Papers

Open electromagnetic-transient datasets are beginning to make reproducible learning-based protection studies possible, but the practical use of these datasets still requires application-level benchmarks that define timing, sensing, and validation assumptions. This paper presents an initial application benchmark on the recently released PROTECT-90 dataset for two protection-oriented tasks: fault-type classification and discrete faulted-line identification. A compact one-dimensional convolutional neural network (CNN) is evaluated using post-inception windows of 0.25, 0.5, 1, and 2 cycles under strict episode-wise splitting. A non-convolutional multilayer perceptron (MLP) is also trained as an architecture-control baseline. The results show that both tasks are nearly saturated under full observability, with one-cycle test accuracies of 99.84% for fault type and 100.00% for line identification. The main performance variation appears under reduced observability: current-only inputs preserve line identification accuracy at 100.00%, whereas voltage-only inputs reduce line identification accuracy to 53.09% with the CNN and 50.57% with the MLP. This indicates that the limiting factor is measurement information rather than neural architecture. Additional stratified checks show stable performance across topology states and fault-resistance bins, while CPU inference contributes only 0.528 ms to the one-cycle total decision time of 20.53 ms.

Designing learning interfaces is often an ill-structured problem: effective solutions depend on learning goals, subject matter, learner characteristics, and interaction context that are difficult to generalize. Designers must integrate knowledge from learning sciences, interaction design, and subject domains when translating learning goals into interface decisions. Yet, prior efforts to connect learning theory and design have largely focused on learning activities or specific technologies, leaving interface-level design knowledge fragmented across systems. In this work, we take a bottom-up approach to surface this knowledge from existing educational interfaces. We present LINK (Linking INstructional Knowledge to interfaces), a library of 27 recurring interface patterns, organized under nine higher-level design commitments, which describe broader learning-science-grounded goals for interfaces. As part of developing LINK, we conducted a user study (N=14) with designers to evaluate LINK's generative power. Our findings suggest that LINK helps designers generate interface ideas and articulate theory-grounded rationales for design decisions.

Topological complexity, introduced by Farber, measures the number of continuous local motion-planning rules on a configuration space. For every simply connected space $X$ of finite rational homotopy type whose rational cohomology is a finite-dimensional Poincaré duality algebra, we prove $\mathrm{MTC}(X)=\mathrm{TC}_0(X)=\mathrm{TC}_0^M(X)$, where $\mathrm{TC}_0(X)=\mathrm{TC}(X_{\mathbb Q})$ and $M$ denotes the monoidal invariant. No formality or ellipticity assumption is required. The result extends to higher topological complexity and gives product additivity under duality. For every map $f$ between simply connected spaces of finite rational homotopy type whose rationalization admits a homotopy retraction, we establish the sharp bound $\mathrm{Msecat}(f_{\mathbb Q})\leq\mathrm{secat}(f_{\mathbb Q})\leq\mathrm{relcat}(f_{\mathbb Q})\leq\mathrm{Msecat}(f_{\mathbb Q})+1$. In particular, $\mathrm{MTC}(X)\leq\mathrm{TC}_0(X)\leq\mathrm{MTC}(X)+1$ holds without a duality assumption. The proofs use controlled Sullivan resolutions and an ideal-valued criterion for relative category. For sections of fibrations with Poincaré duality fibre, we prove that a derived self-intersection formula, vanishing of Thom multiplication on the augmentation ideal, and ideal-valued evaluation are equivalent. This criterion yields equality of the three rational sectional invariants for smooth sections of smooth bundles over simply connected smooth bases of finite rational homotopy type, with closed simply connected manifold fibre, including rationally hyperbolic fibres. Equality also holds when the fibre has finite-dimensional total rational homotopy, without a cohomological dimension bound on the base, and for suitable finite relative Frobenius models. Geometric applications include sphere, Grassmann, and projective bundles.

Video world models aim to preserve scene structure and predict how dynamic objects evolve beyond visual observations. We present Kepler4D, a framework for future video generation through explicit 4D scene state evolution. Given a monocular video, Kepler4D constructs a shared 3D representation of background geometry, object motion histories, coarse spatial supports, and semantic context. Chain-of-Motion summarizes observed motion and uses a vision-language model to select structured speed and heading decisions and decide whether to bound object-center height from below. A deterministic rollout converts these decisions into future object trajectories for inspection and editing before synthesis. We render the evolving proxies into geometric controls for a pretrained video generator, separating coarse object motion from the synthesis of appearance and articulation. Experiments on real-world videos demonstrate that Kepler4D enables controllable object motion and plausible future rollout while preserving scene consistency.

Content-provenance standards such as C2PA let a platform recover a stripped manifest through a soft binding: an invisible watermark read from the content, or a fingerprint looked up in a registry. We benchmarked both families under one protocol, restricted to openly available models whose licences we audited, on public media, with false-match rates calibrated on held-out negatives and source-level bootstrap intervals for performance estimates. For watermarking we evaluated 25 image, 7 audio and 7 video configurations from 12 methods on perceptual quality, robustness, false positives and cost; for fingerprinting, 35 methods on registries of up to 98,985 images, partial edits and adversarial attacks. PixelSeal gave the best balance for image and video watermarks and AudioSeal for audio, but the error-correcting detector of every TrustMark variant fired on 5.9 to 15.3% of unmarked images, so a verifier should test the expected payload. Among fingerprints, copy detectors trained on non-commercial data detected up to 75.5% of transformed images at a pair-level false-match rate of $10^{-7}$ and DINOv2, the best permissively licensed method, 65.0%; near-copies in a product catalogue dominated the false matches, and no detection improvement from geometric verification was observed under a matched calibration false-binding constraint in the evaluated image pipelines. On the same attacked copies the two families failed differently: the union of watermark and fingerprint successes covered 69% of image copies; fingerprints covered more audio queries, whereas expected-key watermark verification covered more video queries. Embedding a watermark moved the ISCC code of 62.0% of images past its match threshold, and platform-dependent colour conversion changed watermark bits between x86 and ARM hosts.

In this paper we prove the weak Hurwitz stability conjecture for the strong Turán expressions of Eulerian polynomials, as well as the corresponding conjecture for Narayana polynomials, which were conjectured by Liu and Yan. We generalize the notion of Turán expressions of polynomial sequences, and provide a criterion for the generalized stability of generalized Turán expressions. Based on our sufficient condition, we not only confirm Liu and Yan's conjectures but also obtain the generalized stability of generalized Turán expressions of a class of recurrent polynomial sequences and generalized Narayana polynomials.

We study the vanishing discount selection problem for critical Hamil\-ton–Jacobi equations posed on ${\mathbb R}^d$ with Hamiltonian $G=H-V$, where $H$ is continuous, periodic in space, convex and superlinear in the momentum, and $V$ is a compactly supported potential of constant sign. For $V\leqslant0$, we prove local uniform convergence of the discounted solutions to a distinguished critical solution, which we characterize in terms of Mather measures and of the solution selected by the unperturbed periodic problem. The proof relies on a decomposition of limiting discounted measures into a retained component and a periodic component accounting for the mass lost at infinity. For $V\geqslant0$, we establish convergence under the additional assumption that the unperturbed selected solution has zero average with respect to every periodic Mather measure. We also give explicit examples in which the discounted solutions converge, but natural formulas for the candidate limit fail to identify the selected solution.

We represent constructive integrals in spatially cut-off $\phi^4_2$ by sequences of exponential-motivic periods. For a nonzero spatial cutoff, a single Galois transformation makes the partition periods diverge in modulus along every sequence of approximations satisfying the constructive estimates. For a fixed cutoff presentation, we characterize the subgroup of transformations admitting a bounded realization on the free field's $L^1$ space and prove that all fixed-observable limits on this subgroup are a common scalar multiple of their physical values. A one-variable quartic example shows that all transformed fixed-observable sequences can converge without such a bounded realization, with limits that are not proportional to the physical values.

A decade after its inception, Inception crop has become the standard crop-based data augmentation method for training deep vision models. Not only is its practice of uniformly sampling crop scale and aspect ratio widely adopted, but also its lower and upper bounds, with the scale lower bound being the sole exception that is sometimes tuned. It is therefore surprising that the standard implementation in the TensorFlow / JAX ecosystem samples crop scale with probability density function $f(A) \propto \frac{1}{\sqrt{A}}$ unlike the PyTorch counterpart, which follows the original description. Motivated by this discovery, we train 522 ViT-S/16 models on the ImageNet-1k dataset with various training budgets and crop scale distributions. We reach $78.78\pm0.09$ top-1 val. accuracy with 90 epochs of training budget and find that 1. Higher training budget requires stronger augmentation; 2. Lower tail of the distribution of the crop scale determines the augmentation strength of Inception crop; 3. Models trained with higher training budget exhibit sparser saliency, regardless of the crop scale distribution or weight decay. Based on 2. we revisit the performance of Beta crop, whose softer cutoff allows it to optimize model performance across training budgets with less compromise. We replicate 1. and 3. with Scion optimizer in addition to AdamW, suggesting that the results may be general.

Imperfect measurements allow different underlying distributions to generate the same observations. Restrictions linking unobserved components can make it difficult to characterize all compatible distributions and compute the resulting range of parameter values. We ask how to construct computable parameter bounds and tighten them without additional data or stronger assumptions. We develop a general framework for partial identification under imperfect measurement based on distributionally robust optimization. We use measurement restrictions to construct a neighborhood around a benchmark distribution, ensuring that every compatible distribution lies within it. Optimizing over this neighborhood gives parameter bounds through a common procedure that can be adapted to different measurement problems. The choice of benchmark affects how informative these bounds are. We establish conditions under which changing it can tighten the bounds while preserving every parameter value consistent with the same data and assumptions. Applications to networks, missing outcomes, and regression illustrate how the framework tightens bounds and can recover the full range of compatible values.

Multiple sclerosis (MS) can impair upper-extremity function through motor, sensory, cognitive, and visual deficits. Although the Nine-Hole Peg Test (9-HPT) is widely used to assess dexterity, completion time alone provides limited insight into the eye-hand coordination underlying performance. We characterized 9-HPT performance in nine people with MS (PwMS) and nine healthy controls using simultaneous eye tracking, markerless hand tracking, and object detection. Each peg cycle was divided into transport, manipulation and insertion, and return-to-bin phases. PwMS had longer overall completion times. Transport was prolonged with the non-dominant hand, manipulation was prolonged with the dominant hand, and return-to-bin duration was longer with both hands, representing the largest phase difference. PwMS also made approximately one additional fixation per peg and showed greater variability in gaze landing position during dominant-hand performance. Saccadic amplitude and mean placement error did not differ significantly between groups, and gaze generally preceded hand movement in both groups. These exploratory findings suggest that MS-related slowing is concentrated in phases requiring precise manipulation and transition to the next peg. Multimodal assessment of the 9-HPT can reveal coordination deficits not captured by completion time alone.

Locomotion on granular slopes such as sand dunes remains a fundamental challenge for legged robots due to reduced shear strength and gravity-induced anisotropic yielding of granular media. Using a hexapedal robot on a tiltable granular bed, we systematically measure locomotion speed together with slope-dependent normal and shear granular resistive forces. While normal penetration resistance remains nearly unchanged with inclination, shear resistance decreases substantially as slope angle increases. Guided by these measurements, we develop a simple robot-terrain interaction model that predicts anchoring timing, step length, and resulting robot speed, as functions of terrain strength and slope angle. The model reveals that slope-induced performance loss is primarily governed by delayed anchoring and increased backward slip rather than excessive sinkage. By extending the model to generalized terrain conditions, we construct failure phase diagrams that identify sinkage- and slippage-induced failure regimes, enabling quantitative risk estimation for locomotion on granular slopes. This physics-informed framework provides predictive insight into terrain-dependent failure mechanisms and offers guidance for safer and more robust robot operation on deformable inclines.

Most feature-based choice models, classical and deep, score items and apply a single softmax. We introduce consideration circuits (CC), feature-based models of multi-stage choice defined by directed acyclic graphs of multinomial logit (MNL) units. Source units assign probabilities to menu items, and internal units combine predecessor distributions using MNL weights computed from their probability-weighted feature summaries. On a three-item compromise task with fixed non-collinear features, menu-independent random-utility models (RUM), including a single MNL unit, suffer an error bounded away from zero. For CC, in contrast, we establish a sharp depth–norm separation: increasing depth from $2$ to $3$ reduces the optimal maximum taste-vector norm for error $\epsilon$ from $\Theta(\log(1/\epsilon)/\epsilon)$ to $\Theta(\log(1/\epsilon))$. The depth-$2$ lower bound holds for arbitrary width and menu-independent routing biases, while a five-node depth-$3$ circuit with zero routing biases attains the logarithmic rate. More generally, we characterize two geometric conditions that are necessary and sufficient for approximating arbitrary deterministic choice tables on finite menu families. Under these conditions, depth $3$ suffices, while depth $4$ achieves optimal logarithmic norm scaling whenever the family contains a non-singleton menu. In experiments, standalone tree circuits with fewer than $600$ parameters attain the lowest mean test negative log-likelihood (NLL) among the evaluated models on four fixed-pool benchmarks and the Expedia temporal split. As output heads, CC generalize the linear MNL readout and lower mean test NLL for every tested encoder on Expedia and Trivago.

Modern training on a new task often starts from a previously trained model rather than from scratch, raising the question of how this initialization affects the subsequent training trajectory. Classical implicit-bias results characterize the direction selected by prolonged training, but this direction alone does not provide information regarding the intermediate behavior. We address this question through a geometric approximation of full-batch logistic gradient descent (GD) trajectories on strictly linearly separable data, with large initialization of scale $R$ motivated by prior training. From any limiting normalized initial position, we use minimum-norm projection rules to construct a unique continuous ideal path consisting of finitely many linear segments. The path has two stages: negative-margin correction followed by minimum-margin growth. We prove that, after an explicit two-stage time reparameterization, the fixed-step GD trajectory divided by $R$ converges uniformly to this path on every fixed parameter interval as $R\to\infty$. Further, our quantitative error bounds account for initialization perturbations and the transition between stages. This approximation provides asymptotic formulas for peak evaluation loss and cumulative training loss. In particular, peak evaluation loss can grow linearly in $R$ even when both endpoint losses tend to zero. The cumulative losses in the correction and margin-growth stages, normalized by $R^2$ and $R$, respectively, converge to explicit limits. Experiments on controlled geometries and fixed image features complement our theoretical results.

Matter aims to unify smart home ecosystems through an open, interoperable, and secure standard, relying on cryptographic mechanisms for device authentication and data confidentiality. However, its openness also exposes protocol details, credential structures, and implementation characteristics to adversaries. We show that an attacker with temporary physical access can exploit design and implementation flaws to extract credentials and impersonate devices and controllers. These replicas integrate seamlessly into the fabric, enabling persistent surveillance and control even after the attacker departs. Notably, the attack is vendor-agnostic and requires no device-specific reverse engineering, as long as the device is not physically secured. We validate its practicality through a proof-of-concept on a simulated smart home with both commercial devices and development boards. Our findings uncover previously underestimated attack surfaces in Matter, particularly under physical access scenarios, and motivate four targeted mitigation strategies.

Mixture-of-experts (MoE) language models send each token to a few experts, so each expert is trained on a different part of the data, and this part changes during training. With a shared learning rate, Muon applies updates of roughly the same size to expert matrices of the same shape, even when an expert's update is poorly aligned with its current gradient. We here propose ExpertMuon-Compass (Compass), which multiplies the Muon step of each expert by two factors. A family factor compares the cosine between the expert's orthogonalized update and its gradient with the same cosine for the other experts in its layer. A scalar radius aggregates the alignment between corresponding rows of the update and gradient into one step-length multiplier. Compass keeps the update direction and the momentum buffer of Muon. In pretraining on FineWeb-Edu, Compass with Nesterov momentum on all matrices performs as well as or better than Muon, NorMuon, and other optimizers, with weight decay matched to NorMuon in the longer runs. Adding its factors to NorMuon gives the same or a lower loss than NorMuon. Compass is the most effective when the data seen by each expert varies during training, for example, as when the languages of a multilingual corpus arrive in separate blocks. With Compass, the expert load stays balanced, and the router assigns tokens to experts more decisively. We also prove a perturbation bound for the two factors.

Predicting future spatial states supports collision avoidance and timely decision-making in dynamic environments. However, existing vision-language models (VLMs) and benchmarks for spatial reasoning primarily focus on observed scenes, leaving predictive spatial reasoning beyond the observed interval underexplored. To this end, we introduce SpatialMind, a metric-scale VLM for spatial reasoning and future prediction. Its metric depth adapter anchors spatial reasoning to real-world scale, while its progressive state chain establishes current spatial states and observed dynamics as the foundation for future prediction. Given a video prefix, SpatialMind predicts distances, motion directions, and spatial relations in both observed and unseen future frames. For training and evaluation, we build a scalable data engine that grounds entity descriptions in metric geometry to generate question-answer pairs and state supervision. Using this engine, we construct the SpatialMind-30K dataset and the SpatialMind-2K benchmark, both covering driving and everyday egocentric scenes. The benchmark spans eight tasks across three levels: current-state understanding, observed-dynamics understanding, and future prediction. Experiments show that SpatialMind substantially outperforms both general and spatially specialized models on our benchmark while achieving competitive zero-shot performance on VSI-Bench, OSI-Bench, and VLM4D.

Account takeover (ATO) fraud is a growing threat to digital banking, requiring effective detection while minimizing friction for legitimate customers. Production systems predominantly rely on tabular models that score sessions in isolation, discarding the relational structure of the underlying interaction network. Although graph-based models exploit relationships among sessions and network entities, they primarily reason over local neighborhoods and therefore capture only part of the problem: fraud risk depends jointly on the local relational structure surrounding a session and the evolving global state of the fraud ecosystem. We present HERMES (HEterogeneous Relational Micro–macro graph transformer Encoder for high-risk Sessions), a dual-scale architecture that jointly models these complementary scales of information. Micro-GT captures local heterogeneous graph structure through structured, relation-aware attention over a temporally safe session neighborhood. Complementing this local representation, Macro-GT models ecosystem-level context using non-anticipative climate tokens that summarize fraud dynamics, platform shifts, and infrastructure reuse, together with adaptive class prototypes that track representative fraud and benign session patterns over time. Evaluated on more than 130 million high-risk transaction sessions from a leading U.S. financial institution, HERMES consistently outperforms production and strong graph-based baselines, achieving a 44.44% relative reduction in customer friction and a 24.66% relative improvement in fraud recall over the production system. Ablation and temporal-stability analyses further demonstrate complementary gains from local relational modeling and global ecosystem context across changing fraud regimes.

Agent harnesses specify the roles, instructions, tools, and communication structure used to solve a task, and the right harness depends on the query. Because the value of each design choice is observable only through execution, tailoring a harness to each query has required either executing alternatives at inference time or costly manual design. We introduce SHIFT, which moves execution out of the per-query search loop. A local LLM architect learns a policy over harness-building actions from search, and a value function that predicts, from measured executions, a utility balancing accuracy against execution cost. For each query, Monte Carlo tree search uses these predictions to construct a harness. Across 9,193 tasks in six benchmarks, from math to document and general-assistant tasks, with a Gemini 3.5 Flash executor, SHIFT attains the highest mean accuracy, about 80%, outperforming 17 baselines that span prompting, prompt optimization, and workflow search, and exceeding the strongest baseline by 7.2 percentage points. A cheaper mode of SHIFT also attains a higher mean accuracy than every baseline while using 32% fewer execution tokens than the strongest baseline. We further show that choosing structure, instructions, and tools jointly beats choosing only instructions or only tools by up to 9.1 percentage points, and that learned value selection identifies more accurate harnesses with lower execution cost from candidate pools.

We study central configurations for $n$ bodies interacting through pair potentials \[ U_\sigma = \frac{\kappa}{\sigma} \sum_{i<j}m_i m_j r_{ij}^{\sigma}, \] with logarithmic limit at $\sigma=0$. For the Newtonian exponent $\sigma=-1$, the classical Chazy–Wintner–Smale finiteness problem asks whether, for prescribed positive masses, there are only finitely many central configurations up to similarity. Finiteness is known in several important special cases and under additional dimensional or genericity hypotheses. It is therefore natural to ask whether an analogous finiteness principle continues to hold for other homogeneous exponents. We show that for every positive even exponent \[ \sigma=2k, \qquad k\ge2, \] the answer is emphatically negative. In fact, these exponents admit an abundance of continuous families of pairwise non-similar central configurations. The mechanism is provided by spherical designs: on a common sphere the force kernel for $\sigma=2k$ restricts to a polynomial of degree at most $k$, and consequently every positive weighted spherical $k$-design is a central configuration. We compute the corresponding central multiplier explicitly.

We study pointwise estimation of a smooth binomial mixing density and characterize its minimax rates. Assuming that the mixing density is $s$-Hölder smooth, we first consider the homogeneous setting with a common number of binomial trials $t$. We derive matching lower and upper bounds that reveal three regimes depending on the relative sizes of the sample size $n$ and the number of trials $t$. When $t$ is small, the finite number of trials creates an identification barrier that persists regardless of sample size; at intermediate $t$, more data reduce sampling uncertainty, while the number of trials limits how finely the mixing density can be recovered; and when $t$ is sufficiently large, the usual nonparametric density estimation rate is recovered. Our lower bounds exploit the polynomial structure of the binomial mixture model, while attainability is achieved by a suitably regularized orthogonal series estimator. We then extend the analysis to heterogeneous trial parameters, where the minimax rate depends on the full trial profile through the effective sample size available at each polynomial degree. These results characterize the fundamental statistical limits of learning smooth populations of binomial probabilities under both homogeneous and heterogeneous trials.

The discovery and development of advanced materials is a challenging process constrained by the high time and monetary costs of synthesis, processing, and characterization. The underlying design spaces can be enormous, often with multiple competing objectives. Bayesian optimization (BO) provides a principled approach for efficiently navigating such spaces, but most workflows rely on fixed exploration-exploitation policies that lack the capacity to adapt to shifting constraints in dynamic campaigns typical of self-driving laboratories. In this work, we develop a multi-objective BO framework for alloy design under resource constraints, benchmarking strategies for adaptive policy tuning at each iteration. Our evaluation covers a septenary refractory high-entropy alloy (RHEA) system focused on maximizing melting temperature and minimizing density, and an Fe-Co-Ni-based soft magnetic alloy system targeting saturation magnetization, coercivity, and hardness. We compare an exploitation-focused strategy, a fixed mixed exploratory/exploitative policy, and two distinct LLM-based adaptive strategies with different approaches to batch allocation and campaign signal interpretation, evaluated across baseline and mid-campaign resource event conditions including budget reductions, timeline cuts, and combined disruptions. Our results show that mixed allocation strategies accumulate substantially more mutual information than the exploitation-focused baseline at a proportionally smaller cost to hypervolume and optimization speed, with adaptive strategies outperforming a fixed-mixed allocation policy by adjusting their allocation in response to both evolving campaign statistics and resource constraints. These findings suggest that adaptive resource allocation offers a favorable tradeoff for materials discovery campaigns in reducing predictive uncertainty on Pareto-optimal compositions.

Multi-agent systems built on large language models (LLMs) are increasingly applied to scientific discovery and hypothesis generation. Both the effect of refinement and the diversity of the delivered set are hard to interpret before experimental ground truth exists, and both are typically reported by deciding whether pairs of generated hypotheses describe the same underlying mechanism. We study two evaluation questions that rest on this pairwise equivalence judgment: (1) how much self-critique changes delivered hypotheses beyond run-to-run variability, and (2) how the equivalence rule used to group hypotheses affects measured diversity. Across four proprietary instances, we hold opening hypotheses fixed, rerun the downstream workflow with 0, 1, and 5 critique rounds, and score matched hypothesis pairs with an LLM-as-a-judge. Relative to matched same-depth reruns, moving from 0 to 1 round produces 34.5 percentage points (pp) of additional mechanism-level divergence, whereas 1 to 5 rounds adds 1.3 pp. We then compare three equivalence rules: term frequency–inverse document frequency (TF–IDF) similarity, dense embeddings, and the same LLM-as-a-judge. We construct controlled hypothesis pairs that either preserve the causal explanation through wording or biological-terminology changes, or replace one component of the causal chain while holding the rest fixed. All three rules are invariant to meaning-preserving edits, but when the initiating event is replaced, the LLM-as-a-judge identifies 83% of valid pairs as different mechanisms, versus 0% for TF–IDF and 8% for embeddings; varying only the rubric that defines same mechanism moves this figure from 38% to 96%. Together, these results show that pairwise equivalence judgments are a measurement choice: how mechanism equivalence is defined affects both the estimated effect of self-critique and the measured diversity of generated hypotheses.

Building LLMs that behave well socially, not merely correctly, requires Building LLMs that behave well socially, not merely correctly, requires more than producing locally helpful responses. A socially competent agent must infer users' unstated goals, respect their preferences, and adapt as the conversation unfolds. These behaviors are inherently multi-turn and social, making them hard to optimize: real interaction data is scarce, and user preferences are typically latent rather than directly observable. To address these challenges, we build on a persona-driven social simulation environment (consisting of a persona library, LLM-based user simulators, and a user-satisfaction scoring system ranging from [0, 1]), to introduce preference-batched GRPO (PB-GRPO), a post-training algorithm that learns socially adaptive policies from conversation-level feedback. Compared to vanilla GRPO, PB-GRPO computes advantages using a normalization estimated across a bucket of users with similar preferences, stabilizing training across a diverse social population. Empirical evidence shows that PB-GRPO improves models' social behavior over strong reinforcement learning baselines in our simulated environment.

We study anomaly constraints in six-dimensional $N=(1,0)$ supergravity with gauged $Sp(1)_R$ symmetry, one tensor multiplet, $n_V$ vector multiplets and $n_H$ hypermultiplets. Imposing cancellation of local gravitational, gauge and mixed anomalies, together with the embedding of the anomaly coefficients in a unimodular rank-two string charge lattice, the number of vector multiplets is sharply constrained: the only allowed positive values are $n_V=12$ and $n_V=96$. For a neutral abelian factor, or drone, the anomaly equations force the corresponding row and column of the abelian gauge-kinetic matrix to vanish. Such a factor therefore cannot be retained as an ordinary propagating vector multiplet. For $n_V=96$, within the class of semisimple non-$R$ gauge algebras containing no $A_1$ or $A_2$ factor, the exhaustive spectra-assembly calculation yields no drone-free spectrum. For $n_V=12$, the semisimple non-$R$ branch is necessarily $A_1^3$. In the doublet/adjoint truncation described in the text, we find $4{,}337{,}331$ consistent models, classified by $102$ anomaly-vector classes. We also consider the reductive branch $A_2\times U(1)\times Sp(1)_R$. For $A_2$ representations in $\{\mathbf 1,\mathbf 3,\overline{\mathbf 3}, \mathbf 6,\overline{\mathbf 6},\mathbf 8\}$ and primitive $U(1)$ charge assignments with $|q|\le 2$, we find $178{,}538$ solutions, classified by $21$ sets of anomaly vectors.

In this paper we investigate whether geometrically aligned astronomical representations are also scientifically interchangeable. We analyze every layer's representation from 36 pretrained models using cross-matched optical images, infrared images, and spectra. All 139 analyzed model-survey comparisons show significant local-neighborhood alignment somewhere in the network after accounting for the search over depth. However, alignment does not generally increase with network depth and can be stronger for the changes between consecutive computational blocks than for the block outputs themselves. We find that geometry-selected stages under-perform label-selected stages in every image-transfer comparison; moreover, every final HSC-COSMOS-Web model pair is aligned while every redshift transfer has negative R^2. Models can therefore recover a similar cross-survey geometry among astronomical objects without recovering a survey-invariant linear encoding of the physical properties tested here.

We introduce InvestigationWorlds, an agentic environment for legal investigation. We build on an underused artifact of U.S. civil litigation: the summary judgment motion. This motion relies upon a record composed of real evidence exhibits, and results in a court-adopted hypothesis that is treated as ground truth for the purposes of deciding the motion. Each environment is built from a real U.S. Federal Court case retrieved from Public Access to Court Electronic Records (PACER) and augmented by an attorney-validated generation pipeline that synthesizes role-tagged documents around the original record. The resulting corpus admits multiple coherent factual readings, only one of which matches the court-adopted hypothesis. Evaluating on 100 cases, we find agents often commit to incorrect hypotheses despite retrieving relevant evidence, struggling to distinguish the court-adopted hypothesis from alternative hypotheses.

Split learning lets a client train a language model on a server without sending its text. The client runs the first layers itself and sends the server only their output, a vector of numbers for each token. During training, the server sends gradients back. We show that an observer at the split can rebuild most of the client's text from this traffic, and we measure how much the gradients help. On GPT-2, an attacker who holds only the publicly released weights of the client's layers recovers 94.20% of tokens from the activations alone and 97.38% when it also sees the gradients, 3.17 percentage points more 95% interval [2.72, 3.64]. Counted by document, the difference is much larger. The attacker rebuilds 13.71% of 32-token documents exactly without the gradients and 37.77% with them, because a document only counts when every token is right. How we count also changes how good a defence looks. Secret mixup, which blends each outgoing vector with a decoy, stops the attacker from rebuilding almost any document exactly, yet the attacker still recovers 83-91% of tokens. In a second experiment on GPT-2 and Qwen3-0.6B, where the server trains only a run of consecutive layers, the layer at which the run starts changes both model quality and leakage, even when the run's length is fixed. We recommend reporting leakage both per token and per document, and treating what a split model sends as being as sensitive as the text itself.

Per-target labeled random-finite-set filters, such as the state of the art $\delta$-generalized labeled multi-Bernoulli filter, cannot represent cooperating, swarm-like, motion between targets. In this work, we present a new filter that can. The new particle filter couples the joint particle filter with ensemble Gaussian mixture filter such that a measurement updates the density at every particle instead of merely reweighting. We prove that the new particle filter converges to the joint particle filter in the limit of particle number and thus to the true Bayesian posterior, under certain assumptions. Experimental results on two coupled systems, the Vicsek model and coupled Brownian motion provide evidence on the efficacy of the filter.

The search for extraterrestrial intelligence (SETI) is conducted across the electromagnetic spectrum to understand the prevalence and distribution of intelligent life in the universe. At radio frequencies, this effort focuses on searching for "technosignatures", observable indicators of technological activity consistent with our understanding of an advanced civilization. Historically, the microwave regime from 1-10 GHz has been the primary frequency range searched for these technosignatures, but recent SETI experiments utilizing low-frequency radio telescopes, such as the Low Frequency Array (LOFAR), Murchison Widefield Array, and MeerKAT radio telescope have explored the insufficiently studied frequency range below 1 GHz. However, the lowest radio frequencies ($<$100 MHz) have been even less explored for technosignatures, despite significant human communication signals occupying this band. The Long Wavelength Array (LWA) is a low-frequency (3-88 MHz) interferometer array telescope collaboration comprised of dipole 'stations' distributed across the Southwestern United States, and the LWA Swarm is the aperture synthesis telescope consisting of interconnected LWA stations operating as a long baseline interferometer. Here we present a narrowband technosignature pipeline using the Breakthrough Listen Interesting Signal Search algorithm for raw beamformed single-station LWA data and outline its planned extension to correlated visibilities from the LWA Swarm. We demonstrate the pipeline on a limited selection of fast radio burst observations to evaluate the data processing and spatial coincidence filtering for a future observing campaign of exoplanet systems.