Human-Computer Interaction
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 14 September
Designing social media experiences that support well-being requires understanding when, how, and for whom use matters. Screen-time totals omit content and context, and connecting these with behavior and experience requires coordinating measurements across timescales. We introduce the CAST framework to connect measurement choices with person-specific models of exposure, behavior, physiology, and experience. Its dimensions specify where observations occur, how they are obtained, what they measure, and at what temporal resolution. Responses to interventions, such as whether to proceed after an app-opening pause, enter as behavioral measurements. We propose four synchronized measurement modules linking mobile and wearable data with self-reports and intervention responses. A synthetic demonstration with 120 simulated participants over 28 days illustrates how daily aggregation can obscure opposing effects of different activities under specified generating assumptions. The framework guides selection of measures and outcomes for evaluating social media interfaces and interventions.
Information Theory
A Hardy-Space Proof of the Filter-Only Gaussian Feedback-Capacity Formula
The feedback capacity of power-constrained channels with additive stationary Gaussian noise was formulated by Kim in [Kim, 2010] as an infinite-dimensional optimization over the spectrum of an independent stationary Gaussian component and a strictly causal feedback filter. Kim further asserted that the independent stationary component can be omitted. However, Derpich and Østergaard [Derpich and Østergaard, 2022] identified a gap in the original proof, leaving the original derivation of the filter-only capacity formula incomplete. In this paper, we establish the filter-only formula at the level of capacity suprema for noise spectra satisfying the Paley-Wiener condition. In particular, we show that polynomial feedback filters suffice in the capacity supremum, without a positive lower bound on the noise spectrum or a rationality assumption. Under a positive essential lower bound, a complementary approximation preserves the output spectrum and rate exactly, while its input power converges to that of the original pair. An interleaved AR(1) example connects these results with the SK(2) coding scheme. The argument does not assert attainment by a single filter.
Read the full abstract
Robotics
MessyMem: Learning-from-Doing Memory for Mobile Manipulation
Mobile manipulators deployed across many rooms and visits should improve with experience: after discovering that a cabinet is locked or finding an object in a drawer, the robot should reuse that knowledge rather than start each task from scratch. Yet today's robots often treat each task as new: compact scene representations omit interaction-derived knowledge, raw video histories are difficult to query, and VLM planners reason at inference time without persistently updating what the robot knows. We present MessyMem, a persistent memory system that enables mobile manipulators to learn from experience and reuse that knowledge across future tasks. It maintains a spatially grounded 3D scene graph of objects and locations, augments it with properties and outcomes learned through interaction, and links visual observations for fine-grained recall. We evaluate MessyMem in simulation and on a real mobile manipulator. In a continuous 25-task simulation spanning over 3 hours, MessyMem achieves 80.0% task progress, outperforming the strongest ablation by 14.8 percentage points and the strongest external baseline by 28.9 points, while retrieving task-relevant evidence from thousands of stored keyframes and over an hour into the past.
Read the full abstract
Computation and Language
Disentangling Representation Evolution in Transformers through Directional Decomposition
Transformer representations evolve through learned additive transformations that either preserve their current direction or redirect it. We study this evolution as a functional geometry, decomposing learned updates into parallel and perpendicular components. Across pretrained models, we find substantial parallel components beyond the residual identity path. We then apply the decomposition in two spaces: to attention and MLP updates relative to the hidden state, and to attention value aggregation relative to the current token's value. Targeted edits reveal a strongly space-dependent asymmetry: exclude-self value-space parallel manipulation is markedly more robust than residual-space and perpendicular counterparts, preserving the direct self message while scaling only the non-self aggregate. The same decomposition gives a component-resolved description of compression-induced update error: perpendicular error separates compression methods more clearly than parallel error. Extensive experiments further demonstrate that full-aggregate parallel suppression during from-scratch pretraining lowers validation-loss trajectories and improves downstream averages, with the value-space variant strongest. Together, these results connect representation geometry to editing robustness, compression diagnosis, and training-time intervention. Code is available in the \href{https://github.com/Shwai-He/Transformer-Geometry}{project repository}.
Read the full abstract
Optics
Critical quantum metrology in the output of an optical parametric oscillator
Many quantum sensors infer parameters from continuously emitted fields, yet in the presence of unmonitored loss, it is difficult to determine how much critical enhancement survives in the accessible output. In this work, we calculate the full frequency-resolved output quantum Fisher information (QFI) of a stationary vacuum-seeded optical parametric oscillator for cavity-detuning estimation. We show that, for any fixed nonzero unmonitored loss, the QFI of correlated sideband pairs grows quadratically with the mean intracavity photon number near threshold, but only within a spectral window that narrows inversely with that number; consequently, the spectrum-integrated output QFI rate scales asymptotically only linearly. We identify a far-detuned, near-threshold regime in which the monitored output alone asymptotically approaches the loss-imposed upper bound on the joint cavity-output QFI. We further determine the operating points that maximize the output QFI rate per intracavity photon and introduce a frequency-resolved homodyne strategy that maximizes the Fisher information obtainable from a single record. Our results establish an operational connection between critical enhancement and metrological information accessible in the output field, while providing concrete guidance for the design and readout of dissipative parametric sensors.
Read the full abstract
Computation and Language
Discovery Foundation Models: Toward Open-Ended Discovery Intelligence
Foundation models have progressed from learning and reasoning over existing knowledge, to increasingly learning through action, tool use, and outcome feedback. We argue that the next frontier is a further transition: from solving and acting within problems specified by humans to participating in the process by which new problems, representations, explanations, and knowledge are created. We refer to this capability as Discovery Intelligence. We formulate Discovery Foundation Models (DFMs) as general-purpose model systems for open-ended discovery. A DFM operates over a revisable research state and supports seven coupled capabilities spanning problem discovery, formulation, representation construction, hypothesis formation, intervention, evidence-grounded revision, and continual discovery improvement. We instantiate this framework with Zetema, which couples explicit research-state dynamics, verification and experimental gating, external grounding, and cross-task Discovery Skill evolution. We further ground the framework with GALILEO, a real therapeutic-discovery system in which Dry-Lab reasoning, robotic and hands-on Wet-Lab experimentation, external biological evidence, and iterative hypothesis and design revision form a closed physical discovery loop. We then formulate a unified approach to capability formation and process-centered evaluation, enabling discovery behavior to be trained, improved, and measured beyond final-answer performance. Together, these components establish discovery as a learnable, executable, and evaluable capability of foundation-model systems. We view this shift as a broader progression in intelligence scaling: from learning over existing knowledge, to learning from action outcomes, and ultimately to participating in the construction, testing, and revision of the structures through which new knowledge is discovered. Code: https://github.com/Gen-Verse/DFM-Plans
Read the full abstract
Computation and Language
Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States
As language models become more capable, long-term collaboration in learning, reasoning, and decision-making calls for a deeper understanding of the people they serve. Yet training such human-aware language models faces a fundamental supervision gap because current datasets for LLM assistant training contain few if any well-informed responses explicitly grounded in users' unspoken beliefs and goals. Scaling such supervision is inherently constrained, as users' underlying states are not directly observable. We thus propose the Mind2Dialogue framework to mitigate this gap by simulating users' mental states and turning them into privileged supervision for human-aware training. Specifically, we first propose a psychology-guided simulator that preserves personal characteristics while updating mental states through interaction to generate coherent conversations. The key idea is to enforce a shared evolving mental state that drives user behavior and guides an Oracle assistant's responses. Our privileged distillation then trains models on the Oracle's well-informed responses to assist users without direct access to their mental states at deployment. Moreover, we propose to evaluate human-aware learning by combining personalization and theory of mind, examining how models understand people and act on that understanding. Training on the full Mind2Dialogue corpus improves every reported personalization metric over the corresponding Qwen, Llama, and OLMo instruction-tuned baselines, including gains of 26.6 to 40.9 percentage points in preference-following generation. The gains extend to belief and action reasoning on Qwen and Llama, beyond personalized assistance. Looking forward, Mind2Dialogue makes user simulation a foundation for genuine AI collaborators that understand beliefs and intentions behind people's words and support their long-term goals across education, work, and everyday life.
Read the full abstract
Representation Theory
Fusion-stable endosplit $p$-permutation resolutions
Let $k$ be a field of characteristic $p>0$, $\mathcal{F}$ a saturated fusion system over a finite $p$-group $P$, and $V$ an indecomposable capped endopermutation $kP$-module. Let $D_k^\Omega(P)$ be the subgroup of the Dade group $D_k(P)$ generated by all the relative syzygies. It is known that $V$ has an endosplit $p$-permutation resolution if and only if the Dade class $[V]$ belongs to $D_k^\Omega(P)$. We show that the resolution can be chosen to be $\mathcal{F}$-stable if and only if $V$ is $\mathcal{F}$-stable. As an application, we prove the following folklore result: if two blocks of finite groups are Morita equivalent via a bimodule with an endopermutation $kP$-source $V$ such that $[V]\in D_k^\Omega(P)$, then these two blocks are splendidly Rickard equivalent.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Less isn't more: Cosmological bounds on the neutrino masses are robust to changes in the neutrino abundance
We investigate how neutrino-mass constraints from cosmology depend on the assumed thermal history of the universe. Photon injection after neutrino decoupling would decrease the neutrino abundance inferred from the temperature of the cosmic microwave background (CMB), potentially loosening the upper limit on their masses. We first evaluate how the cosmological neutrino-mass bound is altered by the decay of massive particles into photons and dark radiation after Big Bang nucleosynthesis. To test the robustness of this constraint more generally, we also assess the impact of varying the temperature of the cosmic neutrino background without assuming a specific physical mechanism. We perform MCMC analyses of both frameworks with primary CMB observations from Planck, CMB lensing measurements from Planck and ACT, and baryon acoustic oscillation data from DESI. For the degenerate mass ordering, the $95\%$ credible limit tightens from $\sum m_\nu<0.0691$ eV in a standard thermal history to $\sum m_\nu<0.0652$ eV if the massive particles decay only into photons, while the injection of dark radiation in addition to photons slightly relaxes this limit to $\sum m_\nu<0.0710$ eV. The same pattern holds for the normal and inverted orderings, and the decay scenario shifts the bound on the sum of the neutrino masses by at most $0.004$ eV for fixed mass splittings. Allowing model-agnostic changes in the neutrino-to-photon ratio yields a $95\%$ credible limit of $\sum m_\nu<0.0724$ eV for the degenerate ordering, indicating that the stringency of our neutrino-mass bounds is not driven by constraints on the decay scenario. We find that the neutrino temperature and the sum of the neutrino masses are positively correlated, which implies that reducing the pre-recombination radiation density will only worsen the emerging tension between cosmological bounds on the neutrino masses and the measured mass splittings.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Constraining quantum-gravity predictions for evolving dark energy
We confront a class of dark-energy equations of state emerging from group field theory (GFT) quantum gravity with DESI Data Release 2 baryon acoustic oscillations and Pantheon+ type-Ia supernovae. We introduce sampling parametrisations that replace microscopic initial-condition parameters by combinations more directly measured by background probes. The GFT solutions separate into logarithmic, power-law and oscillatory branches, determined by the microscopic interaction parameter $m$. The logarithmic branch is constrained to lie extremely close to a cosmological constant, while the power-law branch permits a small phantom deviation. Without perturbative-theory priors, oscillatory solutions can reproduce the mild preference of the distance data for a dip in $w(z)$ near $z\simeq0.5$–$1$. Profile-likelihood constraints favour $m\sim-2$ and $m\sim-5$ from BAO and supernovae, shifting towards $m\sim-3.5$ when CMB information is included. Conservative perturbative priors strongly suppress these deviations from $\Lambda$CDM. The quantum-gravity scale $z_q$, related to the average number of quantum gravity atoms, remains unconstrained, although its role in the time evolution makes higher-redshift observations a promising route to probing it. We further find that strong projection effects highlight the importance of performing likelihood profiling alongside our marginal posterior constraints. Our results provide a first direct test of GFT-motivated dynamical dark energy and demonstrate the potential for cosmological observations to inform quantum-gravity model building.
Read the full abstract
Plasma Physics
Direct observation of electron shedding from a laser-plasma accelerator
Laser-plasma accelerators have demonstrated the ability to produce ultrashort relativistic electron bunches with peak currents suitable for compact light sources, ultrafast diffraction, and strong-field studies. However, their performance critically depends on preserving the longitudinal phase-space density of the beam as it exits the plasma accelerator. Here, we report the first direct observation of a previously unresolved process in which a highly charged electron bunch undergoes significant longitudinal expansion and progressively loses electrons during extraction from a laser-driven wakefield accelerator, a phenomenon we refer to as electron shedding. Using femtosecond relativistic electron microscopy, we tracked the evolution of the beam far beyond the accelerator exit and observed the bunch stretching over many plasma wavelengths before shedding electrons during propagation. Particle-in-cell simulations reproduce the observed behavior and reveal that it originates from a combination of effects when a high-charge-density beam exits the accelerator. These dynamics redistribute energy from the beam head into a low-energy tail, thereby reducing the useful peak charge density and ultimately decreasing the efficiency of the laser-plasma accelerator. Our results provide new insight into beam extraction and phase-space evolution in laser plasma accelerators and highlight the importance of controlling these collective effects for future applications.
Read the full abstract
High Energy Astrophysical Phenomena
A Standard Ultraviolet Continuum Can Hide Supercritical Accretion in GN-z11
The ultraviolet continuum from a high-redshift accreting black hole is routinely used to infer its mass. GN-z11 offers a sharp test of whether that inference is secure. A thin-disk fit to its continuum gives an Eddington mass of $M_{\rm E}=1.12\times10^{7}\,\Msun$, an order of magnitude above the $\log(\Mbh/\Msun)=6.2\pm0.3$ inferred from broad N~iv. We show that this apparent tension is not imposed by the ultraviolet slope alone. We derive a closed-form criterion comparing the radius where observed photons are produced with the radius where a supercritical flow departs from efficient solution. For GN-z11, the criterion places the supercritical transition inside the ultraviolet-emitting region. Composite disks retaining the outer solution confirm this numerically, requiring $\Mdot_{\rm out}=3.1$–$4.4\,\Msun\,\mathrm{yr^{-1}}$ and placing $\Rsph=(1.2$–$1.8)\times10^{3}\,\rg$ inside $\Ruv=(2.6$–$2.9)\times10^{3}\,\rg$. Replacing the inner $\sim10^{3}\rg$ by a supercritical flow shifts the fitted slope by only $\Delta\betaUV=+0.08$ to $+0.11$, comparable to uncertainties from standard spectral modelling. Reversing the radial ordering requires the effective transition to move outward by a factor $1.7$–$2.1$. Within our composite models, the measured slope constrains black-hole mass only to around $10^{6}\,\Msun$, an order of magnitude below the continuum Eddington mass. Continuum Eddington masses should therefore be treated as model-dependent bounds that assume global thin-disk efficiency. The same hierarchy appears in three additional JWST sources under the disk-dominated interpretation, where separation masses are only $4$–$6$% of continuum Eddington masses. This distinction can substantially weaken the seed-mass and early-growth demands inferred from ultraviolet continua at cosmic dawn for the first massive black holes.
Read the full abstract
Mesoscale and Nanoscale Physics
Shot noise signatures of Majorana-assisted transport in an Aharonov-Bohm interferometer
In this work, we investigate the finite-bias charge current, shot noise, and Fano factor in an Aharonov-Bohm interferometer composed of two parallel quantum dots, each side-coupled to a topological superconducting nanowire represented by an effective Kitaev chain hosting Majorana bound states at its ends. The transport properties are calculated using the nonequilibrium Green's function formalism within the equation-of-motion approach. Our results show that coupling to the Majorana bound states significantly modifies the interference pattern, producing distinctive signatures in the current, shot noise, and Fano factor. These signatures are particularly pronounced at the half-flux condition, where destructive interference suppresses the conventional transport background and enhances the visibility of Majorana-assisted transport. These findings demonstrate that finite-bias shot-noise spectroscopy provides a sensitive indirect probe of Majorana bound states in hybrid superconducting-nanowire interferometers.
Read the full abstract
Algebraic Geometry
Wasserstein Metric Normalization of Projective Cycle Spaces
Let $X$ be an irreducible reduced projective subvariety of a Chow variety. On a generic smooth parameter locus, normal motions of resolved components define a quadratic Wasserstein metric, even when the cycles are singular, reducible, or carry multiplicities. Its metric completion is canonically $X^\nu$: resolution fibers collapse, whereas distinct normalization branches over the same Chow cycle remain separated. The completed metric admits an exact ambient action formula. For degree-$d$ hypersurfaces, the inner $W_q$ geometry is compact and induces the projective topology for every $1\le q\le2$. For $d\ge2$, the uniform Hölder exponent $1/d$ is attained and sharp for each such $q$; when $d=1$, the comparison is Lipschitz. Elliptic quartics furnish an explicit model of the boundary branching.
Read the full abstract
Computation and Language
Verifiable by Construction: How Well Do LLMs Substantiate Their Clinical Claims with Verbatim Quotes?
Large language models (LLMs) have been widely adopted for clinical question answering (QA). Current systems can attach citations to their answers, but these often point to broad texts, leaving time-pressed clinicians unable to verify them efficiently. An alternative is to ensure that responses are verifiable by construction: providing fine-grained verbatim quotes from reference material that substantiate claims, so users can verify an answer without opening other documents. In this paper, we evaluate the ability of current models to perform this task end-to-end: from providing citations for every factual claim, to producing verbatim quotes, to ensuring that those quotes fully substantiate the claims. To do so, we build a standardized harness over four clinical practice guidelines and evaluate fourteen LLMs on 222 synthetic clinical questions, measuring each of these stages separately. We find that most models can attach verbatim quotes to over 90% of their claims from prompting alone, apart from some lightweight models such as claude-haiku-4.5. Yet these quotes often fail to substantiate every detail of the claims they accompany. For instance, claude-opus-5 produces verbatim quotes for 98.0% of its claims, but fully substantiates only 37.1%. Our work provides insights into the current capability gap of LLMs in building verifiable clinical QA systems, along with artifacts for future research.
Read the full abstract
Cryptography and Security
Adversarial Testing of Automated Program Repair Agents for Security Vulnerabilities
Software agents with Large Language Models (LLMs) are designed for Automated Program Repair (APR) tasks, raising the possibility that, in the near future, APR agents will fix bugs automatically without much human intervention. Can we trust an APR agent to produce both functionally correct and secure code in such situations? What if attackers target production APR agents with adversarial issues that seem benign but may influence the agents to produce correct but insecure code? In this paper, we took a first step towards answering these questions by conducting an empirical study. First, we created SWEADV, a benchmark of 750 adversarial issue descriptions constructed from 150 repair tasks in SWE-bench Verified. For each repair task, we created five adversarial issue descriptions, one for each attack type: command execution, deserialization, path traversal, denial of service, and weak hashing. Second, we evaluated mini_swe APR agents from three LLM backends on SWEADV: GPT-5-Mini, MiniMax-M2.5, and DeepSeek-R. We found that on average, adversarial issue descriptions can induce malicious behaviors with successful repair in 51.7% of cases. Third, we investigated whether typical detection mechanisms are sufficient to prevent such malicious patches from being accepted. Pre-repair detection with LLM-as-judge on the adversarial issue descriptions resulted in an average detection accuracy of only 62.3%. Post-repair detection on adversarial APR patches using static analysis tools and LLM-as-judge achieved average detection accuracies of only 39.4% and 55.4%, respectively. We conclude that autonomous APR agents cannot be trusted yet in production deployment, given their susceptibility to adversarial attacks.
Read the full abstract
High Energy Astrophysical Phenomena
Can Isotropic Thermal Conduction Heat One of the Hottest Cool Cores?
Thermal conduction can act as a heating mechanism for cool-core clusters, yet the same transport must erode any temperature discontinuity it crosses. Ophiuchus permits both effects to be tested in the same atmosphere. Its temperature rises from about $1$ keV in the innermost core to $9$ keV at $r\sim30$ kpc, making conductive heating especially favorable, while two sharply resolved cold fronts independently constrain the same transport. \textcolor{black}{We construct a steady radial energy balance from deprojected Chandra density and temperature profiles, with XRISM-based limits on turbulent heating and a subsonic inflow contribution.} Closing the balance with conduction requires only sub-Spitzer coefficients — \Ftwentyfive\ at 25 kpc and \Finner\ and \Fouter\ at the inner and outer cold fronts — so conduction can energetically supply the missing heat. Yet the same coefficients, applied normal to the cold-front surfaces, would broaden the fronts to their observed width limits within \Tinner\ and \Touter\ Myr, short compared to the characteristic sloshing timescales that generate and sustain such interfaces. For a representative 200 Myr residence time, the radial conductivity demanded by the energy balance exceeds the cross-front conductivity permitted by front survival by roughly a factor of three, with the required anisotropy exceeding unity in 99.9% and 98.8% of accepted models. The result leaves two broad regimes. Thermal conduction is either subdominant in Ophiuchus or contributes substantially but is strongly anisotropic, with heat transport suppressed across the cold fronts relative to the radial direction. Together they disfavor locally isotropic conduction as a dominant heating mechanism in cool cores generally.
Read the full abstract
Statistical Methodology
Spatially-Indexed Longitudinal Distributional Outcome Regression for Environmental Monitoring
Characterizing longitudinal changes in region-specific distribution of environmental exposures, such as total nitrate (TNO$_3$) concentrations, is critical for understanding localized ecological risks that are otherwise obscured by standard mean-level modeling. However, modeling longitudinal distributional outcomes across spatial regions presents significant methodological challenges. The random objects are spatio-temporally dependent, and mathematical constraints are inherent to distributional representations, such as, monotonicity of quantile functions. To address this, we propose a novel spatially-indexed longitudinal distributional outcome regression model. The distributional coefficients corresponding to the fixed effects of covariates are modeled using Bernstein basis polynomials, while spatio-temporal random effects are flexibly captured via tensor product expansions of splines. We develop a scalable Markov Chain Monte Carlo (MCMC) algorithm to explicitly account for spatial dependencies, and introduce a fast two-stage projected-posterior approach to preserve the monotonicity of the predicted subject-specific quantile functions. Extensive simulation studies demonstrate that the proposed framework achieves superior estimation accuracy and predictive performance compared to standard non-spatial distributional outcome regression. We apply our methodology to predict monthly, site-specific distributions of TNO$_3$ concentrations across the contiguous United States. Accounting for spatial correlation provides substantially lower uncertainty in the estimated distributional effects and improves predictive performance over the non-spatial alternative, offering a robust, interpretable tool for spatio-temporal environmental monitoring.
Read the full abstract
High Energy Astrophysical Phenomena
Validating Inverse Fundamental-Plane IMBH Mass Estimates in the SKA/ngVLA Era
The radio/X-ray/black-hole-mass Fundamental Plane (FP) is widely used to estimate masses of accreting intermediate-mass black hole (IMBH) candidates. Its inverse use is justified only when radio luminosity traces a compact jet core and X-rays arise from the same sub-Eddington hard-state accretion flow. We formulate an accretion-state validity criterion for inverse-FP masses using verified BH X-ray binary radio/X-ray tracks as the empirical reference. Using 424 detections from 21 BHXBs and candidates, we find that both standard and curved tracks are present and the luminosity at which radio/X-ray behavior changes is source dependent. A low X-ray Eddington ratio is therefore not, by itself, sufficient to establish inverse-FP validity. Within a scale-invariant jet/hot-flow framework, track steepening reduces the effective mass leverage of the inversion, while at much lower accretion rates uncertainties in the dominant emission mechanism and compact-core production can invalidate the underlying FP assumptions altogether. We apply the criterion to representative IMBH-relevant systems without fitting a new plane. NGC 4395 is low-Eddington but compact-radio deficient: its VLA-scale and VLBI jet-base luminosities imply formal inverse-FP mass underestimates of about 1.0 and more than 2.8 dex if the deficit is misread as a mass signal. Globular-cluster X-ray populations in NGC 1399 and NGC 4472 are source-identity limited: 71% and 67% lie below $L_{\rm Edd}(10\,M_\odot)$, and the high-luminosity tail alone does not identify an IMBH accretor. The radio non-detection of the $\omega$ Centauri IMBH candidate constrains gas supply, accretion efficiency, and compact-core production rather than excluding the black hole. The resulting test defines which SKA/ngVLA-era radio cores can be converted into FP masses, and which instead constrain gas supply, source identity, or compact-core production.
Read the full abstract
Lattice Field Theory
Continuous Hasenbusch transport towards gauge diffusion with fermions
Incorporating dynamical fermions is a central challenge for diffusion samplers of lattice gauge theories. We propose an analytic pseudofermion sampler based on continuous Hasenbusch transport as a component for gauge diffusion. The construction uses shifted linear solves and a finite-path correction, requiring neither explicit fermion determinant evaluation nor a learned pseudofermion model. We demonstrate the coupling in the two-flavour Schwinger model without neural networks, obtaining corrected physical observables compatible with independent references. We also show why accurate covariance transport can leave large weight fluctuations, tracing them to the backward transition density. This analysis leads to a correction based on the Wilson operator trace that reduces log-weight variance without changing the generated fields or increasing the number of Dirac operator applications. The predicted improvement is verified on previously unused gauge backgrounds. These results provide an analytic option for incorporating fermions in gauge diffusion and a guide to controlling its correction weights.
Read the full abstract
Superconductivity
Stripe-Like Superconducting Enhancement and Coexisting Magnetic Texture in an Infinite Layer Nickelate
Despite their promise as structural and electronic analogs to cuprates, nickelate thin films consistently exhibit broadened superconducting transitions in many experiments that remain poorly understood. Local measurements are ideal for revealing the presence of defects or competing phases, which may cause this transition broadening through a phase-separation. Here, we use scanning SQUID microscopy to investigate local superfluid density and magnetic texture of optimally doped $\mathrm{Nd}_{1-x}\mathrm{Eu}_{x}\mathrm{NiO}_{2}$ (NENO) (x=0.25) grown via molecular beam epitaxy. We observed a weakly-magnetic texture coexisting with superconductivity. Spatially resolved susceptibility imaging reveals a highly non-uniform superconducting state, characterized by a robust stripe-like enhancement pattern appearing near the phase transition. By resolving the mesoscopic landscape of electronic and magnetic inhomogeneities, these findings suggest that a competing magnetic phase is a primary contributor to the unusually broad superconducting transitions of these optimally doped NENO samples. Understanding the uncovered superconducting enhancement may provide a path to raising superconducting temperatures in these materials.
Read the full abstract
Analysis of PDEs
An analysis of aerodynamic properties of delta wings
We consider a sharp-edge delta wing of small aspect ratio $A>0$, which is an example of a 3D airfoil whose aerodynamic properties cannot be modeled using the potential lift}only. An important role is played by a pair of attached vortices, which generate the vortex lift. We review in detail the leading-edge suction analogy, developed by Edward Polhamus (NASA Technical Note D-3767, 1966), which shows that the most accurate description of the vortex lift can be obtained by avoiding any direct analysis of vorticity of the flow around the wing. We also describe a 2D slender wing theory, which was introduced by Jones (NACA report no. 835, 1946) to approximate the potential lift. We show that, neglecting certain two borderline nonintegrability issues and an elliptic regularity issue, the slender wing theory is sufficient to yield accurate prediction of the potential lift as $A\to 0$. We also show that this 2D approximation is necessary to justify the modelling assumptions of the leading-edge suction analogy, so that the slender wing model can also be used to approximate the vortex lift as $A\to 0$. For completeness, we provide additional explanations of the necessary airfoil theory.
Read the full abstract
Dynamical Systems
Dynamical Mordell–Lang Conjecture for Higher-Rank Radially Ramified Skew Products
We establish the dynamical Mordell–Lang conjecture over the complex numbers for a family of radially ramified polynomial skew products in arbitrary base dimension and fiber rank. We assume that one fixed iterate sends the vertical critical locus into the invariant zero section. Our result covers affine-linear bases and, under a degree-gap condition, nonlinear étale polynomial bases. We also establish a uniform Skolem–Mahler–Lech theorem for a fixed nonlinear orbit and a fixed polynomial-coefficient linear recurrence whose trailing coefficient is a nonzero constant. The theorem applies to joint polynomial and rational relations involving the orbit and finitely many successive terms of any solution. A single eventual period works for every such solution and relation, although the finite exceptional set may depend on the relation. Our proof combines controlled nonarchimedean realizations with an orbitwise transfer theorem for superattracting line-bundle dynamics.
Read the full abstract
Networking
Understanding the oversubscription behaviour of DragonFly+ networks
The Max-Host Dragonfly+ topology's original paper proves that there is a 2:1 worst-case oversubscription ratio in expectation for the permutation traffic pattern. We show that the proof only covers a subset of permutation patterns, specifically those in which all host pairs are in different groups, and for any receiver group there are at least two sending hosts in distinct groups. The proof contains two mistakes that cancel out to produce the correct result. Furthermore, the proof implicitly uses an important observation without a backing argument: traffic leaving an indirect group, now forced to follow min-cost paths to receiver groups, can still be split across almost all outgoing global links, in contrast with splitting traffic from a sender group that can naturally select any outgoing group. We prove that the 2:1 ratio holds in expectation for a larger set of patterns, including any permutation. We only need to know that any host sends and receives at most line rate traffic. This constraint can be altered to obtain approximations for expected FCT bounds on any pattern. We also attack the oversubscription problem without the expectation assumption, and find bounds that hold with high probability on permutations for small switch radixes. While in expectation the global layer wiring doesn't affect the oversubscription rate, we experimentally find that it matters in the general case. We find topologies that obtain visible speedups against the default global wiring from the original paper, under $7\%$ for radixes at most $8$.
Read the full abstract
Representation Theory
The coherent KLR sheaf
We construct the KLR algebra (in type $A$) as affine sections of a coherent sheaf $\mathcal{E}$ on a product of projective spaces $\mathbb{P}^n$. It was shown by Elias-Qi that the Lie algebra $\mathfrak{sl}_2$ acts on such KLR algebras; our construction geometrically explains this action as stemming from the $\mathfrak{sl}_2$-action on the twisting sheaves $\mathcal{O}(k)$ of $\mathbb{P}^1$ and gives the maximal finite-dimensional $\mathfrak{sl}_2$-submodule of KLR as global sections of $\mathcal{E}$. We furthermore extend the Elias-Qi action to a Witt action. Kleshchev-Loubert-Miemietz showed that the KLR of type $A$ has a graded affine cellular structure; our construction also gives a geometric incarnation of this.
Read the full abstract
Logic in Computer Science
Continuity-First Lexicographic Optimization for Home-Care Resource Allocation
Home-care allocation must balance continuity of care, caregiver overtime, and caregiver-service compatibility. The published formulation of the Home-Care Optimal Resource Allocation Problem (HCORAP) uses a weighted policy (Weighted) to combine these outcomes, allowing compatibility gains to offset poorer continuity or additional overtime. We introduce LEX-COS, a lexicographic objective for HCORAP that first minimizes continuity violations, then overtime, and finally maximizes compatibility. We give an exact MaxSAT implementation and construct HCORAP-LC, a reproducible 48-instance synthetic evaluation suite with limited caregiver capacity. Compared with Weighted, LEX-COS reduces both continuity violations and overtime in 42 of 48 instances, without increasing either measure in any instance, while yielding lower compatibility. In 36 instances, LEX-COS achieves better continuity and lower overtime than every Weighted-optimal allocation. On the separate 48-instance Original benchmark, Gurobi and CPLEX, two leading commercial MIP solvers, resolve every instance under both policies and are substantially faster than EvalMaxSAT. EvalMaxSAT also resolves the full benchmark under LEX-COS. Within EvalMaxSAT, replacing sorting networks with Totalizer improves runtime under Weighted, with no clear gain under LEX-COS. These results characterize the effects of giving continuity and overtime explicit priority in home-care allocation.
Read the full abstract
High Energy Physics (Theory)
Direct-to-Symbol Integration from Landau Analysis
We propose a Landau-analysis-inspired strategy for directly computing the symbol of integrals that evaluate to multiple polylogarithms in twisted cohomology. The central idea is to identify all ways in which singular hypersurfaces, twisted hypersurfaces, and integration boundaries can interact when external parameters are varied, giving rise to logarithmic or algebraic branch points. By tracking how an integral is modified when analytically continued around each of its branch points, we can recursively construct its symbol. We illustrate this approach by outlining an efficient algorithm for computing the symbol of finite integrals over twisted hyperplane arrangements, when they are expanded around special values of the twist parameter. This algorithm can be used to compute the (transcendental) integrands of cosmological correlators in conformally coupled theories to any order in the twist expansion.
Read the full abstract
Algebraic Geometry
Explicit invariants of the Suzuki and Ree groups in their function fields
The Suzuki and Ree curves are two classical Deligne–Lusztig curves. They are maximal over suitable finite fields and their full automorphism groups are the Suzuki and Ree groups. In this paper we determine explicit generators of the fixed fields of these full automorphism groups in the corresponding function fields. Motivated by Dickson's classical study of invariants of finite linear groups, we use the natural projective representations of the two groups. In the Suzuki case the proof uses restricted Dickson invariants, whereas in the Ree case it relies on an invariant bilinear form of the seven-dimensional representation.
Read the full abstract
Machine Learning
Privacy-Aligned Personalized Federated Learning with Compact Adaptation and Variable-Length Gaussian Communication
Record-level differential privacy exposes a structural misalignment in personalized federated learning when client-specific variation is low-dimensional while training repeatedly releases high-dimensional updates. In this paper, we address this misalignment by releasing a private client context once and confining repeated adaptation to a fixed coefficient space. Beyond dimensionality reduction, the factorized generator induces an adaptive optimization geometry that reshapes noisy updates, and controlled ablations show that most of its private-training gain is retained by radial evolution. To further reduce the communication cost, we realize the Gaussian mechanism for coefficient updates directly through variable-length quantization with finite expected code length, so that the quantization error itself serves as the required privacy perturbation rather than extra distortion. Across MNIST and CIFAR-10, our design matches or outperforms full-model private adaptation across privacy budgets and client heterogeneity, while reducing protected uplink by a factor of 2.67 at \(\varepsilon=16\) on CIFAR-10 with comparable future-client accuracy.
Read the full abstract
Algebraic Geometry
The tautological ring of $M_{25}$ is not Gorenstein
We show that the tautological ring of $M_{25}$ is not Gorenstein, disproving a conjecture of Faber from 1993. This also implies that the Faber-Zagier relations are the complete list of tautological relations on $M_{25}$.