic case studies, including a Shakespeare sonnet, a machine error log, source code, an abolitionist's question, and Indigenous and Devanagari scripts, we show that each text produces a visibly distinct dance, and that scripts marginalized by ASCII-centric computing are amplified into close to three times as much movement per character. We frame this not as a motion-synthesis benchmark, but as a critical and poetic visualization that asks what we choose to count as signal, and what we allow to go unheard.

Recent work has used the method of characteristic gluing to construct spacetimes violating the third law of black hole mechanics. This paper uses numerical methods to investigate two aspects of the gluing construction of solutions describing the gravitational collapse of a charged scalar field to form an extremal Reissner-Nordström black hole. First, for a massless scalar field, we improve significantly the lower bounds on the size of the final black hole for which gluing is possible. Second, for a massive scalar field, it is known that the formation of an extremal Reissner–Nordström black hole is impossible if the mass-to-charge ratio of the scalar field exceeds unity. We present strong evidence that this bound is sharp by constructing gluing solutions with a scalar field mass-to-charge ratio just below unity.

Poor clinical communication can delay care, contribute to errors, and harm patients, yet opportunities for repeated practice with feedback remain limited. Our prior randomized trial showed that practice with the SOPHIE AI patient platform improved serious illness communication, but the system addressed a single clinical context and required human effort for delivery and assessment. We developed SOPHIE 2.0, a browser-based, self-service platform integrating embodied AI-patient interactions, personalized feedback, and automated assessment across 24 clinical scenarios. An automated large language model assessor evaluated three communication skills—Empower, Be Explicit, and Empathize—with agreement comparable to individual human raters ($r=0.759$; ICC$=0.746$). In a study of 59 clinicians and students, participants completed two AI-patient encounters with personalized feedback; 92% found the platform engaging, 86% easy to use, and 83% clinically relevant. Scores were higher in the second encounter, though the uncontrolled design precludes attributing this change specifically to training.

We present the first Gaia-calibrated estimate of the contribution of dynamically processed open clusters (OCs) to the nearby solar-type field binary population and investigate the implications for primordial planet stability. We combine Gaia DR3 observations, empirical models of cluster formation and dissolution, and direct N-body simulations to constrain the fraction of field FGK-type binaries within 1 kpc that originated in OCs over the past 5 Gyr. We find that at least ~53% of solar-type field binaries formed in OCs with initial masses > 200 M_Sun, where dynamical processing becomes significant. We quantify the effects of stellar encounters and binary orbital evolution on the stability of both circumbinary and circumstellar planets by deriving empirical prescriptions for circumbinary planet destabilization as a function of binary separation and for the fractional reduction of stable circumstellar coplanar phase space. These results provide a framework for evaluating how OC dynamical evolution may influence the stability of planets in binary systems and therefore contribute to observed exoplanet demographics.

A hot Brownian particle is driven by thermal fluctuations from a nonuniformly heated solvent and therefore obeys a fluctuation-dissipation relation with a frequency-dependent effective temperature. Existing theories assume an incompressible solvent. We show that finite sound speed qualitatively changes this effective temperature in the kinetic regime. Solving the compressible fluctuating-hydrodynamic problem for a heated sphere, we find that the effective noise temperature develops a pronounced acoustic dip when the sound wavelength becomes comparable to the particle radius. The dip occurs because weakly attenuated longitudinal modes carry mechanical energy away from the heated surface before it is dissipated, so that the dissipation is weighted by colder regions of the temperature field. Its position is set by $a\omega/c=\mathcal{O}(1)$, while its depth is controlled by the dimensionless viscous-acoustic attenuation length $a_c/a$. The result identifies an acoustic window in hot Brownian motion in which compressibility lowers, rather than raises, the kinetic noise temperature.

We define a natural orientation on a spanning-tree-decorated planar map whereby, roughly speaking, each directed edge in the map is oriented to match the direction of the contour exploration of the spanning tree. We study directed distances (lengths of shortest directed paths) with respect to this orientation. We construct the Busemann function which measures directed distances to $\infty$ along a natural interface in the uniform infinite spanning-tree-decorated map. We show that this Busemann function, re-scaled appropriately, converges in law to a $3/2$-stable Lévy process. We also show that in a uniform spanning-tree-decorated map with $n$ edges, directed distances are typically of order $n^{1/3}$. Using a strong coupling argument, we deduce analogous statements for directed distances in other random planar maps in the $\sqrt 2$-Liouville quantum gravity (LQG) universality class, including uniform meandric systems and mated-CRT maps for $\gamma=\sqrt 2$. These results give the scaling dimension for a hypothetical directed version of the $\sqrt 2$-LQG metric. Our proof strategy is inspired by work of Borga and Gwynne (2025) on directed distances in bipolar-oriented triangulations.

An investigation of local aspects of the space of axisymmetric traveling-wave solutions to the three-dimensional Euler equations near Hill's zero-flux spherical vortex and Norbury's small-flux vortex rings is the endeavor undertaken herein. The aforementioned family consists of idealized, swirl-free, and nonsmooth objects, their azimuthal relative vorticity being prescribed by a scaled characteristic function of the compact vortex core. We confront the desingularization question, both within the swirl-free solution subspace and upon admitting swirl, and prove that each member of the Hill–Norbury family of sufficiently small flux is the limit of traveling-wave solutions to the Euler equations that are smooth to infinite order and have compact toroidal vorticity support. In the latter swirling regime, however, a curious flexibility is revealed: total vorticity may diverge along sequences converging in a natural desingularization topology. Our construction, founded upon a five-dimensional reformulation of the Stokes stream function, is an amalgamation of delicate thin-domain analysis, singular-parameter Newton-type nonlinear inversion, and topological fixed-point arguments. To the best of the authors' knowledge, this constitutes the first verification that the small-flux Hill–Norbury family is nowhere isolated from the collection of smooth vortex rings.

Consensus among LLM judges is often taken as strong evidence that a decision is correct. This assumes that judges make their errors independently. In practice, LLM judges are often trained and evaluated in similar ways, so they can make the same mistakes. We study how this dependency affects the reliability of consensus. We find substantial error correlation across both open-weight and frontier LLM judges. In our main bank of ten judges, the average pairwise correlation between judge errors is 0.21. As a result, the ten judges only provide roughly as much statistical information as 3.5 independent judges. The dependency is even stronger among the high-accuracy frontier judges we evaluate, including judges from different providers. In up to 28% of our comparisons, ignoring shared errors leads to the conclusion that one system is significantly better, while accounting for them does not. We also find that the pattern of errors matters. Errors shared by most judges and errors concentrated among a smaller group affect consensus differently and favor different voting methods. Measuring the overall amount of correlation alone is therefore insufficient. Our results suggest a simple approach: use a small set of trusted examples to estimate judge accuracy and identify shared mistakes. These shared errors should then be considered when analyzing the results, and the voting method should be chosen using trusted examples before it is applied to new data.

Identity cues are irrelevant to a fixed programming specification, but raw counterfactual differences can arise from unequal samples and prompt wording. We audit 30.73 million executed Python generations from seven checkpoints on HumanEval+ and MBPP+, supplemented by an exploratory 550B slice, under model-assigned gender, country, and occupation personas. Within-task randomization and false-discovery-rate control identify occupation as the most consistent structural axis: CodeBLEU dispersion exceeds its exchangeability null in 10/14 model–benchmark cells, remains significant in 8/12 full-coverage cells, and exceeds the country ratio in every paired cell, although the median excess is only 0.141 points. In six high-pass-rate cells, occupation dispersion replicates across token similarity, length, comments, reference similarity, and complexity, while pass-rate dispersion is significant in none. Country leads raw dispersion in 12/14 cells but has a median calibrated ratio of 1.00. Gender-associated variation cannot be separated from persona wording in this design. Thus, the evidence supports small, reproducible occupation-conditioned changes in code form, not stable disadvantage to named identities or demonstrated downstream harm. We use this finding to motivate a broader commentary on the potential impacts of bias in code generation, including the possibility that models may condition their outputs on identity information available from prior conversational context.

As LLM applications integrate with external tools, they are increasingly exposed to indirect prompt injection (IPI), where adversarial instructions are embedded in retrieved content. Conventional IPIs fire on contact: the moment an agent ingests the content, it carries out the instruction. We introduce the explosive prompt, a conditional payload that stays dormant until an attacker-chosen trigger is met, in effect a training-free, inference-time backdoor planted in a single piece of retrieved content. This temporal separation reaches where ordinary IPI cannot. On frontier models that refuse the bare imperative almost entirely, rephrasing the same goal as a dormant conditional drives real, state-changing tool execution against a live agent backend (a paired mean of 16.5% vs. 2.4% for the imperative, reaching 34.2% on a proprietary model). In trials on nine production agents (OpenAI Codex, Google Gemini CLI, Anthropic Claude Code CLI, Cursor CLI, GitHub Copilot, Devin AI CLI, Amazon Kiro CLI, Qwen Code, Google Assistant; n=30 each), explosive prompts succeed in 43-83% of cases versus at most 3% for an imperative baseline, and they slip past deployed defenses: off-the-shelf injection classifiers are miscalibrated on them, and a preference-optimized model that closes imperative injection entirely still executes 11.8% of explosive prompts, every one at the trigger turn. The durable defensive lever is ingestion-time detection of the conditional structure, once detectors are trained on explosive-prompt data, which no prior benchmark supplied and our generator does. Retraining cuts live tool-execution attack success from an undefended 34.3% to 7.5-8.1% for the encoder baselines. Our detector, DeFuse, reaches 3.0% at a calibrated 5% false-positive budget with the best detection quality of any method tested (AUC 0.9994) and 25x lower latency, though it needs length-aware thresholds.

We investigate the impact of a monolithic three-dimensional (M3D) integration process-critical $400^\circ$C post-processing forming gas anneal (FGA) on the electrical performance, reliability, and defect evolution of oxide-channel thin-film transistors (TFTs), combining systematic experiments with density-functional-theory (DFT)-based liquid-quench molecular-dynamics (MD) simulations. Indium tungsten oxide (IWO) TFTs are employed as a model system and encapsulated with a thin 3 nm Al$_2$O$_3$ / 3 nm HfO$_2$ hybrid layer that effectively suppresses external hydrogen ingress. We reveal a non-monotonic evolution of device behavior during FGA, governed by initial densification followed by partial crystallization. Short-duration FGA (10 min) induces channel densification and the formation of shallow, delocalized defect states, leading to pronounced positive bias temperature instability (PBTI) degradation and the emergence of a characteristic transfer-curve "kink." With prolonged annealing (>40 min), partial crystallization of the oxide channel occurs, stabilizing hydrogen in deep, localized defect states, suppressing hydrogen mobility, and restoring device reliability. As a result, the PBTI shift is reduced to 10.4 mV after 2000 seconds of stress, accompanied by complete elimination of the transfer-curve kink. These findings provide a mechanistic understanding of hydrogen-defect interactions during high-temperature post-processing FGA and demonstrate that appropriate hydrogen-blocking encapsulation enables oxide-channel TFT integration without compromising electrical performance or reliability.

Generative Adversarial Networks (GANs) have demonstrated strong performance in generating high-quality synthetic data. However, they are limited by no formal guarantees regarding convergence and the effectiveness of the learning process. In practice, this leads to training instability, mode collapse, and sensitivity to hyperparameters. To address this, we propose EmbeddGAN, a novel adversarial training framework based on a dependence-based objective. Instead of relying on a discriminator that classifies samples as real or fake, EmbeddGAN introduces an embedding network that learns a representation in which statistical dependence between samples and their real/fake labels is maximized, while the generator is trained to minimize this dependence. This objective is implemented using the Gini distance correlation (gCor), which equals zero if and only if the embeddings are statistically independent of the real/fake label. Minimizing this objective therefore encourages real and generated samples to become statistically indistinguishable in the learned embedding space. The embedding network projects both real and generated data into a shared low-dimensional space, where distributional discrepancies can be measured directly through pairwise distances. We adopt a minimax training strategy: the embedding network maximizes the Gini distance correlation (maximizing dependence), while the generator minimizes it (minimizing dependence). Experiments on the MNIST, CIFAR-10, and CelebA datasets demonstrate that EmbeddGAN achieves competitive performance relative to established baselines while exhibiting notably stable training dynamics on the evaluated datasets.