Papers

Contemporary data governance has converged on participation. Citizens assemblies, deliberative polls, codesign workshops, and community advisory panels are now routinely suggested to build trust and secure a social license for the reuse of data. Yet this consensus rests on an unexamined assumption: that giving affected communities a voice will, on its own, change what data holders do. This paper argues that the field needs to complement voice with leverage: the capacity of individuals and communities to make their participation consequential by attaching credible legal, institutional, contractual, economic, or reputational costs to being ignored. Drawing on labor relations bargaining theory, the sociology of disputes, and James C. Scott's account of legibility, we develop leverage as an analytic category defined by three properties: it is counterfactual, enforceable, and durable. We further argue that leverage presupposes epistemic legibility affected publics can rarely impose costs on a system they cannot see, document, or contest; and that most existing transparency instruments equip regulators and procurers rather than the governed. We then map six sources from which communities and data subjects can derive leverage (legal and regulatory; institutional decision; contractual; economic; data access; and political and reputational), examine the intermediary structures that aggregate and sustain it, and propose a six dimension evaluative framework: obligation, enforceability, consequence, monitoring equipment, revocability, and institutional durability for auditing whether any social license process actually redistributes power. The paper's central claim is that a social license worth the name must share the properties of a legal license: conditions, a licensor, and enforceable remedies for breach.

Contrastive Language-Image Pretraining (CLIP) is a building block of many machine learning applications. Scaling laws have guided resource allocation for large-scale training, yet prior work treats total CLIP model size as a single variable, without exploring how the capacity split between encoders impacts downstream performance. Here, we train multiple CLIP models with different vision and text encoder sizes, revealing that for most vision encoders, there is an optimal text encoder size beyond which zero-shot performance degrades—even as total parameter count increases. Exploiting this behavior yields efficient configurations that match the zero-shot performance of the standard ViT-B/16 architecture with up to 55% fewer parameters. We further show that this degradation stems from overfitting induced by the oversized text encoder, and that using modality-specific weight decay coefficients not only recovers but improves performance across all degraded configurations. A geometric analysis reveals a trade-off in which scaling the text encoder improves embedding uniformity but worsens cross-modal alignment; we further show that these metrics are predictive of zero-shot performance. We hope these findings motivate CLIP architectures and training methods that counteract this degradation, a prerequisite for scaling CLIP reliably and efficiently.

Carrier Strike Group (CSG) communications must sustain mission traffic across heterogeneous links whose quality varies with distance, weather, fading, and intermittent outages. We present an integrated Software-Defined Networking (SDN)/Software-Defined Radio (SDR) framework that unifies 5G and satellite communications (SATCOM) under a single control plane for dual-path failover in CSG communications. The framework is evaluated via stochastic simulation across four operational regions: Norfolk, the Norwegian Sea, the Philippine Sea, and the North Pacific storm track. The framework couples a centralized SDN controller with a Rician-faded 5G channel under Adaptive Modulation and Coding (AMC), a Ku-band SATCOM channel with ITU-R P.618/P.838 rain attenuation, and Gilbert-Elliott burst-availability processes on both links. The controller routes each packet by traffic class and current link state. Experiments use 60 independent trials of 50,000 packets per parameter setting with common random numbers, paired confidence intervals, and convergence diagnostics across fault, fading, and region sweeps. For the Norfolk weather baseline, SDN routing reaches 93.09% reliability versus 36.22% (5G-only) and 89.87% (SATCOM-only), with paired gains of +56.87 and +3.22 percentage points. The advantage grows under joint low-Rician-K/high-burst-fault stress, and AMC improves 5G robustness as Signal-to-Interference-plus-Noise Ratio degrades. SDN remains the highest-reliability mode across all four regions, indicating that cross-layer SDN-controlled failover improves maritime resilience without relying on deterministic link abstractions.

Overall word error rate hides clinically critical errors: a transcript can be 95% correct and still swap one drug for another. The usual fix weights errors on medical entities, and almost always depends on an evaluation-time named-entity recognition (NER) model or cloud API, which makes the metric's denominator a versioned black box. We present MedWER, an evaluation protocol and open-source tool for medical ASR whose denominator is a fixed, license-clean term list: 19,373 drug, diagnosis, symptom, and injury-mechanism entries projected from public sources. The protocol couples a pinned text normalizer with a phrase-aware term-restricted WER, the MedWER, so the only versioned component is a normalizer dependency held at an exact release and checked against committed golden fixtures. Coverage is validated against an independent provincial drug-benefit file the list was not built from; the matching heuristic is calibrated against ground-truth entity spans. Baselines for Moonshine~base, Whisper~base.en, and MedASR on two open benchmarks are scored with the released tool and reported with 95% confidence intervals from resampled per-utterance scores.

We develop Newton Matching, a unified framework for fine-tuning and sampling in generative modeling. The target is $\pi\propto\mu e^{\tau r}$, where $r$ is the reward, $\tau>0$ the inverse temperature, and $\mu$ denotes the pretrained model's terminal density for fine-tuning or the constant $1$ for sampling. We shift the paradigm from isolated losses to iterative optimization over canonical models: population minimizers of standard conditional matching for terminal densities. Under compatible smooth-realization assumptions, canonical velocities form a manifold diffeomorphic to the density manifold. Transporting the Fisher-Rao metric and mixture connection to this manifold, we show that the reverse-KL Hessian equals the metric, so the Newton direction coincides with the negative Fisher-Rao gradient. At terminal density $\rho$, each stage takes a tangential step generated by the regularized reward $r-\frac1\tau\log(\rho/\mu)$, followed by terminal-density-preserving canonicalization. This canonical retraction yields an exact finite-stepsize density characterization. For the ideal iteration, we prove strict reverse-KL descent away from the target for $0 < \eta \le \tau$, global convergence under mild conditions, and local quadratic convergence for full steps ($\eta=\tau$). Covariance and gradient forms, each with forward or reverse regression-pair constructions, yield sample-wise tangential-update losses with the same population minimizer, without importance sampling or full-trajectory backpropagation. We develop approximate updates and define critical-point consistency as vanishing tangential displacement if and only if $\rho=\pi$. We recover representative methods as exact realizations, critical-point-consistent approximations, or objective-altering variants, enabling modular algorithm design. Our work advances the theory and algorithms of reinforcement learning for generative models.

One way to increase the Energy Efficiency (EE) of 6G wireless networks is to utilize existing network infrastructure more efficiently. This can be achieved by introducing User-Centric Cell-Free Massive Multiple-Input-Multiple-Output (UCCF MMMIMO), which allows for simultaneously serving a single user by multiple Base Stations (BSs). From this perspective, the key challenge is to decide which BSs should serve a given user, known as the Serving Cluster Formulation (SCF). In this paper, we propose to deal with this problem by using an Artificial Intelligence (AI) agent based on a Large Language Model (LLM), targeting improvement of EE. We evaluated the proposed AI agent using a complex, 3D Ray Tracer-based, cellular network simulator, comparing a few GPT models and state-of-the-art algorithms. The results show up to 32% gain in EE of the proposed AI agent compared to the baseline.

Understanding whether language-model embeddings encode structured real-world information is important for both representation analysis and information retrieval. We study this question for temporal and geographic signals using a simple projection-based method that operates directly on output embeddings. Given a small set of seed examples, the method defines an axis in embedding space and ranks texts or entities by their projection onto that axis. Our approach is fully black-box and model-agnostic: it requires only embeddings, without access to model weights, internal activations, auxiliary probes, or additional training. This makes it applicable to modern embedding models available only through APIs and provides a lightweight way to analyze whether temporal and spatial dimensions are present in their representation spaces. We apply the method to temporal and geographic datasets and find that embedding projections recover meaningful chronological and spatial structure. These results provide evidence that output embeddings encode signals relevant to time and space, while also offering a practical tool for interpretability and for downstream temporal and geographic information retrieval tasks, such as temporal ordering, geographic ranking, and tagging.

We study state tomography when each measurement acts on at most $k$ fresh copies and no quantum memory is retained between blocks. We prove a lower bound matching the upper bound in [arXiv:2510.07788]. Thus the copy complexity of estimating an arbitrary $d$-dimensional state to trace distance $\epsilon$ is, up to absolute constant factors, $\max\{d^3/(\sqrt{k}\epsilon^2),d^2/\epsilon^2\}$ for every $k$ and all sufficiently small $\epsilon$. This removes the earlier restriction that $k$ be small as a function of the accuracy. The lower bound applies to arbitrary measurements within each block and adaptive choices between blocks. The lower bound already applies in a small neighborhood of any state whose smallest eigenvalue is of order $1/d$, even when the center is known. The main ingredient is a uniform Fisher information bound for one measurement block that depends only on the smallest eigenvalue of the state. The proof avoids the perturbative expansion responsible for the restriction in [arXiv:2402.16353]. Fano's inequality for metric balls and a log-Sobolev comparison between mutual and Fisher information then reduce the adaptive protocol to this block bound [arXiv:1607.00550, arXiv:1902.08582].

Aggregating heterogeneous vision-language models (VLMs) can improve multimodal reasoning, but neither an individual model's confidence nor that of the aggregated answer measures reliability at the system level. We present CUSP (Collective Uncertainty through Semantic Opinion Pooling), a training-free uncertainty quantification framework that maps multiple VLM responses to a shared semantic response space, pools them into a pooled semantic opinion, and reports two complementary system-level signals: collective uncertainty, the dispersion of the pooled opinion, and Jensen-Shannon divergence (JSD), the conflict among the model-level opinions. Within this pooled semantic opinion, the unnormalized collective entropy decomposes exactly into the mean of the models' individual semantic entropies and the JSD, separating total dispersion from model conflict. Requiring neither token logits nor calibration labels, CUSP applies to open-weight and commercial VLMs alike. In static multi-VLM ensembles, collective uncertainty is the strongest signal in the small-model regime (0.764 AUROC for prediction-error detection, 0.889 AUARC for abstention), outperforming uncertainty baselines majority voting and naive selection by 4.7 to 15.8 points and widening its margin as the ensemble grows; JSD is strongest in the evaluated commercial regime (0.819 AUROC, 0.910 AUARC) and ranks hard-answer model conflict with AUROC up to 0.982. The pooled prediction also improves accuracy over the average single model by 5.6 to 13.0 points. Over the full trajectory of a multi-step, multi-agent system, subagent collective uncertainty ranks system failures above chance (0.619 AUROC) and gives the best abstention ordering among the evaluated signals (0.699 AUARC).

In common-context question answering (CCQA) tasks, multiple input questions share a common context to base their answers from. However, Large Language Models typically generate each answer using an independent prompt. While existing batching and caching techniques help improve parallelism and reduce repeated computations, the separation of questions across prompts limits the achievable speedup, as modern GPUs are underutilized due to a memory bottleneck during attention. We present Intra-Prompt Parallel Decoding (IPPD), a novel inference method that answers multiple common-context questions in parallel within a single prompt. IPPD directly addresses the bottleneck by efficiently sharing both memory and computation during the attention process, as the next token for every question is decoded in a single inference step. IPPD uses virtual position IDs and attention mask manipulation to generate the same output as standard prompting without requiring fine-tuning or any changes to the LLM architecture. Since all parallelism occurs within a prompt, IPPD is fully compatible with batched inference, even when each prompt features a different context. Our experiments show that IPPD delivers up to 7X the effective throughput as standard decoding without quality degradation, and outperforms prefix caching with PagedAttention in most settings.

Platform moderation can disrupt online communities, but the responses that follow do not necessarily take the form of direct migration. Communities may relocate, reorganize, discuss moderation decisions, express censorship concerns, or continue activity on alternative platforms without clear evidence of user-level movement. This study examines Reddit community bans and quarantines and their relationship with activity and migration-related discourse on Voat. Using data from the Multi-Platform Aggregated Dataset of Online Communities (MADOC), we analyze interventions affecting FatPeopleHate, GreatAwakening, MillionDollarExtreme, and CringeAnarchy. We combine computational discourse screening with a stratified independent human-validation procedure and population-adjusted estimates across 90-day pre- and post-intervention windows. The results reveal heterogeneous cross-platform responses. FatPeopleHate experienced a substantial increase in overall Voat activity while the relative prevalence of policy-response and broader relevant discourse declined. CringeAnarchy showed declines in direct migration and policy-response discourse following its permanent ban, although some estimates were sensitive to uncertain human labels. GreatAwakening instead exhibited a modest increase in policy-response discourse without clear evidence of increased direct migration. These findings suggest that platform interventions do not produce a uniform migration process. Overall destination-platform activity and migration-related discourse may move in different directions, highlighting the importance of distinguishing behavioral migration from broader discussion of moderation, censorship, governance, and community rebuilding.

Quantum Machine Learning (QML) has shown rapid advances by utilizing quantum computing for machine learning tasks. Meanwhile, the privacy risks accompanying QML is also starting to be studied, which inherit privacy leakage channels from "classical" ML and also quantum-unique risks. Existing work on privacy-preserving QML largely focuses on a QML-as-a-service scenario, which generally assumes that the QML model owner provides only classical bit outputs to queries, while users (and adversaries) have only classical computing abilities. However, this view is increasingly challenged in a quantum-native world of quantum-capable users/adversaries, which may have access to both quantum computing abilities and access to quantum information output from service providers. In this paper, we aim to bridge this gap by examining membership inference attacks against QML models by demonstrating that increasing quantum access and quantum computing abilities provides provable theoretical privacy leakage and empirical adversarial gain. However, the probabilistic nature of QML introduces a gap between theoretical and empirical adversarial advantage. These results show that existing research on privacy leakage in QML models underestimates privacy leakage in emergent quantum-native access regimes, and we hope to establish a first step in examining potential privacy leakages for QML in the quantum-native world.

One of the biggest risks faced by Software Defined Networks (SDN) is the Distributed Denial of Service (DDoS) attack in which a compromised controller can make an entire network unusable. To address these challenges, we suggest an entropy-guided machine learning framework, called XAI-SDN, for real-time DDoS detection in SDN environments which is lightweight and explainable. The framework extends the flow features extracted by CICFlowMeter with eight Shannon entropy metrics obtained by an $\mathcal{O}(1)$ rolling algorithm and uses a Random Forest classifier with SHAP TreeExplainer for providing transparency at the prediction level. On a fixed temporal split, XAI-SDN achieves an accuracy of 99.9987%, a macro F1-score of 99.9621%, and an AUC-ROC of 1.0000 on the full 3.59 million flows of the CIC-DDoS2019 SYN benchmark. The pipeline sustains 0.0165~ms per flow (60{,}606 flows/s) without the use of SHAP and 0.5122~ms per flow (1{,}953 flows/s) with full support of SHAP under the 99.14% prevalence of DDoS traffic, which is a step towards achieving a balance between the detection performance and operational transparency in next-generation SDN security.

The paper introduces a neural network-based approach for analyzing ECG signals to estimate respiratory rate by leveraging the phe- nomenon of Respiratory Sinus Arrhythmia (RSA). Our method employs a deep learning model trained to predict respiratory waveforms directly from ECG input data. To achieve this, we developed and evaluated three different neural network architectures capable of automatically extract- ing relevant features from ECG signals without the need for manual preprocessing. The proposed approach offers a robust and scalable solu- tion for non-invasive respiratory monitoring, with potential applications in healthcare and wearable technology

Predicting olfactory qualities from molecular structure is an open problem in chemoinformatics. Although linear models can link molecular features to odor descriptors, they often fail when extrapolating to novel chemical scaffolds, extreme molecular weights, or complex odor mixtures. To address this, we introduce GraphNOSE, an open-source graph transformer framework that predicts multi-label odor descriptors from simplified molecular-input line-entry system (SMILES) strings for single molecules and binary mixtures. By integrating positional and structural encodings within a transformer-based graph architecture, GraphNOSE achieves strong performance with six times fewer parameters than standard graph neural network (GNN) baseline while consistently outperforming linear models, molecular language model embeddings, molecular fingerprints, and baseline GNNs by an average area under the ROC curve (AUROC) margin of 4.52% (p < 0.01). GraphNOSE achieves an AUROC of 84% on out-of-distribution compounds (OODs). This exceeds the current state-of-the-art GNN for OOD in olfaction (Open-POM: 81%, p < 0.001), and identifies conditions under which linear models empirically fail. Finally, we apply XAI (explainable AI) methods to identify which substructures and molecular features drive odor predictions, yielding insights consistent with chemical intuition and grounded in the model's learned representations. Together, these results establish GraphNOSE as a scalable and interpretable architecture for olfactory prediction that generalizes to structurally distinct compounds underrepresented in current perceptual databases.

Requirements engineers for agentic-AI domains face challenges in evaluating, specifying, and operationalizing safe autonomy. Mainstream frameworks, such as Goal-Oriented Requirements Engineering (GORE), lack mechanisms to systematically address these challenges under epistemic uncertainty. We contribute an approach that builds on GORE to model and simulate safe autonomy in agentic-AI systems. We introduce a novel Regret-Dominance Mechanism (MS-RGR) to operationalize safe autonomy. MS-RGR uses two signals: epistemic surprise (novelty detection) and cognitive regret (evaluative risk) to address the trilemma problem: should the agent operate in routine autonomy, undergo reflective reasoning, or escalate to human? We instantiate MS-RGR in elderly care monitoring and autonomous driving. A 100-seed stochastic simulation shows MS-RGR reduces silent failures to near-zero and detects risk approximately 17.5 times faster than a sensor-only baseline, remaining formally traceable via LTL safety properties. A retrospective proxy instantiation applying the DRI gate post-hoc over execution traces from 208 AGENTHARM scenarios across seven LLMs shows the gate improves harmful-task refusal only for models with strong baseline safety (over 80% pre-gate refusal, e.g., 84.1% to 90.9%), indicating MS-RGR amplifies rather than substitutes for model-level safety training. We discuss threats to validity, positioning MS-RGR as initial feasibility evidence for design-time safety constraints in agentic-AI requirements engineering.

Finite model finders cannot witness an Austin law: an identity whose finite models are all trivial but which has a nontrivial infinite model. We introduce rank-decreasing sparse trace-tree magmas, finitely presented total operations on a countably infinite constructor-tree carrier. The default product pairs its arguments; finitely many positive Horn clauses define exceptions. Our main procedure derives clauses from symbolic evaluation traces. For every model found, it proves functionality of the exceptional relation by descent on constructor size, proves the identity by exhaustive symbolic case analysis, and emits a self-contained Lean 4 certificate. A least simultaneous fixed point gives an implementation-independent semantics, so bounded search may miss models but cannot invalidate certified results. On ETP's 96 order-five Austin candidates, we discover and Lean-verify infinite countermodels for 28 identities with no prior public classification in our audit. They form 14 duality classes and establish 28 new Austin classifications. Four ALPS-known cases bring the total to 32 certified candidates. On Canonical-4187, the deduplicated union of Order5-130 and the 4,141-row ALPS pool, a fresh trace run produces 636 certificates, all accepted by Judge v3. At equal resource limits, Vampire 5.0.1, E 3.5.1, and complete Twee 2.6.1 jointly prove implications in 94 canonical classes. Only Twee returns trusted counter-satisfiable outcomes, for 18 classes; independent finite-side certificates force 16 to be infinite. None of these ATPs emits an explicit model or Lean certificate, and none decides the 28 new classifications. To the best of our audit, this is the first automated system to synthesize this trace-tree model family, generate well-founded inversion proofs, and emit self-contained Lean 4 certificates.

We study variance-preserving diffusion of the response in mixed linear regression (MLR) with unknown mixing weights. Our analysis separates the statistical guarantees of score matching from the loss geometry and optimization signal at a fixed diffusion noise level. The KL divergence links the denoising score matching objective integrated over the diffusion path with the likelihood and a terminal discrepancy. Under mild regularity conditions and terminal schedule, the resulting estimator converges up to the ground truth parameters of MLR, and its scaled error converges to the Gaussian limit of the maximum-likelihood estimator. At a fixed scale of the diffusion noise level, we derive a decomposition linking the score matching loss to cross-entropy and Expectation-Maximization (EM) operators. This decomposition yields an EM-related low-noise gradient expansion with additional correction terms of latent variance. In the high-noise limit, we further characterize gradient descent on this limiting loss under isotropic covariance. Along fixed high signal-to-noise ratio rays, the score matching imbalance gradient and the latent-variance term tend to zero pointwise. Numerical experiments illustrate our theoretical findings and statistical guarantees.

In the digital online ecosystem, we are surrounded by distinct forms of information pollution, posing significant threats to both individuals and society. Fake news, for instance, wields power to sway public opinion on matters of politics and finance. Deceptive reviews can either bolster or tarnish the reputation of businesses, while unverified medical advice may steer people toward harmful health practices. In light of this challenging landscape, it has become imperative to ensure that users have access to both topically relevant and factually accurate information that does not warp their perception of reality, and there has been a surge of interest in various strategies to combat misinformation through different contexts and multiple tasks. The purpose of the ROMCIR Workshop, for some years now, is precisely that of engaging the Information Retrieval community to explore potential solutions that extend beyond conventional misinformation detection approaches. Key objectives include identifying subjective and objective factors associated with information credibility and truthfulness, respectively, and integrating such factors as fundamental dimensions of relevance within IR Systems (IRSs), achieving early detection of misinformation, and ensuring that the search results retrieved are not only truthful but also explainable to the users of IRSs. Moreover, it is essential to evaluate the role of generative models such as Large Language Models (LLMs) in inadvertently amplifying misinformation problems, and how they can be used to support IRSs, together with the contribution that the human-in-the-loop paradigm can have in this context.

Opioid use disorder (OUD) remains a public health crisis in the United States, yet it is difficult to identify from electronic health records (EHRs) because missing diagnosis codes and supporting evidence are buried in clinical narratives. Accurate OUD identification is critical to support interventions and improve health outcomes. This study developed a rubric-guided large language model (LLM) that incorporated Optimization by PROmpting (OPRO) for OUD computable phenotyping (CP). The framework used an 18-item, expert-identified rubric to instruct LLMs to automatically extract critical text with supporting evidence to determine OUD flags. Two UF Health physicians (GMR and WMG) chart-reviewed 253 patients, including 68 OUD-positive cases. Our LLM-based computable phenotype (CP) achieved the best F1 score of 0.774 and an AUROC of 0.934, outperforming the machine learning-based CP using EHR and natural language processing-extracted variables, and zero-shot LLMs by relative F1 improvements of 12.8% and 44.4%, respectively. The proposed LLM-based CP could link LLM-extracted evidence to OUD phenotyping for better explainability.

We (the author and the AI systems that did the heavy lifting) show that the quantum query complexity of minimizing a convex function over a convex subset of $\mathbb{R}^n$ with evaluation and membership queries is $\tilde\Omega(n)$, nearly matching the best known upper bound. In particular, we show this even for quadratic minimization, which is equivalent to inverting an $n \times n$ real matrix using matrix-vector queries. We also show linear or nearly linear lower bounds on the quantum query complexity of computing the trace, the sign of the determinant, and the magnitude of the determinant of a real matrix in the matrix-vector query model. We use a novel quantum lower bound technique, the determinantal witness method, based on identifying a witness whose Fourier transform vanishes on low-rank matrices and that correlates well with the function being computed.

Lifelong LLM agents increasingly rely on external skill artifacts as one element for preserving and reusing capabilities over time. These skills (usually portable Markdown files such as SKILL.md) describe when and how to apply a capability and must be corrected, expanded, and consolidated as tools and usage patterns shift over deployment. Recent work seeks to automate skill curation, but it largely evaluates against automated baselines and treats human maintenance as an unmeasured bottleneck. We study that missing process directly. We mine the full commit histories of five public AI-skill repositories, a purposive sample of AI-tooling organizations, covering 873 commits, 143 skill files, and 254 substantive post-creation edits from October 2025 to June 2026. We code each edit with pre-registered governance, operation, and trigger-evidence codebooks. Three findings emerge. First, every substantive edit is authored or merged through a named human account, while 62% carry an AI co-author trailer, with large repository-level variation. Second, these edits are genuine curation: an audited sample shows that most change skill content, and the coded operations are dominated by additions and corrections. Third, a pre-registered rule-likeness axis fails its reliability gate; reliably coding rule-likeness from commit artifacts remains an open measurement problem. We release the corpus, codebooks, mining scripts, and a replay protocol for automated skill curators. For self-evolving agents, public skill maintenance currently looks less like an autonomous pipeline than a human-governed, AI-assisted loop that future curators must measure against and operate within.

Language models adapted on private text are often served through APIs, so privacy leakage occurs through generated outputs rather than exposed weights. Private prediction protects these releases. Methods such as PMixED incur privacy cost at each release and increasingly rely on the public model over long horizons. PAC privacy instead calibrates noise to output variability across possible secrets, adding less noise when predictions are stable. To our knowledge, PAC-private prediction has not previously been extended from classification to autoregressive generation. We construct $m=128$ overlapping worlds from the private corpus, with each record appearing in exactly $m/2$ worlds, and train one adapter per world over a frozen public model. The realized world is the secret. At each token, the public model defines a candidate set, the worlds vote, and their posterior-weighted disagreement determines the PAC noise; unanimity requires no calibration noise. We prove $I(S;Y_{1:T}) \leq I(S;H_T) \leq bT$. Our contributions are extending PAC privacy to autoregressive generation, handling adaptive self-generated contexts, and introducing coupled decoding that preserves privacy accounting while avoiding greedy degeneration. On WikiText-103 with GPT-2-small, we retain 74% of the fine-tuning gain at a per-token budget of $2^{-32}$, while membership-inference success is bounded by 51.08% after $10^6$ tokens; posterior-entropy estimates of leakage are roughly 17% of the charged budget. Inference privacy is not content protection: even when membership advantage on a memorized canary is indistinguishable from zero, the canary is emitted at the same rate. Against PMixED under matched membership-inference bounds on the same data universe and test set, we retain 98% of non-private headroom from $10^2$ to $10^6$ tokens, versus at most 56%, with no crossover.

Recent advances in artificial intelligence (AI), particularly large language models (LLMs), are transforming how we design and build systems by increasing access to domain knowledge and by providing automation support to software engineering (SE). As implementation becomes less expensive through generalist SE agents, engineering effort shifts away from writing correct code and toward expressing, curating, verifying, and evaluating requirements. In this paper, we survey the state of the art in AI for requirements engineering (RE) research leading up to the transformation, before reviewing advances in LLMs. We survey two subsequent research areas: prompt programming, which treats LLM instructions as a program in SE vernacular, and generalist SE agents, which combine multiple LLM advances to yield semi-autonomous processes that complete SE tasks. Finally, we explore the future of requirements engineering along two axes: matters changing how we interact with requirements through the SE process, and matters changing how requirements are experienced by software developers and stakeholders more broadly, including end-users. This article aims to inform how RE researchers can navigate this transformation in the selection of future research priorities.

We present a continuous, population-scale measurement record of autonomous language-model trading agents operating in production across two systems with one design lineage: DX Terminal Pro (3,505 user-funded vaults trading real ETH in Base memecoin markets for 21 days, February to March 2026) and the DXAP live alpha fleet (500 to 599 user-created agents all-history, 91 to 117 concurrently active, trading Hyperliquid perpetuals, June to August 2026). The record spans roughly six months, 7.5M single-model invocations with about 300K onchain actions, and a further 231,638 multi-tool turns producing 14,596 fills. Four findings carry the paper. First, the operating layer determines behavior more than anything written in strategy text: a risk slider explains leverage (+0.425 per level), agent fixed effects absorb 60% of variance, and a leaderboard render boundary causally routes selection (regression discontinuity 1.75x at the top-3 cut). Second, sizing is volatility-blind: median leverage is 5.0x in every volatility sextile, and one posture-slider cell (11% of the book) holds 62% of liquidations. Third, agents capture almost none of the upside they reach: 43.2% of positions saw at least +300 bps of favorable excursion within 24h, yet 49.3% of those closed with a negative trade return; a mechanical bracket recovers +39.0 bps per position. Fourth, neither fleet shows a directional edge. The DXAP fleet is not profitable and trails a matched Hyperliquid retail benchmark (41% vs. 50% roundtrip win rate). A paired-replay league of frontier models on 416 captured production scenarios finds decision quality statistically indistinguishable at this horizon, while choice stability differs sharply across model families. Every headline survives day-clustered inference, permutation nulls, and a common-fee restatement; the paper closes with a 17-rule methodology canon bought with our own retractions.

Full-page end-to-end Optical Music Recognition seeks to transcribe entire music pages directly into symbolic notation, avoiding the limitations of traditional pipelines that rely on accurate staff segmentation. Recent Transformer-based architectures have achieved strong performance on typeset scores, relying on large-scale synthetic data for pretraining. However, their applicability to handwritten music remains largely unexplored. In this work, we study full-page transcription on handwritten monophonic collections and analyze the impact of synthetic pretraining in this setting. To investigate which factors are most relevant during pretraining, we introduce a generator capable of producing visually coherent full-page scores in both typeset and handwritten styles. Experiments on three real handwritten datasets provide a comparative evaluation of several full-page pipelines and different synthetic pretraining strategies. The results suggest that the benefits of synthetic pretraining are primarily associated with learning structural layout conventions rather than with visual similarity to the target handwriting.

We present Srijika, a system for producing installable OpenType fonts for nine Brahmic scripts: Devanagari, Tamil, Bengali, Telugu, Kannada, Malayalam, Gujarati, Gurmukhi, and Odia. Rather than generating fonts from scratch, Srijika restyles glyph outlines from shaping-complete template fonts. It preserves the template's cmap and GSUB closure and its GPOS data under a documented metric policy, making every output a complete font by construction. This addresses a central challenge of Indic font generation: hundreds to thousands of conjuncts, half forms, and matra variants must remain mutually consistent under OpenType shaping. Srijika produces 66 TTFs: 57 curated presets and nine open-vocabulary showcase fonts. All pass the OpenType Sanitizer, while HarfBuzz and CoreText reproduce the template glyph-ID sequences on conjunct-heavy probes. A full-closure audit covering 80,915 glyphs and 54,812 anchors quantifies metric changes. Natural-language style selection uses Lipika, a retrieval index over approximately 650 open-license font families. A reference-conditioned latent diffusion model redraws template glyphs in the selected style, followed by content gating, harmonization, and shaped-cluster verification with fallback to template outlines. We evaluate against no-learning baselines. On diffusion-training-family-held-out SSIM gates, template copying outperforms generation on 50 of 56 faces. Style movement is measurable only with an internal same-model embedding whose training corpus includes the held-out families, so these results require caution. A learned baseline, independent style metric, and human study are outside this report's scope. Our contributions are the layout-reusing formulation and pipeline, its nine-script audit and benchmark, and a negative-results catalogue covering failed conditioning, objective choices, and data-hull limits of reference-guided restyling.

Large language models (LLMs) are increasingly being explored for automating SystemVerilog Assertion (SVA) generation, yet most evaluations report correctness on a single syntactic representation of an input. Such point accuracy does not reveal whether a model's correct output is stable when the same RTL behavior is written differently. This paper presents a controlled metamorphic evaluation of LLM-based SVA generation under semantics-preserving RTL transformations. Starting from the VERT dataset, we construct a quality-filtered conditional-control pool and a stratified 40-program evaluation set containing 295 assignment behaviors. We evaluate two open code models, Qwen2.5-Coder-7B and DeepSeek-Coder-V2-Lite, with an identical evaluation prompt and greedy decoding. Three transformations are studied: operand reordering, deterministic identifier renaming, and redundant parenthesization. Beyond baseline and transformed accuracy, we measure conditional robustness, invariance failure, and any-flip rate, with 10,000-sample clustered bootstrap intervals at the RTL-program level. Across all six model-transformation conditions, 9.7%-27.0% of behaviors that were correct on the original RTL become incorrect after a semantics-preserving transformation. Aggregate accuracy can therefore hide substantial instability: under identifier renaming, DeepSeek-Coder-V2-Lite improves from 53.9% to 63.7% accuracy while 19.5% of its originally correct behaviors fail. Manual review of 30 sampled correct-to-wrong transitions identifies dropped path predicates, branch-polarity errors, Boolean-structure corruption, and output-contract violations. The results show that point accuracy alone is insufficient for characterizing LLM reliability in assertion generation and motivate robustness-aware evaluation for AI-assisted hardware verification.

Thermoacoustic instability in premixed flames results from the coupling between the flame's heat release, as determined by combustion parameters, and surrounding acoustics, as determined by combustor geometry. A primary instability results in flame flattening as intrinsic flame instability modes are stabilised. Secondary thermoacoustic instability results in a parametric flame instability and drastic growth of acoustic amplitudes. Due to their relative expense, numerical simulations of these phenomena remain scarce. In this work, Direct Numerical Simulations (DNS) of thermoacoustically unstable idealised premixed flames in a tube with acoustically closed upstream and open downstream ends are presented. Results herein demonstrate nonlinear saturation of the primary instability as the flame flattens as well as an oscillating flame fingering characteristic of the unsteady Rayleigh-Taylor effect. To reduce computational cost, we perform DNS only on the region surrounding the flame. Acoustics at in- and outflows are described using the Navier-Stokes Characteristic Boundary Condition (NSCBC) method to model their delayed reentry into the domain in a formulation referred to as the Acoustic Delay Characteristic Boundary Condition (ADCBC) method. A new Averaged Proportional and Integral Linear Relaxation (APILR) method is also introduced, which modifies the Classic Linear Relaxation (CLR) method to maintain time-averaged values of inflow velocity and outflow pressure. Here, an integral control term is used to remove non-zero equilibrium time-averaged inflow velocities which impinge control over flame position. Both new methods demonstrate their capability in inert and counterflow flames test cases. These methods enable the numerical simulation of combustion instabilities at significantly reduced computational expense.

We propose a reinforcement learning framework in which exploration is driven by intrinsic curiosity, designed for scenarios where environments are non-stationary and rewards are sparse, delayed, uninformative, or absent. In our model, action selection is guided by a combination of external rewards and an epistemic motivation mechanism that biases the agent toward structured exploratory directions. The central hypothesis is that effective exploration emerges at intermediate levels of incoherence, while performance degrades under both overly rigid and overly disordered dynamics. To test this idea, we implement the framework on top of a Liquid State Machine (LSM) substrate and evaluate it on two standard benchmarks: the discrete-action LunarLanderv2 and the continuous-control BipedalWalkerv3. The proposed method achieves competitive performance on both tasks relative to established deep RL algorithms, including Proximal Policy Optimization (PPO) and Intrinsic Curiosity Module (ICM). We further show that the curiosity window is not recovered in Active Inference agents under the same analysis, suggesting that the proposed dynamics capture a distinct exploration regime