Papers

A deep understanding of quantum information theory is essential for quantum science and technology. However, quantum mechanics and particularly quantum information theory suffer from a lack of intuitive images. Visualization tools offer novel perspectives on many problems in quantum physics by revealing important properties of quantum states and processes that are otherwise obscured by abstract mathematical formalisms. Here, we extend the previously established BEADS representation of pure states to the so-called generalized BEADS representation of mixed quantum states. We introduce spherical functions, so-called beads to visualize correlation structures of mixed quantum states. A recursive optimization scheme is used to obtain minimum-entanglement pure state decompositions of all subsystems of a given N-qubit state. We show how correlation function components can be decomposed into individual contributions that directly correspond to different types of correlations of mixed states. This allows us to visually characterize the total correlation, genuine n-partite entanglement, nonclassical quantum-discord-type correlations as well as classical correlations. After outlining the mathematical details relevant for constructing the representation, we examine the generalized BEADS representations for various classes of mixed states including product, classically correlated, separable but nonclassically correlated, as well as entangled states. As exemplary applications of the generalized BEADS representation, we showcase entanglement distillation and the effects of noise channels.

Inference of phylogenetic trees from sequences plays a central role in understanding evolutionary relationships, with applications ranging from tracking pathogen spread to reconstructing the history of life. Conventionally, practitioners obtain posteriors over trees using likelihood-based methods like MCMC. However, likelihoods are only available for simple evolutionary models with restrictive assumptions. For realistic evolutionary models, inference using conventional methods is prohibitively expensive, inaccurate, or even impossible. In practice, practitioners settle for using as flexible a conventional method as possible, even though they expect it is misspecified. In principle, we could use simulation-based inference: simulate tree and sequence data from an evolutionary model and train a neural network to predict the posterior over trees given sequences. However it is a mathematical challenge to represent and train complex posterior distributions over trees. Here we present a flexible family of conditional generative models over trees. Our models iteratively build trees using three natural paradigms – top-down, middle-out, or bottom-up – and are trained using a discrete diffusion framework. For real and synthetic datasets of up to $65$ taxa and $>10^5$ tokens, our models recover MCMC posteriors much more accurately than the previous state-of-the-art simulation-based method Phyloformer 2. Finally, we compare to IQ-Tree, a popular flexible-but-misspecified conventional method. Our models match its tree estimates on simple data, but on more realistic data, our method is reliably more accurate.

Video-based remote physiological measurement (RPM) has broad applications, while Vision-Language Models (VLMs) provide a promising opportunity to enable end-to-end physiological monitoring for future human-aware intelligent systems. However, existing RPM methods rely on task-specific models and cannot be seamlessly integrated with the multimodal understanding and interaction capabilities of VLMs. We revisit RPM with VLMs and provide empirical evidence for two major bottlenecks: generic visual semantic representations fail to preserve fine-grained physiological cues; while discrete token prediction in autoregressive generation is inherently misaligned with continuous numerical measurement. To address these challenges, we propose VITAL (Visual-physiological Token Alignment for Language Models), a visual-physiological token alignment framework for RPM and preserving general vision tasks. VITAL retains the original visual pathway for semantic understanding and introduces an independent physiological pathway to jointly model semantic and physiological information. Moreover, we propose a structured numerical alignment approach to improve the reliability of generative heart rate estimation by enhancing numerical relationship modeling. Extensive experiments show that VITAL improves cross-domain estimation performance on multiple rPPG datasets while preserving the general visual understanding capability of VLMs, demonstrating its effectiveness as a unified visual-physiological model. Code will be released.

Coordinators that act on what they infer about teammates are usually certified by the accuracy of those inferences. We argue that accuracy is the wrong certificate, for two reasons. First, the accuracy needed to act depends on what a wrong action costs, so the bar belongs to the deployment, not the module. Second, a module's errors are rarely independent, and dependence has a sign: positively associated errors help where their damage overlaps and hurt where it compounds, so an accuracy threshold fails in a predictable direction, and once wrong beliefs interact and can do harm, no rule based on accuracy is right for every module. We test these predictions in task assignment, Level-Based Foraging and Hanabi, with injected errors, trained and Bayesian modules and five language models, and in human judgments. Accuracy fails in both directions, approving harmful language-model deployments in Hanabi, where shared errors compound, and rejecting beneficial ones in task assignment, where costly errors are rare. In our tasks it also buys far less redundancy than independence predicts: from GPT-5.4-mini to Claude Opus 5.5, error on teammates the models have evidence about falls from 32% to 11%, but the chance that all three observers misread one falls only from 16% to 9% (independence: 3.4% to 0.15%). We propose certifying the deployment, with a Blind-Spot Gate that prices error costs, pairwise interactions and measured dependence without deploying the module. On held-out data from the cells it was developed on, it makes no significant false approval in 20 splits (accuracy: 3.2 per split), and in preregistered tests on modules it was not developed on, it recovers what accuracy rejects (seven classifiers: regret 0.004 against 0.479) and catches what accuracy approves (a Bayesian module: significant false approvals in 17 of 20 splits for accuracy, 1 for the gate).

Dense cardiac dynamics are essential for assessing cardiac function, yet in practice clinicians often observe only extremely sparse measurements. However, recovering full dynamics from such sparsity is a highly ill-posed problem due to the lack of temporal cues. While auxiliary time-series signals like ECG provide physiological guidance, effectively exploiting them is challenging due to the cross-modal representation gap and phase misalignment. In this work, we present CardioFAD, a Focus–Align–Diffuse paradigm that reformulates sparse cardiac dynamic synthesis as a recoverability–alignability anchor discovery problem. Our core insight is that resolving extreme temporal ambiguity requires a compact set of anchor states that are both sufficient to recover the visual trajectory and reliable for ECG–visual alignment. We therefore learn motion-salient keyframes under two complementary principles: trajectory sufficiency, which preserves trajectory-critical intra-modal motion content, and anchor alignment, which promotes consistent inter-modal correspondence at these anchors. We implement the Focus and Align modules optimized by the two aspects. Diffuse then performs two-stage generation by first synthesizing keyframes and subsequently interpolating the remaining frames. Extensive experiments on cardiac MRI and echocardiography videos demonstrate superior visual quality, temporal coherence, and improved functional consistency under the extreme one-shot regime.

Axion-like particles (ALPs) with sufficiently low masses are dark matter candidates because they could account for much or all of the observed dark matter energy density and help explain observations of galactic substructure. Such pseudoscalar fields could couple to photons, causing a time-dependent rotation of the polarization of the light with an oscillation frequency equal to the ALP mass. We use measurements of the polarization of the cosmic microwave background (CMB) by the Atacama Cosmology Telescope (ACT) between 2017 and 2022 to search for this effect using an EB power spectrum estimator. By probing oscillation periods from days to years, the data constrain the coupling between ALPs and photons for ALP masses between $2.5\times 10^{-23}$ eV and $4.8\times 10^{-20}$ eV. We find no evidence for such a signal and set a 95% upper bound on the axion-photon coupling constant, $g_{\phi\gamma} < 1.1\times10^{-11}\left({m_\phi}/{10^{-21}\,\mathrm{eV}}\right)\,\mathrm{GeV}^{-1} .$ This limit assumes that ALPs make up all of the dark matter and that the ALP field amplitude is stochastically distributed, which leads to more conservative bounds than the deterministic-field assumption used in some previous CMB polarization analyses. Among searches for ALP-photon coupling using CMB polarization, we set the strongest bounds to date under the stochastic-field assumption. This work develops a methodology to analyze over 10,000 time-sliced "depth-1" polarization maps from ACT, a data product that will be standard for time-domain astrophysics with the Simons Observatory, enabling searches with a higher sensitivity and observing span.

Record linkage and demographic inference from names are routine demands in social science. Most approaches to these problems require labeled training data that are difficult or costly to acquire. I show that pre-trained language models perform both tasks accurately without requiring task-specific training. For record linkage, I introduce EnsembleLink, an algorithm that uses pre-trained embeddings alongside string-based scorers to retrieve a short list of candidates and then identify the likeliest matches. EnsembleLink matches or exceeds the performance of most prior approaches while generalizing across scripts and formats that those methods are unable to handle. For ethnicity inference from names, I show that large language models (LLMs) achieve up to 83.8% accuracy on U.S. voter files, compared to 78.0% for name-supplemented BISG. Open-weight models offer reproducible implementations that approach these models in accuracy. Moreover, LLMs are able to infer group membership for religious, caste-based, and other ethnic group categories outside of the United States, and reduce bias in downstream estimates of geographic racial gaps. Where pre-training encodes the world knowledge required by each task, task-specific labels add less information.

Mixed-precision algorithms exploit reduced-precision hardware, however computing with the same matrix at distinct precisions typically requires duplicate storage, repeated conversions, or operation-emulation. We introduce TOMOS, a composite-precision matrix format that provides implicit 32-bit and 16-bit access from a single representation. Each FP32 entry is split into a 16-bit primary word for low-precision computation and a complementary word that enables exact FP32 reconstruction. The encoding scheme improves numerical quality of the low-precision view through subnormal representations, and allows exact recovery of high-precision inputs. TOMOS uses FP32-sized storage, with data movement scaling with compute precision. Optimized GEMM and GEMV kernels achieve throughput comparable to vendor libraries on NVIDIA A100 GPUs, while a mixed-precision least-squares solver achieves up to 1.6x speedup over FP32 without redundant matrix allocations.

Information-theoretic exploration is of primary interest in methodological studies of reinforcement learning (RL) and computational cognitive neuroscience, while the established RL theory builds regret guarantees on confidence sets. This situation leaves open the conditions under which the information-driven exploration found in nature is more advantageous than the exploration of the established RL theory. We put the state-of-the-art methods of these two paradigms, Active Inference (AIF) and Bayes-UCBVI, on an equal footing: both act through the same state abstraction of a tabular MDP with $S$ states, and their greedy forms differ only in the exploration term. We find that AIF's information gain decays too fast to serve as a confidence width, locking greedy agents that explore by it out of the optimal action. Calibrating the information gain by a square root yields calibrated AIF, which attains regret of order $\sqrt T$ for an exact abstraction, but no better guarantee than Bayes-UCBVI for a fixed reward. When an adversary chooses rewards from the agent's data, Bayes-UCBVI's value-specific widths lock out, whereas calibrated AIF pays a factor $\sqrt S$, which no goal-memoryless agent avoids on some MDP whose start states the adversary also chooses. Calibrated information gain thus acts as a reward-agnostic $\ell_1$ confidence set, preferable to value-specific widths when rewards are chosen from the agent's data, not when they are fixed.

Pediatric vaccine procurement for low- and lower-middle-income countries (LICs and LMICs) requires long-term coordination across heterogeneous vaccine markets, competing stakeholder objectives, and uncertain demand, yet tenders are currently scheduled as a series of independent renewals. We extend the Hypothetically Coordinated Vaccine Market (HCVM) framework to schedule multi-year tenders with flexible lengths and all-unit quantity discounts, and compare three objectives: a balanced coordinatingentity model (UG), a social-benefit model that minimizes missed doses (SB), and a producer profit-maximizing model (MP). Using UNICEF ten-year demand forecasts and 30 demand replications across six discount scenarios, we find that the procurement objective, not demand variability, drives cost-to-tender and schedule complexity. Quantity discounts favor large producers, concentrating the market and increasing supply chain vulnerability. Discount effectiveness depends on market maturity rather than valency: pentavalent and hexavalent vaccines are already near their price floor, and among monovalent vaccines only rotavirus responds to discounts, whereas HPV and PCV remain too expensive to compete. Removing pentavalent and hexavalent vaccines from the procurement pool sharply increases missed doses. These results give planners practical guidance on objective selection, discount design, and protecting access to combination vaccines.

Tiny robots have limited computation and memory, yet visual navigation systems often maintain more state than a task requires. We ask whether persistent memory can instead be shaped by the task. We present RingNav, a \(10.6\)k–parameter visual heading estimator that constrains its memory to the circular manifold \(S^1\). RingNav learns only the turn between consecutive images, while heading is composed directly on the circle. On the TartanAir Visual SLAM Challenge, RingNav reduces mean absolute turn error by \(27\%\) and accumulated heading error over $4$ sec by \(15\%\) relative to a similarly sized learned-memory model. The structured update also gives exactly zero turn for identical frames and exact cancellation under sequence reversal. Without fine-tuning, RingNav transfers to real-world drone experiments and continues estimating heading under severe visual degradation. RingNav runs at $27.3$Hz on Raspberry Pi Zero 2W and can be deployed on an ultra-low power GAP8 processor at $33.4$Hz.

Because its binary optimization landscape contains exponentially many local optima, MAX-CUT is highly sensitive to initialization. We study Particle Polynomial Relaxation (PPR), a moment-based nonconvex continuous reformulation possessing an advantageous optimization landscape. We warm-start PPR with Combinatorial Multigrid (CMG), a spectral graph-coarsening method that produces structure-aware initial solutions in near-linear time without training data. On t2g20 ($n=400$), benchmark instance with known exact optimum, the combined CMG+PPR pipeline reaches $97.5\%$ of optimum on average and $98.4\%$ on the best run. An extended variant with discrete local-search refinement reaches the published optimum on all sixty Biq Mac instances evaluated by a recent neural branch-and-bound method. Using a sparse CSR representation, the same approach scales to G-set instances with up to $20{,}000$ vertices reducing memory requirements by up to $3333\times$ while maintaining a consistent improvement over the CMG warm start. Gains are largest on sparse weighted graphs; on dense graphs we complement the pipeline with spectral sparsification.

Understanding the mechanism behind why diffusion models memorize their training data remains an open question in generative modeling. In this paper, we examine this question from two perspectives: (a) the model-level perspective concerning the onset of memorization in a model during training, and (b) the sample-level perspective concerning the proclivity of an individual sample to be memorized by a given model. We show that the so-called input loss curvature, or the Frobenius norm of the Hessian of the diffusion model's training loss, is an effective metric for quantifying both perspectives on memorization. In particular, we empirically demonstrate that input curvatures averaged across training and validation samples can identify when a model is likely to start regurgitating its training data when tracked over training time. However, this metric is not sufficient for measuring an individual sample's likelihood of being memorized. To address this perspective, we introduce a counterfactual definition of memorization in diffusion models, extending the analogous notion from supervised learning proposed by [Feldman, 2019]. We prove that this counterfactual memorization is controlled by the input curvature of the leave-one-out model, thus establishing a direct theoretical link between a sample's local loss landscape and its generation likelihood. We validate our derived bounds through several empirical experiments and show that input-dependent curvature provides a proxy for memorization.

Understanding the dynamics of language acquisition in Large Language Models (LLMs) is crucial for uncovering how statistical representations evolve into structured linguistic capabilities. However, most existing studies focus primarily on fully trained models, leaving the developmental trajectory across intermediate training stages largely underexplored. In this work, we leverage intermediate checkpoints from open-source models (e.g., Pythia and OLMo) under teacher-forcing evaluation to systematically investigate language acquisition across training steps and model scales. We focus on four core characteristics: morphological structure acquisition, embedding drift, repetition dominance, and co-occurrence effects. Our findings reveal that models acquire morphological structures well before semantic embeddings stabilize. Furthermore, we show that training dataset size has a positive impact on divergence recognition, while co-occurrence during pre-training and repetition during inference have a negative impact on divergence recognition and associative link generation.

Simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RIS) enable programmable wave control on both sides of a surface, extending wireless coverage beyond single-sided operation. However, practical STAR-RIS demonstrations in literature have narrow bandwidths, because it is challenging to achieve wideband reflection and transmission simultaneously from the same aperture. To overcome this limitation, a wideband STAR-RIS architecture is proposed based on shared-aperture polarization conversion and diagonal switching topology. Here, multiple resonances enable wideband magnitude responses and mirror-symmetric current reversal maintains $180^{\circ}$ phase difference in both reflection and transmission modes. Accordingly, a diagonal four-branch unit cell configuration is developed, where the tailored top patch directs the received wave either through a pair of blind vias connected by a buried strip to the shared top aperture for reflection, or to the bottom ring patch using a through via for transmission. Based on the proposed design, a 10 by 10 STAR-RIS prototype is fabricated and experimentally characterized. The measured 3-dB bandwidths reach 19.8% in reflection mode and 21.9% in transmission mode. Near-field measurements show average signal enhancements of 19.8 and 22.6 dB for reflection and transmission, respectively, over the wide frequency range of 6.8-8.4 GHz. Far-field measurements further demonstrate average received-power enhancements of 11.2 dB in reflection and 11.9 dB in transmission. To the best of the authors' knowledge, this work presents the first experimentally demonstrated wideband STAR-RIS covering the 7.125-8.4 GHz band within the emerging 6G FR3 spectrum.

Post-quantum cryptography (PQC) and quantum key distribution (QKD) address the quantum threat from opposite directions: PQC replaces vulnerable public-key algorithms with schemes believed to resist quantum cryptanalysis, while QKD derives keys whose secrecy rests on physical assumptions rather than computational ones. Because each approach has a distinct failure mode, a growing body of work combines them so that a session key remains secure if either component holds. We first establish the deployed baseline, hybrid classical-plus-PQC key exchange, against which QKD-augmented designs must be judged. We then propose a taxonomy of hybrid PQC-QKD designs along eight dimensions (integration layer, combiner construction, authentication scope and source, QKD topology, trusted entities, key-delivery interface, classical component, and evidence type), and use it to compare published protocols, including Muckle and its successors, MACsec and TLS integrations, and IKEv2 mechanisms standardised in RFCs 8784, 9370, and 9867. We examine key-combiner constructions in light of the KEM-combiner literature and of NIST, ETSI, and ITU-T guidance, and map the security arguments for hybrid PQC-QKD protocols onto a common grid of security model, assumptions, permitted compromise, established guarantee, and limitations. We synthesise deployment evidence from metropolitan and national QKD networks, distinguishing the few testbeds that have run PQC in the key path from pure-QKD deployments, and summarise the standards and policy landscape, including the positions of national security agencies that recommend against relying on QKD alone. We close with open problems: authentication of the QKD classical channel, shared-infrastructure threat models, device certification versus formal proofs, formal verification of hybrid handshakes, and the lack of peer-reviewed deployment economics.

We study online task assignment with reusable resources. This problem arises in peer reviewer assignment, skill-based call routing, rideshare matching, emergency dispatch, and crowdsourcing: Tasks arrive according to a geometric process and must be immediately assigned to servers that become temporarily busy, yielding rewards that depend on the task-server pair. The goal is to maximize long-run average reward. We cast the problem as a contextual restless multi-armed bandit (CRMAB). We derive an occupancy-measure LP (OCC-LP) that upper-bounds the reward of any work-conserving policy. From its dual we obtain Lag, a closed-form index policy. We introduce a scaling regime in which the number of copies of each server and the number of arrivals per step grow together. This regime yields the first deterministic fluid limit for a contextual bandit. In it, Lag is asymptotically optimal. Finally, we evaluate our algorithms on peer review datasets and call center routing. The Lagrangian index achieves 83-99% of the LP upper bound across all load levels. Moreover, the Lagrangian index policy significantly outperforms the natural greedy assignment scheme, improving reward by up to 12.3% on journal data and 36% on structured synthetic instances.

We study reversible Higgs bundles on a compact Riemann surface $\Sigma$ of genus $g\geq 2$, that is, $G$-Higgs bundles $(E,\Phi)$ with $(E,\Phi)\cong(E,-\Phi)$. For a complex reductive group $G$ realized as a self-adjoint subgroup of $GL(n,\mathbb{C})$, we show that if $(E,\Phi)$ corresponds to a representation $\rho$ of $\pi_1(\Sigma)$, then $(E,-\Phi)$ corresponds to $(\overline{\rho}^{\,t})^{-1}$. As a consequence, the moduli space of reversible Higgs bundles is the fixed locus of an involution of the character variety and carries the structure of a real affine algebraic set. For a connected semisimple group and every topological type, we prove that the reversible locus in the smooth moduli space is a $(B,A,A)$-brane. We then describe reversibility through spectral data: in even rank, for a smooth spectral curve, reversible Higgs bundles correspond to line bundles invariant under the involution $\eta\mapsto-\eta$; in odd rank, under natural genericity assumptions for which the spectral curve is nodal, we give a gluing criterion together with an explicit polystability condition. Finally, for $SL(2n,\mathbb{C})$ we construct a relative Fourier–Mukai transform over the regular locus of the Hitchin base and prove that it maps the structure sheaf of the reversible locus to an object whose support is exactly the regular degree-zero $PSp(2n,\mathbb{C})$-locus in the Langlands dual $PGL(2n,\mathbb{C})$-moduli space. This support is of type $(B,B,B)$.

Motivated by the standard 2-dimensional 3-segment snake model, leading to the famous $(2,3,5)$ geometry, we discuss a series of 3-dimensional models of similar snakes. In particular, we seek for rules allowing to steer them exclusively by the two joints, which expects 4-dimensional control freedom. This leads to interesting $(4,9)$ and $(4,10)$ models. Finally, we discuss modifications of the previous snakes which lead to $(4,8)$ models. All of the models open further questions and should be a good baseline for further research.

Pluralistic alignment needs to accommodate diverse preferences under limited human feedback. Using AI simulators to label preference comparisons reduces annotation costs, but variation in simulation accuracy across individuals can leave some people's preferences poorly represented in training. We study how to allocate human feedback across people to construct training data for pluralistic alignment, and how this allocation affects the trained policy. To make this allocation explicit and controllable, we define a social welfare objective over individual supervision utilities that measure label quality. Each query's marginal welfare gain depends on its expected supervision improvement, the recipient's current utility, and application-designated priority. Welfare-Aware Preference Acquisition (WAPA) uses auxiliary reference models to estimate query gains and individual utilities, then selects queries by estimated marginal welfare gain while updating recipient utilities. We prove exact optimality of the acquisition algorithm for the estimated welfare objective under a fixed query budget and model predictions. Experiments on six Roleplay collections and four OpinionQA regions demonstrate improved supervision and controllable allocation tradeoffs. Policy training on four Roleplay collections improves preference prediction for both seen and unseen individuals. When prioritizing a designated group and poorly represented individuals, acquiring human labels for 20% of comparisons recovers 75-86% of the average predictive utility gap between policies trained entirely on AI-generated labels and entirely on ground-truth labels.

Quantum error correction (QEC) can strongly suppress noise, but predicting whether a computation succeeds requires understanding not only how much effective noise persists after QEC, but what form that noise takes. To thoroughly characterize this noise, we introduce logical gate set tomography (“logical GST”). Logical GST produces detailed representations of noisy logical operations that can be used to derive higher-level metrics, such as gate fidelity and diamond distance, and to test for non-Markovian and coherent logical errors. We use logical GST to obtain the first full, self-consistent characterization of a universal single-qubit logical gate set, implemented in the seven-qubit Steane code using Quantinuum System Model H1, a 20-qubit trapped-ion quantum processor. In the experimental data from H1, logical GST identified predominantly logical Pauli-stochastic errors, found no evidence of logical non-Markovianity, and detected weak potential signatures of coherent logical error. By contrast, higher-shot datasets generated with Quantinuum's detailed system emulator, identify mechanisms for weak but statistically significant logical non-Markovianity and strong evidence of coherent logical error. Finally, we use logical GST to perform error attribution with the system emulator, revealing how specific physical error sources shape logical error mechanisms.

Low-rank adaptation (LoRA) applies every rank-one component of its update to every token, and mixture-of-experts variants replace this with a learned router that activates a fixed number of components. We argue that the number of components a token needs is a detection question with a classical answer. A token's responses to the rank-one components of a trained LoRA consist of a few large values on a Gaussian noise floor whose width matches that of an untrained adapter, and the floor carries cross-talk rather than adaptation. We introduce SQUELCH, a router-free mixture of rank-one experts that (i) soft-thresholds each token's responses at a multiple of its own robust noise level, so the active rank varies per token; (ii) chooses that multiple for each module by Stein's unbiased risk estimate, so no rank budget is specified; and (iii) replaces the resulting marginal keep-or-drop decisions by one explaining-away step on the value Gram matrix, which suppresses redundant co-active components and recruits complementary ones. SQUELCH adds no trainable parameters to LoRA. We prove that the squelch is within a logarithmic factor of the oracle input-dependent allocation, that the selected threshold is risk-consistent, and that explaining away is a descent step toward the joint sparse code whose error is controlled by the coherence of co-active components. On commonsense reasoning, knowledge, math, code, and GLUE with LLaMA-3.1-8B, Qwen3, and DeBERTaV3, SQUELCH improves over LoRA by 2.2–3.2 points and over the strongest mixture baseline by 0.8–1.2 points while activating 8.6 of 32 components per token on average.

Coronal mass ejections (CMEs) are large expulsions of plasma and magnetic fields from the solar corona into the heliosphere. Coronal shock waves are driven by CMEs, and accurate Solar Energetic Particles (SEP) forecasting requires knowing when and where shocks form, how fast they propagate, and how they interact with the coronal magnetic field. By measuring the speed, geometry and temporal evolution of coronal bright fronts (CBFs), one can create more realistic predictions of SEP onset and intensity at Earth, thus velocity-field estimation techniques supply these critical parameters. High-resolution velocity field maps can provide even deeper insight. Instead of yielding only an overall speed of the front, techniques like FLCT and optical flow can show the spatial distribution of velocities across the entire wave dome or front. These analyses reveal which parts of the front are moving fastest and how the front might be deforming as it expands. Here we discuss a set of practical parameters for Lucas–Kanade, Horn–Schunck, FLCT, and Farneback algorithms, arrived at by combining analytic reasoning with empirical measurements on a single event—and demonstrate the potential of these techniques for on-disk shock-wave propagation during the X1.5-class event that took place 10 May 2022, processed by the wavelet-based Wavetrack technique. A better understanding of how to fine-tune these parameters could improve the precision of algorithms running in pipelines, and is especially valuable where per-event hand-tuning is impractical.

The exact local exchange-energy density, obtained from the Slater potential of the Hartree-Fock exchange hole, is a central ingredient of Becke's B05 and B13 functionals and of many other local hybrids. It is usually evaluated in a Gaussian basis, so the local quantity retains a pointwise basis-set incompleteness error. We describe here a basis-set-free numerical code that solves the Kohn-Sham problem on real-space grids for atoms, diatomics, and polyatomics, and returns the exact local exchange-energy density together with the density, its derivatives, the kinetic-energy density, and the reverse Becke-Roussel hole quantities. Reconstructed from the published record of Becke's non-public NUMOL code, AXEL reproduces numerical atomic references through krypton, scalar-relativistic uranium orbital energies, and Becke's diatomic spectroscopic constants. We then extend the original local-density numerical framework to generalized-gradient and hybrid functionals. Comparisons with the independent adaptive-multiwavelet code MRChem give sub-millihartree agreement for Hartree-Fock and PBE0. The same calculations agree with large Gaussian-basis references to a similar level of accuracy. Atom-centred eXact Electronic Limit (AXEL) is intended as both a basis-set-free numerical reference and a platform for exchange-hole-based density-functional development.

Understanding how galaxies populate and grow within dark matter haloes is essential for linking the observed evolution of galaxies to the hierarchical growth of structure. We present a galaxy clustering analysis using optical and near-infrared data from the UltraVISTA and VIDEO surveys covering $9.3$ deg$^2$ across the COSMOS, XMM-LSS and ECDF-S fields, $\sim 5\times$ larger than previous comparable studies, providing sufficient volume to determine the halo occupation of the most massive galaxies to $z\sim 4$. We measure the angular two-point correlation function for stellar-mass-selected galaxies over $0.25<z<4$ and by combining the three fields, we reduce the cosmic variance by a factor of $\sim2$ compared to the individual fields. The stellar-to-halo mass ratio peaks at a halo mass of $M_{h}\approx10^{12} {\rm M}_{\odot}$ and is a maximum at $z\sim 1$. For the most massive galaxies, the peak halo mass ($M_{h}^{\rm peak}$) increases from $z=1\rightarrow 0$, while the peak stellar mass in such haloes remains approximately constant, indicating that at $z<1$ massive galaxy formation becomes less efficient. We also find a steady increase in the satellite fraction since $z\sim 1$, from which we infer that massive halos continue to grow hierarchically while their central galaxies experience suppressed stellar-mass growth. We suggest this is due to the central massive galaxies having their growth controlled by feedback from active galactic nuclei feedback, likely from low-power stochastic radio emission.

A central hypothesis in aging biology proposes that the gradual accumulation of stochastic molecular damage drives transcriptional deregulation (TD), ultimately leading to functional decline. However, existing single-cell analyses largely rely on static expression measurements and provide limited resolution of how regulatory responses change at the level of individual cells. Here, we leverage DKO, a deep learning framework that predicts cell-specific virtual gene perturbation effects from unperturbed transcriptomes, capturing cell-to-cell differences in regulatory context without requiring a reconstructed gene regulatory network or matched perturbation data. Applying DKO to single-cell transcriptomic datasets from humans, mice, Drosophila, and C. elegans revealed that the transcriptional consequences of perturbing the same gene became increasingly heterogeneous across cells with age, indicating that gene regulatory responses become less consistent and more dependent on cellular context. Furthermore, we find that cells with higher somatic mutation burdens have greater knockout-response variability, linking accumulated molecular damage to regulatory inconsistency. Within the same cellular populations, higher impact variability was also associated with higher senescence scores, linking transcriptional dysregulation to senescence-associated states. Our findings identify a progressive loss of regulatory robustness at both the gene and cell levels as a conserved feature of aging, linking stochastic molecular damage to increasing heterogeneity in cellular regulatory responses.

Science fiction is widely credited with unsettling gender. This article asks whether that holds at the linguistic level, in how readily gendered reference gives way to its counterpart, a neutral term, or other social category, and whether it separates the genre from general fiction. It does so by masking man/men, woman/women, and person/people across same-period science fiction and general fiction corpora and classifying RoBERTa's five likeliest substitutes by gender. Retention and displacement are permutation-tested over five resampling rounds, and the underlying sentences are read closely. Both corpora cross the gender binary in both directions, but unequally: where woman is masked, the model predicts a masculine term about twice as often as it predicts a feminine term where man is masked. Each mismatch marks a statistical surprise, a word the corpus did not expect there. Against this shared pattern, science fiction shows greater gender permeability than the general fiction corpus, placing man in feminine contexts more often than general fiction while the reverse rate holds steady. Its degendering tendency is stronger too, drawing both gendered terms toward the neutral. Beyond the binary, a masked woman also draws substitutes naming a social category and no gender far more often than a masked man does. In measuring the gender permeability of man, woman, and person in linguistic distribution and testing whether genre changes it, the study gives science fiction studies linguistic evidence for thematic claims and computational literary studies a pipeline applicable to other category vocabularies, such as race and class.

Chance-constrained stochastic programs (CCSPs) require solutions that satisfy uncertain constraints with high probability. Standard approaches either embed scenarios directly into optimization by considering sample average approximation (SAA)-based reformulations, growing prohibitively expensive and often failing to find feasible solutions; or employ conservative reformulations that guarantee feasibility but sacrifice objective quality by design. We propose ACCESS (Adaptive Chance-Constrained Exploration via Structured Sampling), a structured sampling framework that decouples optimization from feasibility evaluation. First, high-quality candidate solutions are generated by solving many inexpensive single-scenario subproblems, avoiding sampling directly in the constrained high-dimensional solution space. Second, feasibility is assessed separately via large-scale batched simulation over a massive number of scenarios, which is naturally amenable to GPU parallelism. That reduces the search from a high-dimensional solution space to a two-dimensional parameter space controlling conservatism and scenario diversity; while concentrating on candidates near the feasibility boundary, where high-quality feasible solutions lie. We also introduce a variant, ACCESS($\delta$), which provides certified feasibility guarantees with probability $1-\delta$. Experiments across three benchmark classes and five scenario distributions show that ACCESS matches or exceeds the best objective found by any method in nearly every setting, while ACCESS($\delta$) is empirically feasible on every instance. The conservative reformulations remain several percent worse in objective, and SAA is frequently infeasible while taking an order of magnitude longer on the harder benchmarks.

Sparse attention reduces the KV-cache traffic that dominates long-context LLM decoding, but practical methods largely operate at coarse granularity. Token-level selection offers finer control, yet presents two challenges: low-dimensional scoring may lose query–key structure needed for accurate ranking, while gathering scattered selected tokens can offset the resulting latency benefit. We introduce Joint Subspace Sparse Attention (JSSA) to address both challenges. For selection quality, JSSA learns a shared low-dimensional query–key subspace by minimizing token-score preservation error. We upper-bound the resulting non-convex objective with a tractable marginal objective that admits a closed-form spectral solution; direct refinement of the exact objective provides no measurable improvement in the tested settings. For execution efficiency, we develop a gather-free token-level indexed-attention kernel that supports distinct per-KV-head selections under the standard grouped-query attention (GQA) KV-cache layout. Across standard GQA, hybrid sliding-window/global attention, and Multi-head Latent Attention (MLA) models, JSSA consistently remains close to full-KV accuracy while avoiding the substantial generation-length inflation and architecture-dependent failures observed with competing methods. On an NVIDIA B200 GPU, JSSA achieves up to $6.43\times$ the decoding throughput of full-KV attention at 128K context. JSSA achieves reasoning accuracy comparable to model-native DSA without model training and reaches 96–99% of its decoding throughput even at 700B scale large models.

Modern image generation largely relies on specialized generative models that leave the process of image construction implicit in learned parameters. We explore an alternative: general-purpose coding agents that construct images through executable programs. Our study of zero-shot drawing on AgentDrawBench, a 200-prompt benchmark, reveals a distinctive capability profile. Agents excel at preserving structure and satisfying precise visual constraints, achieving competitive results in domains where compositional accuracy matters more than photorealistic appearance, while remaining limited in perceptual richness. Their visual expressiveness varies with the drawing interface, and a qualitative study on reconstruction shows how procedural skills extracted from human demonstrations support different artistic media. We then examine code as an external visual language of thought: a structured, executable representation that agents can construct and inspect. We test it in visual question answering task and find reconstructed visual code retains substantial information for question answering and, in some domains, supports higher accuracy than the original image. These abilities raise a further question: do agents draw on a shared internal representation of scene content beyond a particular drawing program? We explore an internal visual language-of-thought hypothesis, in which a shared, compositionally structured representation supports visual operations across drawing languages. Cross-harness drawings and images assembled from independent single-pixel queries exhibit related layouts and tones. Agents also maintain self-generated room arrangements across viewpoints after visible drawings are removed. These results demonstrate that the coding agent would maintain a consistent and coherent world inside its imagine even before the harness gets involved and the generation starts.