Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Saturday 7 November

Strongly Correlated Electrons

Dipolar Exchange and Emergent Electric-Dipole-Carrying Quasiparticles in Electronic Ferroelectrics

Electronic correlations underpin diverse spontaneously ordered states in matter, whose collective excitations emerge as quasiparticles that encode the underlying order. While spin exchange interactions give rise to magnetic order and its quasiparticle excitations known as magnons, their counterparts in electronic ferroelectrics remain poorly understood. Here, we derive an analogous quantum dipolar-pseudospin exchange model for electronic ferroelectricity and identify the associated bosonic quasiparticles that carry intrinsic electric dipole moments. Unlike ferrons in conventional lattice-based ferroelectrics, these dipolar quasiparticles arise entirely from collective electronic dynamics without involving lattice vibrations. Their excitation spectrum exhibits a linear Stark shift under a static electric field, while coupling to an oscillating electric field enables their coherent excitation, leading to colossal, resonantly enhanced second-harmonic generation and optical rectification in the terahertz regime. These dipolar quasiparticles offer promising opportunities for electrically tunable nonlinear optics and open a route toward extending the paradigm of transport in electronic systems beyond conventional charge currents to electric dipole currents.
Read the full abstract

Strongly Correlated Electrons

Quartet ferromagnetism in the small-exchange limit

Rare-earth magnets are often regarded as classical owing to their large angular momentum J, but this picture breaks down when exchange is smaller than the crystalline electric-field (CEF) energy scale, allowing the degeneracy of low-lying CEF manifolds to dictate the emergent quantum phases. Here, we uncover such a regime in the RV6Sn6 (R = rare-earth) kagome metals with strong magnetic anisotropy. Thermodynamic and neutron-diffraction measurements reveal successive axial and tilted ferromagnetic (FM) phases in DyV6Sn6, arising from independent ordering of the axial Mz and transverse Mx dipolar components enabled by a rare quasi-quartet ground state. On the other hand, HoV6Sn6 has a similar tilted magnetic anisotropy, but remains axial down to 0.03 K. This discrepancy can be traced back to the non-Kramers nature of Ho3+ where the CEF lifts the quartet degeneracy, leaving a quasi-triplet ground state. Unlike quartets or higher multiplets, we find that the effective spin-1 triplet does not allow independent dipolar ordering without strong quadrupolar (nematic) interactions that act to break rotational symmetry, revealing an essential multipolar prerequisite for tilted magnetic ground states. These results establish a quartet CEF manifold as the minimal degeneracy required for dipolar-driven independent ordering. Mapping the surrounding CEF parameter space further reveals competing magnetic, non-magnetic singlet, and potential quadrupolar phases, driven by multifold ground state degeneracies. These results establish CEF tuning as a versatile route to unconventional quantum phases across the broader RT6X6 (T = transition metal; X = Sn, Ge) family.
Read the full abstract

All authors Samuel Ogundeji, Tianxiong Han, Tyler J. Slade, Olivia Medina, Joanna M. Blawat, Filip Ronning, John Singleton, Sean M. Thomas, Alimamy F. Bangura, Wei Tian, Eleanor M. Clements, Bing Li, Adam A. Aczel, Paul C. Canfield, Benjamin G. Ueland, Rebecca Flint, Robert J. McQueeney, Aashish Sapkota

Machine Learning

CALFA-KV: Joint Calibration and Fisher-Guided Anchors for 1-Bit KV Cache Compression

1-bit key-value (KV) cache compression can combine quantizer calibration with selective high-precision storage to reduce reconstruction error. We propose Calibration and Fisher-guided Anchors for KV cache compression (CALFA-KV). Offline, we jointly fit a key center, radial normalization parameters, and K/V vector codebooks to match the frozen model's projected attention output. We also estimate channel-wise Fisher-type sensitivities from squared gradients of a reference next-token cross-entropy loss. Online, we use these sensitivities to weight the squared reconstruction errors of each KV state. Here, a KV state is one token's key-value pair within one layer and KV head. We retain the highest-scoring states at high precision under a per-layer, per-KV-head budget. Under exact restoration, we bound the excess risk of score-based anchor selection relative to the optimum at the same anchor budget, where risk is the conditional mean square of the first-order loss change. With a budget of 56 additional FP16 anchors per layer and KV head, CALFA-KV (d8b8) outperforms all evaluated nominal 1-bit baselines on the mean score across seven benchmarks for each of the three models. At the same anchor budget, CALFA-KV (d8b10) uses a nominal index rate of 1.25 bits per channel and narrows the gap to BF16 to 0.90–2.01 points on the same metric.
Read the full abstract

All authors Xiancai Tian, Leichao Zhang, Shenhui Zhang, Yi Shan, Hailong Li, Feng Lin, Xiaotian Yu, Aijun Li

Robotics

ReSkill: Selective Self-Improvement for Robot Manipulation

Self-improvement on compositional manipulation tasks requires deciding what to reuse, where to improve, and which changes to retain. We present ReSkill, a selective self-improvement framework that concentrates revision on capability gaps. A coding agent constructs an object-centric task graph from an instruction and a skill library, then uses node-level execution records to locate failures. At the task level, it revises stage organization, hand-off conditions, and recovery routes; at the skill level, it strengthens perception, motion, manipulation, and verification capabilities that repeatedly fail. Candidate revisions undergo full-task evaluation and paired comparisons before adoption, allowing local changes to accumulate into validated skills and an independently executable policy. The agent participates only during development; the deployed policy calls the retained skills without invoking the agent at run time. On the same development set, average success across ten RoboDojo tasks rises from 30.0% for the initial versions to 86.2% for the final adopted programs; retained programs achieve 83.6% in recorded robustness and model-selection evaluations. Across eight real-world tasks, ReSkill succeeds in 72 of 80 rebuilt-simulation trials and 62 of 80 real-robot trials (77.5%). The median revision changes only three functions, approximately 0.5% of a program, while 39% of task-specific functions in later-added tasks directly reuse existing implementations. These results show how locating local deficiencies, validating candidate changes, and accumulating reusable skills can improve execution of compositional manipulation tasks.
Read the full abstract

All authors Zheng Sun, Zhihao Li, Zhuo Li, Wenrui Liu, Xin Wang, Fei Chen

Computer Vision

Scaling Optimization-Oriented Hypernetworks for Implicit Neural Representations

We present a hypernetwork that generates the parameters of an implicit neural representation (INR) for a given signal in a single forward pass, without test-time optimization. Existing hypernetworks for this problem either ignore the chain-rule structure of the test-time optimization that produces an INR or, when they preserve it, instantiate the structure with MLPs whose parameter count grows quartically in target-network width. Our central contribution is to show that this parameter-count cost is structural to the MLP instantiation, not to the optimization-oriented principle of preserving chain-rule structure, and to resolve it by replacing the MLP body with attention. The resulting hypernetwork combines three design choices: an output-neuron-centric tokenization that aligns the token axis with each layer's output-neuron axis; an information-flow consistency rule that fixes the role of every cross-attention module from the representation its output should occupy; and a token-space iterative rollout that realizes the five quantities one optimization step on the target INR computes as five cross-attention modules with parameter count independent of target-network width. Across five 2D and 3D INR generation benchmarks, our method outperforms the state-of-the-art single-pass and optimization-oriented baselines; scales to target-network widths at which the MLP-based optimization-oriented paradigm exhausts GPU memory; and produces the highest result on NeRF generation without ground-truth INR weights supervision.
Read the full abstract

All authors Lulu Cai, Hanxiang Ren, Siyan Dong, Li Sun, Jiefeng Wu, Youyi Zheng, Yi Ma, Yanchao Yang

Computation and Language

Imperfectly Imperfect: Evaluating Graded Cognitive Diversity in LLM Persona Simulations

Large language model (LLM) agents are increasingly used as synthetic participants in social, behavioural, and decision-making research, yet it remains unclear whether they can express graded cognitive diversity as assigned when engaging in simulated behaviour. We introduce a cognition-suite evaluation that tests whether simulated populations reproduce human variation in reasoning, self-reported traits, their associations, and downstream behaviour. Agents conditioned on synthetic or human-derived profiles of cognitive reflection, Need for Cognition, and Big Five personality answer the same suite as a cohort of 473 adults: two newly assembled tests, a Cognitive Reflection Test (CRT) and a Behavioural Reflection Test (BRT) whose decision scenarios measure reflection in behaviour, alongside an established CRT and the two trait questionnaires. Across seven full-battery simulation runs of four models, agents recover Need for Cognition and Big Five trait distributions on self-report questionnaires, but fall short on simulating diversity of cognitive task performance and downstream behaviours. In the four Claude Sonnet 4.6 runs, every agent scores full marks on both CRTs, whatever reflection level its persona was assigned. In the other three runs agents make intuitive-lure errors (the tempting wrong answers CRT items invite), but 49 to 85% score full marks, against under 10% in the human sample. Claude Sonnet 5 comes closest to people's item-by-item pattern of lures; the other two runs give fewer lures and vary less between items. In the BRT scenarios, every run over-produces 8 to 11 of 15 coded behaviours by at least 20 percentage points, and people's links between reflection and behaviour carry over only in part. Useful synthetic populations must preserve not only trait descriptors but the structured variation in reasoning, error type, and behaviour those descriptors are meant to organise.
Read the full abstract

All authors Sion Weatherhead, Ben Newell, Aaron Belbasis, Flora Salim

Quantum Gases

Thouless pumping and Haldane-model topology in a driven extended SSH chain

Thouless pumping provides a powerful route for probing Chern topology through quantized motion in one-dimensional lattices. We investigate such a topological pump in an extended Su-Schrieffer-Heeger (SSH) model with tunable next-nearest-neighbor hoppings. A graphical representation in Bloch space is used to construct pumping schemes. In this framework, the Bloch vector traces a closed loop that evolves in time. The Chern number of the synthetic band is then determined by the motion of this loop. Quantum pumping phenomena are demonstrated for a filled band and for a bosonic wave packet driven by Bloch oscillations. We further show that the Haldane model can be mapped to the driven extended SSH chain by replacing a quasimomentum by the pumping time. The quantized charge pump indicates that the Chern number of a two-dimensional model can be read out from a one-dimensional pumping protocol, offering an accessible route for detecting higher-dimensional topology, such as in photonic and cold-atom platforms.
Read the full abstract

Computer Vision

SRA: Spatial Reasoning Adapter via Evolving Social Interaction Graphs for Trajectory Prediction

Predicting where multiple agents will move next is fundamental to autonomous driving, robotics, and any system that must share space with people safely. Modern stochastic predictors read off inter-agent interactions once from the observed past and keep that view fixed throughout the generative rollout, so the reciprocal geometric constraints that emerge as the joint future sharpens are not re-examined. We address this gap with SRA (Spatial Reasoning Adapter), a modular graph denoiser that is integrated into existing diffusion and flow-matching predictors without redesigning their generative cores. At every reverse step, SRA reads the host's current future estimate, builds a sparse Euclidean interaction graph on the predicted trajectories, and returns a residual correction to the host's decoding path. Our adapter consists of three core components: a sparse neighborhood retrieved with a learned query-key score and a geometric bias, a future-conditioned relational feature recomputed on the evolving prediction, and uncertainty-weighted message passing where confident agents act as anchors for uncertain ones. We instantiate SRA on three structurally different hosts (LED, MID, and MoFlow) across four multi-agent benchmarks: SDD, NBA, Soccer, and Football. SRA improves ADE for every host on every dataset, with average gains of 5.8% in ADE and 7.0% in FDE. A cumulative ablation shows that the three components act as one mechanism: only together do they improve every host on every sports dataset, indicating that spatial reasoning over the model's own evolving future, weighted by how far that future can be trusted, is broadly useful across denoising architectures. Code is available at https://github.com/Jaewoo97/SRA.
Read the full abstract

All authors Jaewoo Jeong, Seonkyu Song, Hyeonwoo Park, Jegyeong Cho, Yujin Bae, Giwon Lee, Daehee Park, Kuk-Jin Yoon

Fluid Dynamics

Reconstructing Interacting Blood Flow and Wall Interface Deformation from Noisy Flow Data Using PirateNets

Blood flow and elastic vessel-wall motion interact with each other: fluid stress drives wall deformation, while the moving wall changes the domain in which blood flows. Given noisy flow images, we study the inverse problem of reconstructing the coupled, unsteady three-dimensional motions. We solve the reconstruction through Gauss–Seidel-type alternating fluid and geometry updates. For the fluid, we use physics-informed networks to recover velocity and pressure by fitting the observations while enforcing the incompressible Navier–Stokes equations on the moving domain, which is described by an arbitrary Lagrangian–Eulerian (ALE) map. For the geometry, a reference-domain correction and sequential registration recover the vessel shape and wall displacement. Interface traction continuity and an elastic momentum equation regularize the registration so that the recovered displacement is also physics-informed. The networks evaluate fields and physical residuals at sampled space–time points without remeshing the moving domain. Experiments on a converging channel and a patient-like aorta demonstrate reconstruction of flow and wall motion from noisy data. Ablation and parameter-sensitivity studies examine the role of structure regularization and imperfect physical coefficients. A further experiment with a three-element Windkessel outlet condition demonstrates how the framework accommodates a more physiologically realistic boundary setting.
Read the full abstract

All authors Han Zhang, Xue-Cheng Tai, Alessandro Veneziani, Raymond H. Chan

Optics

Dynamics of squeezed-state preparation in quadratic and Kerr nonlinear microresonators

We develop a dynamical model of squeezed-state generation in nonlinear optical microresonators that takes into account the back-action of the signal mode on the pump mode. The resonator is treated as an open quantum system, while the pump modes are described by time-dependent classical amplitudes, which allows pump depletion to be included beyond the standard undepleted-pump approximation. For a $\chi^{(2)}$ microresonator, the model reproduces the conventional 3~dB intracavity squeezing limit below threshold and predicts a transient above-threshold regime in which this limit can be overcome for a finite time before pump depletion becomes appreciable. The dynamical approach is extended to a three-mode Kerr microresonator with two pump modes. Separately, we investigate stationary squeezing in a reduced five-mode model with fixed pump amplitudes, including a parasitic single-pump spontaneous four-wave-mixing channel. Within this model, the parasitic channel reduces the achievable squeezing and shifts the instability threshold to lower pump amplitudes. These results provide a compact framework for estimating the validity range of the undepleted-pump approximation and for analyzing transient squeezing dynamics in nonlinear microresonators.
Read the full abstract

All authors S. D. Chikalkin, V. L. Gorshenin, V. E. Lobanov, I. A. Bilenko, D. A. Chermoshentsev

Machine Learning

Learning a Unified Cross-Model Semantic Dictionary via Gated Bottleneck Sparse Autoencoders

Foundation models pretrained on similar data distributions tend to learn overlapping semantic structure, but their representations are distinct due to different model architecture, pretraining objectives and protocols. This prevents unified interpretability and cross-model feature reuse. Existing approaches extract interpretable concepts from individual models and align these concepts in a shared coordinate system using cross-reconstruction objectives, but the alignment only applies to latent coordinates while the concepts themselves remain model-specific and cannot be shared or reused. This paper introduces the Gated Bottleneck Sparse Autoencoder (GB-SAE), which casts unified cross-model interpretability as a representation factorization problem and solves it by simultaneously learning a unified shared dictionary and multiple private model-specific dictionaries via a gated bottleneck sparse autoencoder framework. The shared dictionary encodes the cross-model common semantic structure in unified concepts that are reusable across all participating models (and beyond), while each private dictionary comprises residual concepts that capture model-specific biases. We evaluate GB-SAE with distinct foundation models including ViT, SigLIP, DINOv2, and ConvNeXt. GB-SAE demonstrates superior reconstruction capability, unified and faithful interpretability, strong reusability and generalization to unseen models and downstream tasks. The code is available at https://github.com/trustAIoT/GB-SAE_NeurIPS.
Read the full abstract

Computation and Language

Distinguishing Extraction Failures and Clinical Discordance with LLM Uncertainty on Radiology Reports

Large language models (LLMs) are increasingly used to structure data from unstructured text, but extraction errors remain inevitable. We evaluated whether LLM uncertainty could distinguish extraction failure from true clinical discordance. In 5{,}225 vessels from 1{,}045 patients undergoing coronary CT angiography (CCTA) for suspected coronary artery disease (CAD), we evaluated token-level and self-reported LLM uncertainty during extraction of vessel-level stenosis severity. Extraction failure was defined as disagreement between LLM-extracted stenosis and expert annotations. Clinical discordance was defined as disagreement in obstructive CAD status between CCTA and subsequent invasive coronary angiography (ICA). Incremental discrimination was assessed after adjustment for stenosis severity, calcium burden, and coronary vessel. Extraction failure was evaluated in 1{,}077 vessels from 253 patients with expert annotations, and clinical discordance in 1{,}641 vessels from 729 patients undergoing ICA within 90 days. Token-level Stenosis Margin showed the strongest association with extraction failure (adjusted OR, 4.15 [95% CI, 2.57–6.69]; $P<0.001$) and clinical discordance (adjusted OR, 1.22 [95% CI, 1.04–1.43]; $P=0.118$). Stenosis Margin significantly improved discrimination of extraction failure ($\Delta$AUC, +0.06 [95% CI, +0.03 to +0.09]; $P=0.001$), but not clinical discordance ($\Delta$AUC, +0.01 [95% CI, $-0.00$ to +0.02]; $P=0.245$). In a nested cohort of 138 vessels from 57 patients, 38 of 42 clinically discordant vessels (90%) were extracted correctly. LLM uncertainty improved identification of extraction failure but not clinical discordance, suggesting that uncertainty may identify errors introduced during data extraction but may not generalize to discordance in underlying clinical data.
Read the full abstract

All authors Jerjes Aguirre-Chavez, Shant Malkasian, Roshun Sankaran, Brendan Crabb, Albert Hsiao

Artificial Intelligence

Sampling Is Not Curiosity: Why LLM Agents Should Investigate

LLM agents are increasingly used for open-ended tasks, very common in research scenarios where they are required to explore, investigate, or act curiously. At inference time, an LLM is a fixed map from context to next-token distribution, the pretrained weights determine the function, and only its conditioning changes as context accumulates. Mechanisms like temperature, best-of-N, tree search, or RAG vary how the distribution is sampled or what it is conditioned on, but do not update the agent's model of the environment. We argue that current LLM agents do not explore curiously; they explore stochastically. From how curiosity is understood in the RL literature, it requires a predictive structure over how the environment responds to the agent's actions, which improves online from current interaction, drives action selection through its predicted improvement, and persists beyond the context window. Additionally, in its open-ended form, the agent should intrinsically originate what to investigate. Thus, we propose a framework defining curiosity in LLM agents, reserving the term for systems that satisfy it. In doing so, we identify the research directions its implementation opens up.
Read the full abstract

All authors Alfonso Amayuelas, Piotr Piękos, Xin Eric Wang, William Yang Wang, Jürgen Schmidhuber, Alex Pentland

Quantum Physics

On the estimation of linear-optical amplitudes and probabilities with entangled input state

Most proposed applications of boson samplers, including photonic machine-learning models, do not use the samples themselves but estimates of output probabilities, marginals or overlaps, obtained to additive precision from a finite number of shots. For product inputs such as single photons, these quantities can already be estimated classically with Gurvits's algorithm and its generalisations. We ask whether entangled inputs change this picture. For any $n$-photon input that is a matrix-product state of bond dimension $\chi$, with one photon per site and $L$ modes per site, we give an unbiased estimator of any output amplitude of an $m$-mode interferometer, whose mean-square error after $N$ draws is at most $\|\psi\|^2/N$ whatever the interferometer, the output pattern and the entanglement of the state. Each draw costs $O(nL\chi^2+nL^3+mn)$ operations after $O(nL\chi^3)$ preprocessing, and an additive error $\varepsilon$ with probability at least $1-\delta$ requires $O(\|\psi\|^2\varepsilon^{-2}\log(1/\delta))$ draws, with memory of the order of the input size. The main application is to light emitted sequentially by quantum emitters, such as spin-based sources of photonic cluster states, for which $\chi$ is bounded by the dimension of the source's retained quantum memory: one spin gives $\chi=2$ at any length, and a memory of $k$ qubits, held in emitters or in photons stored in a delay line, gives $\chi\le2^k$, so the cost stays polynomial as long as $k=O(\log n)$. The same sampler also estimates traced-memory probabilities, overlaps and individual marginal probabilities. All guarantees are additive: the estimator gives neither relative estimates of exponentially small probabilities nor a sampler of the output distribution.
Read the full abstract

Number Theory

A Goldbach-Type Problem for the Unitary Möbius Function

Let $g(n)=(-1)^{\omega(n)}$, where $\omega(n)$ denotes the number of distinct prime factors of $n$. We prove that, for every sufficiently large integer $N$, each ordered additive sign pattern $(g(a),g(N-a))=(s,t)$ occurs at least $N/(128(\log N)^3)$ times. We also prove that, for every $N\ge18$ and every $s,t\in\sgnset$, there is at least one integer $1\le a<N$ such that $(g(a),g(N-a))=(s,t)$. In addition, we establish a one-quarter asymptotic formula for almost all target integers. Finally, we discuss Goldbach-type problems for the ordinary Möbius function and formulate related questions.
Read the full abstract

Robotics

EmbodiedIntentBench: Benchmarking Intent-Driven Runtime Adaptation for Mobile Robotic Missions

Mobile robotic sensing systems must adapt to operator requirements, limited resources, changing connectivity, and observation opportunities shaped by physical motion. Evaluating runtime strategies therefore requires assessing whether their decisions fulfill mission objectives, beyond isolated perception accuracy or execution efficiency. We present EmbodiedIntentBench, a benchmark for intent-driven runtime adaptation in mobile robotic missions. It represents mission intent through structured contracts for coverage, temporal evidence, reporting, urgency, and resources, and couples simulated unmanned aerial vehicle (UAV) workloads with causal, resource-accounted execution. Computation decisions affect mission time, resource use, and which observations can contribute evidence. We evaluate fixed and adaptive perception strategies in multi-phase missions and under changing urgency, connectivity, and battery conditions. In the multi-phase Depot study, a configuration with 87.5% full-bank target recall across visible target observations completes only 1 of 16 missions, whereas a lower-recall configuration completes 14. In an ablation, ignoring execution-induced observation loss increases aggregate mission success and reverses the ordering of two runtime strategies. Replays using Jetson AGX Orin CPU and GPU profiles change the relative success of local and remote strategies. In selected Jetson GPU-profile replays, intent-driven adaptation reduces modeled task energy by up to 28% with unchanged mission success. EmbodiedIntentBench provides an extensible framework for evaluating runtime decisions against mission fulfillment under physical and resource constraints.
Read the full abstract

All authors Rajat Bhattacharjya, Changwoon Lee, Hyeonjong Ju, Sing-Yao Wu, Hamidreza Alikhani, Zhuoyi Yang, Ian Harris, Nalini Venkatasubramanian, Nikil Dutt

Computer Vision

Disentangling Channel Semantics in Vision Transformers via Token Decorrelation and Composition-Aware Modulation

Multi-channel images encode heterogeneous channel semantics aligned over a shared spatial structure, which requires models to jointly capture globally shared structure and channel-specific variations. Recent multi-channel Vision Transformers attempt to address this challenge by augmenting the [CLS] token with memory tokens to organize multi-channel representations under varying channel compositions. However, we find that these context tokens collapse onto a few dominant channels under high inter-channel redundancy, producing redundant representations and a biased global summary that fails to reliably support composition-aware channel interpretation. We propose MuCa-ViT (Multi-Channel Composition-Aware Vision Transformer), a unified framework that addresses these limitations through two coupled mechanisms. Factorial Token Learning (FTL) enforces token-level decorrelation among context tokens, encouraging the [CLS] token to capture globally shared semantics while memory tokens preserve complementary channel-specific cues. Channel Composition-Aware Modulation (CAM) uses the FTL-refined [CLS] as a semantic anchor to generate sample-wise modulation signals, which enable composition-aware interpretation of channels within a unified backbone rather than relying on static channel embeddings. Experiments on JUMP-CP, CHAMMI, and So2Sat show that MuCa-ViT outperforms prior multi-channel ViTs across all three benchmarks, and the magnitude of improvement aligns with the inter-channel redundancy of each dataset ($\rho \in [0.19, 0.77]$), validating the diagnostic analysis that motivates our design. Our code is available at https://github.com/kimdanni/MuCa-ViT .
Read the full abstract

All authors Daeun Kim, Hyesong Choi, Hyejin Park, Dongbo Min

Systems and Control

On Real Eventual Exponential Positive property of complex-valued Laplacians and their pseudoinverses

The property of real eventual exponential positivity guarantees the state trajectories of a linear time-invariant system to evolve asymptotically to right half of the complex-plane [1]. It has applications in the characterization of herdability and reachability of states in networked systems. Networks with complex-valued edge-weights arise in AC power systems and opinion dynamics. Recent work has introduced real eventual exponential positive (rEEP) Laplacians, for which the trajectories of the associated network flow eventually converge to consensus. We establish further properties of rEEP Laplacians and their pseudoinverses. Firstly, we show that the rEEP property for complex-valued matrices exhibits invariance under specific rotational transformations. Secondly, we establish that the property of rEEP is preserved under the MoorePenrose pseudoinverse operation, highlighting its robustness under algebraic manipulations. These results apply to effective impedance computation in AC networks and to consensus in multi-agent systems. Finally, we provide numerical examples to validate our theoretical results.
Read the full abstract

Number Theory

Multivariate Pellarin $L$-values on algebraic curves

We establish a generating trace formula for multicharacter ideal-type $L$-values associated with rank-one Drinfeld modules. For a sign-normalized Hayes module with free period lattice, the formula realizes the $L$-values at weights $n\equiv s\pmod{q^{\deg_\infty}-1}$ as a trace over the ordinary Hilbert class field. We further characterize the polynomials in the exponential variable $X$ arising in multicharacter lattice generating identities over arbitrary function rings. Their coefficients are given explicitly via the shtuka basis, which gives concrete control over their supports and degrees. This completely settles an open problem of Pellarin and Perkins: these polynomials are all constant if and only if $s<2q$. The results on coefficients apply to curves of arbitrary genus, infinite places of arbitrary degree, and nonprincipal period lattices.
Read the full abstract

Statistical Methodology

Generalized Laplacian in Spectral Seriation on Manifold Data

Noisy high-dimensional observations may have conditional means lying on a smooth closed curve, indexed by temporal labels or phases. We want to recover the latent phases for each observation and thus retrieve a circular ordering. Spectral seriation uses the graph Laplacian's spectral embedding to recover the phases and ordering. We consider generalized graph Laplacians in spectral seriation, where a parameter $\alpha$ controls the degree normalization and affects the accuracy and stability of both tasks. We develop and compare three generalized Laplacian constructions: a shifted generalized Laplacian, a graph-difference Laplacian, and an extra-normalized generalized Laplacian. Among these, the extra-normalized variant is less sensitive to the choice of $\alpha$ and achieves the strongest overall recovery performance. For this variant, we establish a theoretical error bound relating phase estimation accuracy to eigenspace perturbation, accounting for the inherent rotation and reflection ambiguity. In synthetic experiments across various scenarios, our new graph Laplacian outperforms the classic spectral seriation method and other seriation methods. On the Pink Angel data, which records photos from different viewpoints, our method achieves zero ordering error and the lowest phase recovery error. On FUCCI cell-cycle data, our method performs the best when all 6 cells are mixed, demonstrating the robustness to heterogeneity.
Read the full abstract

Optimization and Control

Geometric Stochastic Dynamics and Fokker-Planck Equation on $\mathsf{T}\mathbb S^2$

This paper develops a geometric (coordinate-free) stochastic model of a controlled point mass evolving on the unit two-sphere $\mathbb S^2$, together with the Fokker-Planck equation governing the probability density of the resulting stochastic process. The state manifold is the tangent bundle $\mathsf{T}\mathbb S^2$. The Levi-Civita connection of the round metric is used to construct the horizontal-vertical splitting of $\mathsf{T}(\mathsf{T}\mathbb S^2)$ and to convert the intrinsic second-order dynamics of the system into a first-order controlled vector field on $\mathsf{T}\mathbb S^2$. Endowing the tangent bundle with the Sasaki metric then permits stochastic perturbations to be introduced intrinsically and independently in horizontal, i.e., position, and vertical, i.e., velocity, directions. The corresponding infinitesimal generator is obtained in horizontal-vertical Hessian form. The probability density is defined with respect to the Sasaki Riemannian volume measure. The Fokker-Planck equation is derived intrinsically in both Stratonovich and Itô forms. In the Itô representation, a globally defined horizontal-vertical diffusion tensor yields a compact double-divergence Fokker-Planck PDE. No coordinates or embedded Euclidean representation of the sphere dynamics are used in the development. Numerical simulations illustrate the stochastic motion of a controlled particle ensemble on $\mathbb S^2$ toward an antipodal target configuration.
Read the full abstract

Complex Variables

A Zero-Free Sum of Cauchy Kernels with Positive Coefficients

We construct a zero-free meromorphic function \[ f(z)=\sum_n\frac{c_n}{z-a_n},\qquad c_n>0,\qquad |a_n|\to\infty,\qquad \sum_{a_n\ne0}\frac{c_n}{|a_n|}<\infty, \] thereby disproving the Clunie–Eremenko–Rossi conjecture. The construction starts from a product of two rotated Airy functions. The zeros of this product are shifted simultaneously so that all residues of $f$ are positive. The poles lie on the rays $e^{\pm i\pi/3}[0,\infty)$. The resulting function has order and lower order $3/2$, and the sum of its residues in the disc $|z|\le r$ is asymptotic to $\frac{4}{3\pi}r^{3/4}$.
Read the full abstract

Numerical Analysis

A Trajectory-Aware Domain-Decomposed Multispace Reduced Basis Preconditioner for Parametric Helmholtz Problems

Repeated solutions of parametric Helmholtz problems arise in multi-query applications such as sensitivity analysis and uncertainty quantification. Classical preconditioned iterative solvers do not automatically reuse low-dimensional structures across parameters. Multispace reduced basis (MSRB) preconditioners can exploit such structures by constructing iteration-specific reduced order models (ROMs) from training data, but storing global iteration-dependent bases incurs substantial memory costs. Building on trajectory-aware MSRB training, we develop a domain-decomposed MSRB (DD-MSRB) preconditioner for parametric Helmholtz systems truncated by perfectly matched layers (PMLs). Within flexible GMRES, each reduced space is trained on inverse-action snapshots along the Krylov trajectories generated by the ROM-based preconditioners applied in preceding iterations. To reduce storage, problem-adapted spatial partitions exploit heterogeneous local snapshot compressibility across the physical interior and the PML. Local basis dimensions in each subdomain are determined greedily under a global snapshot-error budget to balance approximation accuracy and storage. The resulting hybrid solver applies DD-MSRB preconditioning during the initial iterations to capture dominant cross-parameter modes and then switches, when necessary, to a shifted-Laplacian preconditioner to reduce error components poorly represented by the reduced spaces. Numerical experiments on two-dimensional half-space propagation, elliptic-obstacle scattering, and three-dimensional waveguides demonstrate significant acceleration over the shifted-Laplacian baseline and substantial memory savings relative to global MSRB. For the largest waveguide system, with 11.0 million unknowns, DD-MSRB reduces deployed ROM storage by 44.9% and mean online solve time by 30.2% compared with global MSRB.
Read the full abstract

Robotics

Tactile Sensors Are an Effective Substitute for Wrist Cameras in Contact-Rich or Occluded Manipulation

Learned manipulation policies often rely on wrist-mounted cameras for close-range visual feedback that complements external views of the workspace. However, contact-rich manipulation requires establishing and maintaining contacts that can remain hidden even from wrist cameras. Tactile sensors mounted on fingers can reliably and directly measure contact, without needing visual access. In this work, we show that this feedback enables policies to recover the performance advantage of wrist cameras using scene and egocentric vision together with touch. Our approach incorporates two low-cost tactile arrays into a visuotactile ACT policy, without wrist images during training or deployment. We evaluate six real-world task families spanning placement, insertion, occluded retrieval, and sustained-contact tool use. On three contact-rich tasks, we independently vary wrist-camera and tactile availability while keeping the demonstrations, policy architecture, and optimization settings fixed. Adding wrist cameras improves success by 18.3-28.3 percentage points over vision without wrist images. Adding tactile sensing instead improves success by 31.7-60.0 percentage points and yields higher observed success than vision-only policies with wrist cameras on all three tasks, with significant advantages on two under Holm-corrected tests. These results support finger-mounted tactile sensing as an effective alternative to wrist cameras in the evaluated contact-rich settings.
Read the full abstract

All authors Aijia Yang, Arsh Tangri, Sergey Zakharov, Krishna Murthy Jatavallabhula

Number Theory

Local vanishing density of order $1$ and Divisor moments

We establish a general framework relating local vanishing densities of order $1$ to logarithmic growth of divisor moments on regular arithmetic curves. As applications, we recover classical divisor moment estimates for polynomial values and establish their function field analogue. Furthermore, under GRH, we obtain a result on divisor moments for polynomials in the Fourier coefficients of non-CM cuspidal eigenforms with trivial nebentypus and trivial inner twist. The Fourier coefficients need not be integers, and the polynomials need not be squarefree. Under additional analytic hypotheses, we also allow a Sato–Tate angle restriction.
Read the full abstract

Human-Computer Interaction

Safer for Whom? A Systematic Exploration of AI Safety in Eating Disorder Interactions

AI assistants advise users about diet, exercise, and weight loss—areas in which unsafe guidance can worsen body dissatisfaction, dietary restriction, compulsive exercise, and other behaviors associated with eating disorders (EDs). Existing evaluations emphasize diagnostic accuracy, overtly harmful prompts, or synthetic scenarios, providing limited insight into whether models recognize implicit vulnerability in natural interactions or protect users consistently. We evaluate responses from three open-weight large language models to 100 questions sourced from an online pro-anorexia community. We grade responses using clinically informed criteria that extend a seven-dimensional taxonomy of ED-related harms. We also construct matched conditions in which questions are accompanied by information about the user varying race, gender, body mass index (BMI), and explicit ED vulnerability. Safety varies substantially across models and harm dimensions. High-risk responses are approximately twice as frequent without bios, suggesting that models do not consistently recognize vulnerability from language alone. Explicit signals of ED vulnerability improve safety by 10–14 points across models. Responses also differ significantly by race, gender, and BMI, with body size producing the largest disparities. These patterns are consistent with reliance on the stereotype that EDs primarily affect thin, White women, potentially leaving users outside this profile with less protective responses.
Read the full abstract

All authors Jinyuan Wang, Minh Duc Chu, Frances Ross, Ellen Fitzsimmons-Craft, Kristina Lerman

Optimization and Control

GEAR: A GPU-Accelerated Global Solver for Bounded Continuous Nonlinear Programs via Linear Bound Propagation

We present a novel GPU-accelerated global solver for constrained nonlinear programming (NLP) inspired by the efficient linear bound propagation framework in neural network (NN) verification. Existing work on NN verification, a problem that can be cast as feasibility detection for a nonlinear objective, demonstrated orders-of-magnitude speedups with the linear bound propagation-based framework when compared to traditional solvers. Extending this paradigm to global nonlinear programming requires addressing many challenges: a global constrained NLP solver must effectively exploit all, possibly unstructured, nonlinear constraints to improve the objective, whereas NN verification typically handles simple input box constraints over a well-structured NN and does not directly optimize an objective function. We address these challenges by fully exploiting the linear bounds produced by bound propagation: these linear bounds can be viewed as relaxations of nonlinear constraints and used to provide lower bounds, check infeasibility, and guide the search for a primal solution. In addition, we propose tighter linear relaxations for bilinear, quotient, and other commonly used nonlinear functions in NLP problems, and a projected augmented Lagrangian method tightly coupled to our dual solving procedure to produce primal solutions. Our method solves 421 of 505 bounded and continuous NLP problems in the GAMS and MINLPLib benchmarks under a 180-second limit, more than each of the six global solvers we compare with, including Gurobi and BARON (commercial), and SCIP (open-source). Sixteen of these problems are certified by none of the six global solvers; all of them contain exponential, logarithmic, or trigonometric terms, which are often challenging for traditional approaches. Project webpage is at https://verified-intelligence.github.io/GEAR.
Read the full abstract

All authors Duo Zhou, Hesun Chen, Xiangru Zhong, Grani A. Hanasusanto, Huan Zhang

Computation and Language

Test-Time Scaling for Diffusion Language Models with Frequency-Aware Remasking

Diffusion language models (DLMs) generate text through iterative denoising, where decisions about which tokens to commit or remask strongly affect output quality. Existing certainty-aware strategies rely on instantaneous confidence or entropy, but these local signals can be misleading on difficult reasoning problems. We formulate DLM test-time scaling as adaptive allocation of both inference compute and decoding strategies. Our two-stage approach uses a fast base sampler to solve easy questions, then applies frequency-aware remasking to the remaining hard questions. This strategy tracks how often each token's prediction changes across denoising steps and preferentially commits frequently changing tokens, encouraging samples to explore alternative reasoning paths rather than collapse onto the model's preferred trajectory. Experiments with LLaDA-8B-Instruct and Dream-7B-Instruct show consistent improvements over a strong adaptive allocation baseline, with absolute accuracy gains of up to 3.73 percentage points (LLaDA) and 7.33 percentage points (Dream) on MATH-500, and absolute coverage gains of up to 11.39 percentage points (LLaDA) and 10.98 percentage points (Dream) on HumanEval. On questions unsolved by the base certainty-aware strategy after 100 samples, frequency-aware remasking recovers solutions for 3.75% to 18.75% of examples.
Read the full abstract

All authors Bowen Zuo, Yue Yu, Dongruo Zhou, Yinglun Zhu

Computer Vision

Decoupling Direction and Magnitude: Language-Steered Flow Matching for Super-Resolution in the Dark

Image super-resolution in the dark is fundamentally challenged by extreme spatial heterogeneity: severely underexposed and noisy regions demand aggressive generative enhancement to reconstruct missing details, while relatively well-exposed areas require conservative updates to prevent hallucinated textures. Standard diffusion and flow-matching models, however, rely on globally uniform integration trajectories, making them sub-optimal for such spatially varying degradations. In this paper, we propose DeLang-SR (Degradation-Language steered Super-Resolution), a one-step flow-matching framework that explicitly decouples the restoration direction and magnitude. Rather than treating low-light corruption as arbitrary latent features, DeLang-SR translates measurable physical statistics into structured degradation language, harnessing the semantic prior of pretrained text-to-image models. The degradation-language intent, complemented by image-specific visual tokens, steers the velocity field (direction) toward restoration-relevant manifolds. Simultaneously, a continuous spatial update-strength field acts as a locally varying Euler step-size field (magnitude), applying stronger integration steps in heavily degraded shadows while enforcing conservative updates in reliable regions. Experiments on paired and real dark images, together with ablations and diagnostic analyses, show that DeLang-SR provides an effective and interpretable one-step route to super-resolution in the dark. Code and pretrained models are available at https://github.com/moriyaya/DeLang-SR.
Read the full abstract

All authors Jiaxin Gao, Ziyu Yue, Yaohua Liu, Danchen Cui, Zhihui Zhao, Zhixun Su

Computer Vision

Can General-Purpose Generative AI Supply Training Data for Medical Vision Models? A Multimodal Retrospective Study

Background: Well-annotated, diverse medical imaging datasets remain a major constraint for vision-based medical AI. We evaluated Nano Banana, a general-purpose image generation and editing model, as a source of synthetic medical images for model training, subgroup augmentation, and model auditing. Methods: In this retrospective study across four imaging modalities, chest radiography, fundus photography, and CT models were trained only on synthetic images and externally validated on real-world data against real-data-trained models. Carotid ultrasound models were trained on a private Human Phenotype Project cohort (aged 40-75 years) with and without synthetic augmentation. Clinicians rated image plausibility. Evaluation used AUROC, AUPR, calibration, and decision curve analysis. Findings: Synthetic-only models achieved AUROCs of 0.877 (95% CI 0.843-0.910) for pneumonia (n=14863), 0.963 (0.948-0.975) for retinal detachment (n=1149), and 0.711 (0.600-0.823) for renal calculus (n=4122), comparable to real-data-trained models, but their calibration and net benefit were not consistently more favorable. With mixed synthetic-real training, a carotid ultrasound model produced an ordinal vascular age index that stratified 633 external subjects into subgroups differing in BMI and glucose (both p<0.01), linking vascular aging to cardiometabolic health. To stress-test shortcut learning, editing a moderate tessellated pattern into fundus photographs reduced the true-class probability for central serous chorioretinopathy from 0.996 to 0.822. Clinician-guided editing raised low-quality CT plausibility from 3.10/5 to 4.30/5. Interpretation: Synthetic data can complement real data in model training, subgroup representation, and model testing via controlled editing, but its clinical value should be assessed with calibration, decision curve analysis, and clinician review, not discrimination alone.
Read the full abstract

All authors Jiachuan Peng, Kyle Lam, Jierui Qu, Lin Li, Haining Zheng, Bang Zheng, Aniruddha Agarwal, Zaki Almallah, Min Chen, Vicente Grau, Hagai Rossman, Eran Segal, Jianing Qiu