Science & Mathematics

Cubic boron arsenide (c-BAs) is a promising ultrahigh thermal conductivity semiconductor, but its direct synthesis on lattice-mismatched substrates remains challenging. Here, we demonstrate the synthesis of nanocrystalline c-BAs thin films on GaAs(100) substrates using gas source molecular beam epitaxy. X-ray photoelectron spectroscopy reveals an optimal growth temperature window of 250-400 C to maintain near stoichiometry and suppress severe arsenic loss. Surface and structural characterizations via atomic force microscopy, reflection high-energy electron diffraction, and grazing-incidence X-ray diffraction demonstrate a Volmer-Weber 3D island growth mode and a preferential crystallographic alignment with the GaAs substrate. High-resolution transmission electron microscopy confirms a chemically sharp interface without gallium interdiffusion into the c-BAs layer. These findings establish an initial step for the integration of c-BAs into advanced thermal and optoelectronic platforms.

Electrophysiological source imaging is severely underdetermined because the lead-field matrix has limited rank relative to the source dimension and strongly correlated columns. Scalp EEG measurements constrain only the observable row-space component, leaving a high-dimensional null space unobservable. We propose a two-stage framework coupling posterior diffusion refinement with deterministic null-space completion. A minimum-norm-initialized, singular-coordinate Posterior-DDIM sampler combines a spatiotemporal diffusion prior with direction-dependent measurement reliability to refine the observable component. A frozen ridge predictor and nonlinear temporal residual network then estimate the null component under explicit projection constraints, preserving the refined row estimate and its predicted EEG exactly. All learned components are trained exclusively on simulated cortical activity and its row-null decomposition, without paired EEG-source training data. Across simulations with varying noise levels, the method achieved higher frame-level area under the receiver operating characteristic curve and average precision, together with lower localization error, than conventional and learning-based baselines. Exploratory evaluations on evoked EEG and simultaneous scalp EEG-stereo-EEG recordings showed condition- and event-dependent concordance with cross-modal and intracranial references.

Reconstructing a radio map (RM) from sparse received-power measurements supports non-cooperative interference monitoring and spectrum management. However,sparse measurements leave uncertainty about both the global RM and the interfering emitter location, making informative sampling and accurate reconstruction challenging. To address these challenges, a novel joint sampling and reconstruction method named RMSampler is proposed in this paper to improve measurement selection and RM reconstruction under limited measurement budgets. For measurement selection, RMSampler evaluates global RM error reduction using the covariance of diffusion-generated RMs and localization error reduction using candidate emitter locations. Because these errors have different units, their expected relative reductions are combined into a scale-invariant sampling criterion, with each acquired measurement updating both uncertainty estimates. After sampling, the updated emitter-location probabilities guide a separate conditional diffusion ensemble for final reconstruction. To improve measurement agreement, an effective sample size (ESS) target guides robust posterior correction, followed by a minimum-distortion projection enforcing measurement and building constraints. For the reconstruction ensemble, analysis establishes that increasing likelihood power reduces expected robust measurement loss, while pretrained neural parameters and architecture remain unchanged. Experiments using simulated and real-world measured data demonstrate improved reconstruction under matched measurement budgets. In the Chongqing real-world measured data experiment, root mean square error (RMSE) decreases from 3.54 to 2.73~dB relative to random sampling under a common reconstruction method.

Agent platforms are starting to improve their own agents: they read a deployed agent's execution traces, derive a change, have an automated verifier approve it, and release it with no human author. How much of this loop can safely run without a human? We propose two quantities a platform can measure. Verifier strength $σ$ is how well the available verifier -- often an LLM judge -- agrees with a trusted reference verifier; it limits how much a step can improve the agent. Recoverability is how cheaply a bad change can be undone: whether undoing it needs a new verifier approval, and how much of the old behavior a rollback restores. A step should run unattended only if both are adequate. On our platform this rule allows prompt, tool and skill edits to run unattended and keeps weight updates under human sign-off. Our main contribution is these two definitions and how to measure them. We built a platform that applies the rule and tested it in two ways: on the platform's own approval gate, and in controlled experiments on the $τ^2$-bench agent benchmark. On verifier strength, the platform's LLM judge sometimes gave the same score no matter what it was shown (9 of 37 settings) and reported no error, so a verifier can look healthy while approving anything; its strength must be measured, and two simple tests catch the failure. On recoverability, restoring a saved copy brought the agent back to its original behavior, but a regenerated undo did not, so only changes whose old version is kept can be reversed without a human; and while a change was deployed, it altered the outcome of 17 percent of tasks beyond normal run-to-run variation, which no undo repairs. Together these support the rule: a change can run unattended only when its old version is kept and its verifier's strength has been measured.

On-policy learning is central to post-training: training on a model's own generations limits distribution shift and helps preserve existing capabilities. On hard problems, however, on-policy RL provides no reward-based learning signal when all attempts fail. On-policy distillation provides next-token supervision from a teacher, but on hard problems, if a rollout goes wrong early, subsequent supervision is conditioned on prefixes that already contain the mistake. Correct behavior can instead come from a gold or teacher solution, but such off-policy targets can differ substantially from the model's generations and, as we show, may erode existing capabilities. We therefore treat on-policyness as a continuum and propose projected distillation, which minimally edits the student's failed attempts into correct targets. The method uses contrastive updates to favor corrections over failed attempts and refreshes these corrections as the student evolves, enabling near-on-policy learning even when all student attempts fail. Theoretically, we prove that the student's learning regret grows with the edit size: it scales linearly with the horizon under minimal correction, as in DAgger, and quadratically under full-answer replacement, as in behavior cloning. Empirically, across two students (Qwen3-8B and OLMo-2-7B) and five domains spanning math, code, science, and tool use, projected distillation gives the highest area under the specialization-retention frontier (best on 8 of 9 cells against GRPO and OPD, and 7 of 9 against SPOT and SODA). Our ablations highlight the importance of online corrections, the contrastive loss, and on-policyness: moving targets farther off-policy steadily worsens the specialization-retention trade-off, with retention falling from 0.916 to 0.744 when minimal corrections are replaced by student-agnostic gold solutions.

Adapting a forecasting system to a changing world classically means retraining. We ask whether the task of continually learning regimes for time series forecasting can be represented by a piece of text: a guidance block, evolved by outcome-driven reflection and used in a frozen LLM forecaster's prompt. On a refreshed multimodal benchmark (weekly search indices, daily currencies, daily commodities, each window paired with web-retrieved news), we run GEPA-style prompt evolution under strict contamination control to prevent the data leakage common in timeseries benchmarks: the forecaster's knowledge cutoff predates every evaluated window. A 900-word evolved guidance carried by Llama-3.3-70B beats TimeSeries Foundation Models (TSFMs) on held-out data (GM MASE 0.776 vs. 0.898 for TabPFN-TS-3), transfers to more recent models without measurable loss, and costs dollars rather than GPU-days to update. We characterize the loop: out-of-loop gains concentrate in the first reflection, in-loop improvements overstate out-of-loop gains 2-3x unless an untouched monitor arbitrates, and error-only objectives silently suppress the text modality. We provide two extensions. First, compiling the guidance into executable code with a TSFM as a callable tool recovers most of the gain deterministically. Second, ablations show the advantage widens at short horizons and that removing the news text does not hurt the guided forecaster, consistent with the text's value being distilled into policy during learning. We argue that continual prompt evolution is a practical, training-free alternative to retraining for multimodal forecasting.

Magnetic fields are widely used to brighten spin-forbidden dark excitons, but the inverse effect on an optically bright exciton has remained largely unexplored. We demonstrate the complementary phenomenon of magnetic-field-induced darkening of a bright exciton in Mo$_{0.68}$W$_{0.32}$Se$_2$. At 4.2 K, an in-plane magnetic field up to 25 T activates a dark-exciton line 14.2$\pm$1.3 meV below the bright exciton while suppressing the bright-exciton emission by about 85\%. A bright-dark mixing model quantitatively captures the coupled energy and intensity evolution with an effective in-plane $g$ factor of $|g_{\parallel}|=5.2\pm0.2$. Alloying is crucial to the effect, as it brings the system close to the bright-dark crossover, where a small bright-dark splitting is combined with unusually strong in-plane magnetic coupling. This places the system in a regime where field-induced level repulsion raises the bright-like branch and strongly depletes its thermal population. These results identify a general route to magnetic control of exciton populations through engineering of the bright-dark hierarchy.

Federated learning (FL) is a distributed machine learning paradigm that reduces transmission costs, mitigates cross-border data sharing restrictions, and enhances privacy by keeping raw data on local devices. Clients in FL systems train local models and share only updates, which are aggregated by a central server. While this setup limits direct data exposure, FL remains vulnerable to privacy attacks. Existing attack and defense strategies have largely been developed for centralized systems, and adapting them to FL introduces new challenges due to its unique architecture and training workflow. In this work, we first propose a model inversion attack on FL systems using Deep Convolutional Generative Adversarial Networks (DCGANs) in the context of large-scale medical imaging. Our method operates under realistic assumptions about the attacker's knowledge and capabilities. We then examine the effect of client-side differential privacy (DP) in mitigating privacy leakage and incorporate personalized FL to improve model performance under DP settings. Experiments on the Brain MRI ND-5 dataset show that DCGAN-based attacks can reconstruct semantically meaningful synthetic images, revealing significant risks to the privacy of training data. Applying DP reduces attack success but introduces an accuracy trade-off, which is further overcome by personalized FL at each client side. We also evaluate how different aggregation algorithms affect both attack vulnerability and defense effectiveness, underscoring the critical role of aggregation strategies. The proposed attack-and-defense framework is model-agnostic and applicable to a wide range of FL-based systems.

Mobile legged robots offer unique advantages for on-site construction, including the ability to traverse unstructured terrain, climb stairs, and operate independently between spatially separated work zones. This paper presents a proof-of-concept system for autonomous brick stacking using a Boston Dynamics Spot quadrupedal robot equipped with a manipulator arm. The system operates between a supply area and an assembly area, using AprilTag fiducials for localization and autowalk missions for navigation. Target brick positions and orientations are specified in CAD software and exported directly for robotic execution. We evaluate placement accuracy through ten repeated assemblies of a curved four-layer wall comprising 14 wooden bricks. Results show mean translation errors below 10 mm and mean yaw errors below 1.5 degrees, while the overall RMSE across all runs is 27.72 mm (translation) and 2.06 degrees (rotation). The system was also tested outdoors with stairs between the two areas. The workflow is modular and easily extendable to different wall geometries.

Text-to-image models can generate visual concepts that their prompts never mention, and recent work shows that such associations can be implanted so that ordinary prompts trigger them. Auditing this behavior requires recovering the description that a prompt omits. Existing prompt optimizers are not well suited to this task. Gradient-based methods require backward access to the scoring encoder. Discrete evolutionary search treats semantically unrelated tokens as equally close. Language-model-guided search requires one inference call per candidate, which limits exploration. This paper presents EmbGA, an embedding-space genetic algorithm that recovers concepts present in a set of images but absent from the prompt that generated them. EmbGA keeps the population structure of genetic search but applies its operators in the token embedding space. Crossover interpolates between parent embeddings, and mutation moves an embedding toward a randomly drawn vocabulary embedding. Each offspring is then projected back to discrete tokens before scoring, so every evaluated candidate is a valid token sequence. We evaluate EmbGA on two benchmarks against six optimizers under two exploration budgets. EmbGA achieves the highest alignment when the exploration budget is large, recovers the most grounded tokens on natural images, and runs much faster than language-model-guided search.

The represented gradient flow of a parameterized model depends on a mobility that can retain hidden parameter information. We study when one entire represented mobility map works for every autonomous loss from one fixed initialization. For two-factor matrix representations wider than both output dimensions, with individually full-row-rank factors, we classify all admissible joint ranks: closure occurs precisely at effective scalar-balanced minimal joint rank or a two-value signed spectral family at full joint rank. All intermediate ranks are excluded; positive represented flows belong to known square-root families. For general representations with constant ambient representation and gradient-Lie ranks, ordinary closure is equivalent to a local invariant-level metric in the analytic case and specified smooth regimes. An endpoint section forces the initially unrestricted mobility map to admit a locally smooth representative on the represented image. We also separate testing families: linear losses locally admit closure at every submersion, whereas one fixed translated strongly convex quadratic family fails it on a delayed-bracket example. A fixed output readout gives an ordinary squared-loss witness. These supporting results do not characterize quadratic-only closure in general. We retain the unresolved smooth output-rank-two case with Lie rank at least four. A frozen deterministic experiment checks a separate engineered-loss construction; it is distinguished from the squared-loss proof, universal theorems and empirical learning performance.

Open-vocabulary object detection aims to detect both seen and unseen categories by leveraging the semantic knowledge of vision-language models. However, existing methods often struggle with small objects due to insufficient preservation of high-resolution details, ineffective information aggregation across feature levels, and limited region-level discriminability. To address these challenges, we propose FocusOVD, an open-vocabulary object detector that enhances small-object representations without modifying the frozen VLM backbone. FocusOVD introduces three complementary components: a Reweighted Cross-scale Feature Fusion module, which learns normalized level-wise weights to adaptively balance shallow appearance details and deep semantic cues during feature pyramid aggregation; a Residual Pinwheel-shaped Gated Convolution module, which strengthens local geometric responses in high-resolution features; and a Residual Convolutional Gated Linear Unit module, which suppresses background-dominated activations and enhances RoI discriminability before prediction. Extensive experiments on COCO-OVD and LVIS-OVD demonstrate the effectiveness of FocusOVD. On COCO-OVD, FocusOVD achieves a novel-category AP of 36.3%, outperforming CORA, UniDetector, and SAMP by 1.3%, 1.1%, and 1.5%, respectively; on LVIS-OVD, it achieves an overall AP of 33.7%, surpassing the strongest competing method by 0.6%. These results demonstrate the strong generalization ability of FocusOVD for open-vocabulary small-object detection.

Language-model agents increasingly participate in operational workflows where approvals, binds, and payments create commitments. An approval valid when issued may no longer authorize execution after relevant facts, evidence, policy, or authority change. We call this the evaluation--execution gap. We specify a state-bound runtime assurance layer that separates probabilistic measurement from deterministic adjudication, binds an execution allowance to the decision-relevant basis of an approval, invalidates that allowance after relevant logged changes, and rechecks admissibility at dispatch. This distinguishes the continued validity of a prior approval from a current-state policy check. We describe eight failure modes, an assurance model, a reference architecture, and prospective interventions with retrospective remediation. In a scripted 42-case replay, the state-bound configuration identifies 15 later-invalidated approvals that a deliberately non-re-evaluating point-in-time comparator does not revisit; it holds 14 affected actions before dispatch and compensates one landed effect. These results illustrate behavior under the stipulated policies and event orders, not superiority over dispatch-time enforcement designs or production effectiveness. Under explicit interception, ingestion, dependency, and serialization assumptions, the model establishes dispatch-time validity within the registered execution surface.

A semilinear set is stratifiable if it is a finite union of linear sets whose periods have at most two nonzero coordinates, with no two periods in one component connecting alternating pairs of coordinates. We give a computable many-one reduction from the existence of a rational zero of an integer polynomial to deciding stratifiability. After the classical reduction to quadratic systems, we reduce rational solvability of these systems to a problem we call finite separation by Cartesian products of rational linear forms. Its input consists of rational subspaces $D_1,\ldots,D_k\le\mathbb Q^r\times\mathbb Q^r$. A separator is a map $(\boldsymbol x,\boldsymbol y)\mapsto(u(\boldsymbol x),v(\boldsymbol y))$, where $u,v\colon\mathbb Q^r\to\mathbb Q$ are nonzero rational linear forms. The problem asks whether finitely many separators suffice so that every pair outside $\bigcup_iD_i$ receives, from some separator, a value attained at no pair in that union. We then reduce this problem to stratifiability. Ginsburg and Spanier's characterization of letter-bounded context-free languages} also induces a reduction from rational polynomial solvability to context-freeness of letter-bounded languages specified by semilinear exponent sets. An OpenAI preprint claims undecidability of rational polynomial solvability; if correct, our result settles the decidability of general stratifiability, a problem open since 1966.

The growing use of autonomous artificial intelligence agents in digital commerce creates new security concerns involving agent impersonation, unauthorized payments, compromised agents, and misuse of delegated authority. This research introduces the Autonomous Commerce Agent Guard System (ACAGS), a cybersecurity framework combining supervised machine learning, behavioral analytics, agent-specific authorization controls, and explainable AI. The framework was evaluated using synthetic financial transaction data augmented with simulated agent characteristics and five adversarial scenarios. XGBoost was selected following comparisons with alternative classifiers. Across ten experimental runs, behavioral features improved mean F1-score from 52.12% to 67.49%, with agent-specific features increasing performance to 69.46%. The framework achieved detection rates of 99.77% for abnormal agent behavior, 99.73% for impersonation, and 100% for malicious delegation and authorization-scope violations. Compromised-agent detection reached 72.27%. ACAGS also supported APPROVE, REVIEW, and REJECT authorization outcomes, with SHAP explanations providing model interpretability. The results demonstrate the potential of combining behavioral intelligence and delegated authorization controls to strengthen simulated agentic-commerce payment security, while highlighting the need for validation using real-world transactions.