Peer Revenue Returns subject to peer review

Papers

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

All of arXiv

Wednesday 2 September

Robotics

Following a Unique Path: A Fast Certifier Applied to Outlier-Robust Pose Registration

Certifiable methods have arisen as a means to guarantee global optimality of solutions to non-convex problems using convex semidefinite programming (SDP) relaxations. The most performant of these methods use a local solver to obtain the candidate solution, and then certify its optimality using efficient linear algebra techniques. However, for many problems of interest in robotics, this local-solve-then-certify approach is impeded by a form of degeneracy in the relaxation, leaving a costly optimization of the relaxation as the only recourse. In this paper, we introduce our Central-Path Certifier (CP-Cert), a certifiable method explicitly tailored to certify candidate optima to problems that exhibit this form of degeneracy. Using a candidate as a starting point, our approach seeks a nearby region of the feasible space – known as the central path – where a valid certificate can be readily obtained. The approach is kept efficient by exploiting indirect linear algebra techniques, problem sparsity, and parallelism. We apply CP-Cert to both matrix-weighted pose registration and pointcloud data association, whose novel SDP relaxation is of independent interest. On simulated examples, we explore the properties of this novel relaxation and show that CP-Cert is fast and scalable, achieving runtimes that are up to three orders of magnitude faster than state-of-the-art direct solvers. Finally, we combine these contributions into a certifiable, outlier-robust pose-estimation pipeline, which we apply to real-world data.
Read the full abstract

Computation and Language

Instability Floors: Separating Bias from Noise in Fairness Audits of Clinical LLM Agents with FairMedAgent

Counterfactual fairness audits of clinical language-model agents report a flip rate: how often an action changes when only the patient's demographic descriptor changes. Part of that rate is not demographic. A stochastic agent also changes its own action when nothing changes, and a flip rate cannot be interpreted without knowing how often. We measured it. Re-running one condition ten times over sixteen synthetic vignettes at default sampling changed a clinical agent's action in 8.7 percent of replicate pairs, from 2.2 percent for intensive-care escalation to 17.9 percent for controlled-substance caution, an output given no operational criteria. Across six models from five vendors, pooled floors ranged from 2.5 to 23.7 percent; in this panel neither disclosed size, vendor, nor hosting ordered them. The floor depends on the decoding configuration: majority voting over five draws removed 39 percent of it (95 percent confidence interval (CI) 18 to 64); at temperature 0 three of four locally served models showed no disagreement, but a hosted model still did. We also show that, for a binary action, the flip rate expected under no demographic effect equals the floor and a real effect adds only its square, so a flip rate inside the floor is not evidence of fairness, and direction must be tested with a signed paired test. We give a four-step reporting procedure and release FairMedAgent, the harness, with its protocol, vignettes and analysis scripts, so any team can measure the floor for its own agent.
Read the full abstract

All authors Rohith Reddy Bellibatlu, Manpreet Singh, Deepak Parashar, Rahul Joshi

Quantum Physics

Discretization-Aware Fine-Tuning for Quantum Machine Learning with Chemical Foundation Models

A key challenge in practical quantum machine learning (QML), particularly for discriminative tasks such as classification, is the limited capacity of near-term quantum devices to encode high-dimensional classical data into small quantum registers. In optimized basis-encoded (bit-bit) settings, this constraint leads to cross-class collisions, where samples with different labels are mapped to the same discrete bit-string and thus become indistinguishable to any downstream model. In this work, we investigate how data representation affects QML performance under such severe information bottlenecks. We introduce discretization-aware fine-tuning (DAFT), a method that adapts a pre-trained chemical foundation model to produce representations that remain informative after quantization. DAFT reduces collision probability through a differentiable soft collision loss. We evaluate both quantum and classical models under a controlled setting in which they receive identical discretized bit-string inputs, isolating the effect of representation from model architecture. On the blood-brain barrier penetration (BBBP) molecular property prediction benchmark using ChemBERTa-77M, DAFT reduces collision counts by several orders of magnitude and improves quantum classification accuracy by more than 12 percentage points compared to a frozen backbone. Importantly, without DAFT, classical models outperform QML under the same input constraints. With DAFT, however, this comparison reverses at higher qubit counts. At 10 qubits, the quantum model surpasses a matched classical baseline trained on identical bit-strings (0.883 vs. 0.855, $p = 0.026$). These results show that, in information-constrained regimes, achieving a quantum advantage critically depends on aligning continuous representations with discrete quantum encodings.
Read the full abstract

Optimization and Control

An Exact Combinatorial Branch-and-Bound Algorithm for the Job Sequencing and Tool Switching Problem

The Job Sequencing and Tool Switching Problem (SSP) is a well-known combinatorial optimization problem arising in the context of flexible manufacturing. Since the seminal work of Tang and Denardo (1988), the SSP has received significant attention in the literature, leading to the development of numerous exact and heuristic approaches. Despite these efforts, several benchmark instances proposed decades ago and containing only 20 jobs have remained unsolved to proven optimality. In this work, we propose an exact algorithm for the SSP, namely the Combinatorial Branch-and-Bound (C-B&B) algorithm, which combines two distinct branch-and-bound algorithms, each introducing novel features compared with the existing literature. The former relies on a new branching scheme designed to reduce the size of the implicit enumeration tree, together with a collection of new bounding functions. The latter builds on the branching scheme introduced by Laporte et al. (2004) and strengthens it with a new bounding function and two dominance rules. Within C-B&B, these exact algorithms are complemented by a preprocessing phase that incorporates a new branch-and-bound-based heuristic capable of rapidly generating a high-quality initial incumbent solution. Extensive computational experiments show that C-B&B represents a strong breakthrough over previously published approaches, proving optimality for more instances with significantly less computational effort and closing several benchmark instances that have remained open for decades.
Read the full abstract

Computation and Language

The Analyst in the Prompt: Role, Retrieval, and Memory Biases in LLM Financial Analysis

Large Language Models (LLMs) increasingly use user context such as memory, profiles, and role prompts to personalize their responses. This personalization can affect evidence-based judgment: the same evidence may lead to different conclusions under different user contexts. Finance provides a high-stakes setting to study this problem because decisions often depend on interpreting long and complex documents. We test this using 3,575 SEC filings across twelve LLMs. We compare persona-conditioned retrieval, neutral retrieval, and memory-framed context to separate the effect of evidence selection from the effect of interpretation. We find that most user-context spillover comes from how models interpret the same evidence under different roles, rather than from retrieving different evidence. We then test two simple mitigation strategies: expressing the same investor mindset as a user profile instead of an assistant role, and separating evidence-based and personalized outputs. Both reduce spillover, but neither removes it completely, and their effectiveness varies substantially across models.
Read the full abstract

All authors Ahmed Asaad, Amr Mohamed, Yang Zhang, Omneya Abdelsalam

Statistical Mechanics

Optimal Control of Periodic Nonequilibrium Mechanochemical Systems via Automatic Differentiation

Biological molecular machines are mesoscopic systems that act repeatedly and periodically to perform important cellular tasks while contending with strong fluctuations and operating in an overdamped regime. Optimal control theory is a tool that can be used to understand the design principles behind efficient operation of these machines; however, most studies on optimal control of classical mesoscopic systems have focused on control problems that do not repeat periodically. Here, we automatically differentiate Fokker-Planck simulations to design efficient nonequilibrium control strategies for simple models of periodic molecular machines with and without explicit changes in the machine's chemical state. The designed protocols and theoretical analysis provide insight into the design principles governing efficient driving in these nonequilibrium systems. Designed control protocols should seek to reduce mechanical heat by rotating the entire angular probability distribution at a constant speed without changing its shape, and should reduce chemical heat by reducing the proportion of chemical transitions with large heat.
Read the full abstract

Computer Vision

ProgResViT: Progressive Resolution and Width for Adaptive Vision Transformers

Vision Transformers (ViTs) typically process every image using a fixed input resolution and model width, even though many images can be classified with substantially less computation. We introduce ProgResViT, an input-adaptive ViT that performs inference progressively across multiple rounds. The first round processes a low-resolution image with a narrow subnetwork. Inference terminates when the prediction is sufficiently confident; otherwise, the model reuses the representations produced in the current round and proceeds with a higher-resolution input and a wider subnetwork to refine its prediction. As all rounds share a single backbone, we propose Progress-Conditioned Soft Gating (PSG), which conditions token fusion and layer outputs on the current round, block, and input resolution. On image classification, applying ProgResViT to DeiT yields better accuracy-compute trade-offs than adaptive-width, adaptive-depth, and dynamic-token baselines. With knowledge distillation, a DeiT-based ProgResViT achieves 84.9% top-1 accuracy, slightly exceeding the reported DeiT-III-S accuracy under a comparable evaluation setting. We show that the same design also provides favorable accuracy-compute trade-offs for self-supervised DINO representations and downstream semantic segmentation. Code is available at https://github.com/ds-kiel/ProgResViT.
Read the full abstract

Computation and Language

SWIM: Student Writing Simulation via Proficiency-Conditioned Generation

Writing proficiency manifests in how students develop content, organize ideas, choose words, and use language. Despite growing interest in LLM-based student simulation, whether LLMs can reproduce such multidimensional variation in extended writing remains largely unexplored. In this work, we explore if language models can realistically simulate student writing, and introduce SWIM, a task that formulates Student Writing sIMulation as proficiency-conditioned essay generation. We evaluate prompting, supervised fine-tuning (SFT), and reinforcement learning (RL) methods for writing simulation using automated essay scoring as a measure of profile alignment. Extensive experiments reveal that prompting provides limited proficiency control, even for strong proprietary LLMs with rubric-grounded strategies. In particular, while models can adjust content-oriented traits, they struggle to reproduce the lexical, grammatical, and organizational variation in different proficiency levels. SFT substantially improves alignment, while RL with the proposed proficiency-alignment reward yields further gains across all writing traits and essay prompts. Our findings suggest that explicit supervision enables substantially stronger profile alignment than prompting alone, while authentic low-proficiency writing remains challenging to reproduce.
Read the full abstract

Numerical Analysis

A novel parallel approach for solving some free boundary value problems

This paper introduces a novel, efficient class of parallel direct and indirect iterative schemes to solve general obstacle and free boundary value problems. The uniqueness of the solution for the direct parallel method is established under the assumption that the model problem yields an $M$-matrix. The convergence analysis of the indirect approach is predicated on minimizing the discrete energy functional at each iterative stage of the linear approximation process. Under the aforementioned setting, we establish theoretical convergence results for both smooth and nonsmooth energy functionals defined over a convex set, in the sense of Ferris and Mangasarian \cite{ferris1994parallel}. For the numerical computation of the one-dimensional obstacle problem, we adopt a direct generalized parallel approach based on the SPIKE algorithm \cite{polizzi2006parallel}. To accelerate convergence and reduce computational complexity, a fast recursive version of the scheme is generalized to the nonsmooth case. For two- and three-dimensional obstacle problems, we employ a directional splitting method that treats each directional subproblem as a one-dimensional obstacle minimization problem. By framing the energy functional minimization as a parabolic time-dependent problem and utilizing an Armijo time-stepping rule during the solution update process, we successfully obtain results for higher-dimensional obstacles. Additionally, this study explores the possibility of extending the algorithm into a constrained quadratic programming optimization solver. We also evaluate the framework on image deblurring phenomena under the aforementioned setting, successfully recovering the original images. Finally, numerical illustrations are provided to validate the theoretical results.
Read the full abstract

Computation and Language

LLMs Learn Better In-Context from Rules than from Examples

Large language models (LLMs) exhibit in-context learning capabilities, where they can learn new tasks from prompt contexts without weight updates. We compare the learning efficacies of two prominent modes of in-context learning: (1) learning from descriptions of rules (instruction following); and (2) learning from examples of input-output demonstrations (few-shot prompting). Through five learning tasks that cover diverse domains (games, arithmetic, linguistic inferences), we compare two modes of learning (rules vs. examples) specifying the same underlying task. We furthermore explore model and task properties that modulate the learning efficacies. We find that models generally learn more reliably from rules than from examples alone, and additional examples on top of rules or simply scaling up the number of examples do not lead to consistent and significant gains. Instruction tuning amplifies the benefit of rule-based learning while keeping example-based learning capacities intact. Surprisingly, we find no privileged effect of example-based learning in base models, and rules still lead to gains in algebraic task domains. Overall, the comparative efficacy of rules over examples is larger when the task recruits algebraic abstractions and computations, and smaller when the task requires distributional sensitivity and/or recruits parametric knowledge.
Read the full abstract

All authors Xiang Fu, Seungmin Cho, Yukyung Lee, Najoung Kim

Cell Behavior

Coarse-Graining Agent-Based Models of Bacterial Infections

Agent-based models (ABMs) provide a natural framework for representing cell-level rules and spatial heterogeneity in bacterial infections, but their computational cost limits their use for macroscopic tissue-scale simulations and broad parameter exploration. We derive a deterministic coarse-grained description for a class of bacterial-infection ABMs in which immune cells and extracellular bacteria diffuse, immune cells ingest nearby bacteria, and intracellular bacterial loads evolve through prescribed birth and clearance processes. The first coarse-grained model is a semidiscrete reaction–diffusion system that retains a discrete internal state for each immune-cell bacterial load while representing cell and bacterial populations by continuum concentration fields. The key technical step is the derivation of state-dependent effective ingestion rates from the microscopic ABM parameters: these rates are obtained by solving an auxiliary diffusion problem around a single bacterium and computing the flux of immune cells into the interaction region. We then take a continuum limit in the internal state variable, yielding a state-structured reaction–diffusion system in which intracellular dynamics appear as advection and diffusion in state space. Numerical comparisons with ensemble-averaged ABM simulations show close agreement in biologically motivated parameter regimes. The resulting framework preserves the rule-based structure of the ABM while producing PDE models that are substantially more tractable for large-scale simulation and parameter studies.
Read the full abstract

Materials Science

Thermal history controls the optoelectronic response of lead halide perovskites through structure and dynamics

Lead halide perovskites are promising optoelectronic materials for photovoltaics, light emission and detection. Their efficiencies in PV now approach the detailed-balance limit, leaving stability as the principal barrier. The intrinsic instabilities studied to date centre on ionic motion within a fixed, homogeneous lattice. Here we identify a further source of intrinsic structural instability, hidden in the lattice dynamics. Mapping caesium, methylammonium and formamidinium-based compositions with Cl, Br, I and mixed X-sites through all accessible phases, using single crystal X-ray and neutron diffuse scattering, machine-learning-assisted molecular dynamics, a phenomenological octahedral tilt model and hyperspectral photoluminescence, we find that nearly every composition hosts equilibrium local structural fluctuations: dynamic nanodomains of correlated octahedral tilts, a few nanometres in size, that locally break the crystallographic symmetry. Three complementary levers control them. The A-site cation sets their symmetry, shape and anisotropy, from sparse, isotropic and tetragonal in formamidinium-based compositions to dense, anisotropic and orthorhombic in nominally cubic caesium-based ones, the most locally disordered we studied. The halide controls the dynamic disorder and the phase-transition sequence. Thermal history is the third: different ramp rates drive nominally identical compositions into distinct crystallographic phases, each with its own hidden local order. In MAPbI3, the heating rate alone changes the photoluminescence quantum efficiency across the phase transition. Because these transitions lie within device operating ranges, from terrestrial thermal cycling to the extremes of space, thermal history may shape the local structure, and hence the optoelectronic response, throughout fabrication and operation, establishing it as a design variable alongside composition.
Read the full abstract

All authors Milos Dubajic, Xia Liang, Johan Klarbring, Yang Lu, Thomas A. Selby, Erik Fransson, Philippe Holzhey, Benjamin M Gallant, Qichun Gu, Ganbaatar Tumen-Ulzii, Khasim Saheb Bayikadi, Isaiah Gilley, Martin v. Zimmermann, Christian Orr, Chwenhaw Liao, Josh S. Moon, Jacek Jasieniak, Makhsud Saidaminov, Michael P. Nielsen, Tom Wu, Stephen P. Bremner, Anita Ho-Baillie, Julia Wiktor, Paul Erhart, Steve Albrecht, Mercouri Kanatzidis, Henry J Snaith, Aron Walsh, Samuel D. Stranks

Atmospheric and Oceanic Physics

Improving global precipitation forecasts with an AI weather model trained on satellite observations

Precipitation forecasts shape decision-making across the global economy, particularly in sectors such as agriculture. However, unlike variables such as temperature, precipitation is highly intermittent and localized, making it difficult to forecast. While recent advances in AI weather prediction systems have enabled them to surpass physical models on globally averaged metrics, improvements in mean error rarely translate to actionable forecasts of severe flooding or dry crop fields. Furthermore, most of these models are trained and evaluated against a reanalysis data product, ERA5, which has well-known biases. Here we retrain AIFS, ECMWF's widely-used, open-source operational 0.25{\deg} probabilistic graph-transformer weather model, on satellite-based precipitation observations. Our model, Laxmi, improves global probabilistic accuracy by 19% and resolves systematic distributional biases in ERA5. Specifically, Laxmi reduces drizzle overprediction by 33% for amounts less than 3 mm per day. It also mitigates extreme rainfall underprediction, improving the global 95th percentile Brier skill score by 57%. Across a case study of 10 Indian tropical storms, Laxmi delivered the most accurate forecast of 150 mm event-total precipitation in 7 events, compared to 1 for AIFS and 2 for the leading physical model, IFS. Our results demonstrate that incorporating observation-based precipitation data directly into training can substantially improve forecasts.
Read the full abstract

All authors Julian F. Schmitt, Bertrand Delorme, Robert C. King, Yashica Patodia, Tapio Schneider, Aditi Sheshadri, Ravi Jain

Artificial Intelligence

MasterControl Seventeen Every Time

We study a governed approach to enterprise analytics: a language model interprets the question, while deterministic policy selects and runs a pre-approved analytical program that returns both results and evidence. We show that this restriction can remain expressive within a defined analytical class, using relational operations plus aggregation, comparison, windows, ranking, and similarity. Fixed meaning, policy, data, and execution rules also make results replayable. Across 440 runs, three 8B models generated SQL and selected tools at runtime, while Qwen3-8B interpreted intent only and policy executed the approved program. None of 330 runtime-planning episodes matched the full answer-and-evidence contract across all test datasets; the policy-executed analyzer matched 110 of 110. This is a configuration-specific result, not evidence that runtime agents cannot succeed under other designs.
Read the full abstract

Statistics Theory

G-Exponential Families through Probability Coordinates

We develop a probability geometric construction of generalized exponential families based on probability coordinate charts. A classical exponential family on the unit interval is transported to value space through a chart, producing what we call a G-exponential family. The construction changes the geometry in which exponential structure is represented while leaving the exponential function itself unchanged. This distinguishes it from the algebraic deformations used in the Tsallis and Kaniadakis frameworks. The transported families inherit the structure of the probability coordinate framework. The probability coordinate of each family member follows exactly the original exponential family on the unit interval. Initial Kolmogorov moments are therefore obtained by pulling classical coordinate moments back to value space. For the canonical family, the probability barycenter is the pullback of the mean parameter, while the Fisher information coincides with the Kolmogorov variance. Tail behaviour is inherited from the chart. Every family member is tail equivalent to the chart density, so heavy tails arise from geometry and remain unchanged in order under probability coordinate tilting. This is the opposite of classical exponential tilting. The sufficient statistic is bounded by construction, maximum likelihood reduces to moment matching in probability coordinates, and the maximum likelihood estimator has a bounded influence function. Thus robustness arises from geometric compactification. We also characterise G-exponential families as minimisers of relative entropy to the chart law under coordinate moment constraints and prove that transport preserves the information geometry of the family. A Cauchy chart example and a Monte Carlo study confirm efficient, calibrated, and bounded influence estimation even though every family member has an infinite mean.
Read the full abstract

Signal Processing

Underwater Acoustic Channel Library

The development of communication systems critically depends on realistic channel models, yet there are no widely accepted standards in the underwater acoustic communications community. The situation is in stark contrast to terrestrial radio communications, where channel models have been standardized and are widely available. To address this gap, we present an open-access library of underwater acoustic channels derived from field experiments conducted across geographically distinct locations and varying propagation conditions, including shallow and deep water, short and long range, and fixed and mobile platforms. Each channel is described by a time-varying impulse response extracted from at-sea recordings using an adaptive algorithm that separately identifies the multipath structure and the Doppler-induced phase and delay drift. Each channel is also accompanied by a site-specific ambient noise model, which captures the statistics of colored Gaussian noise and impulsive noise. Spatial diversity reception across an array of hydrophones is supported for most channels, while time diversity is included for single-hydrophone scenarios. The library is accompanied by a simple replay interface through which a user supplies an arbitrary transmit signal, passes it through a chosen channel, adds noise at a desired signal-to-noise ratio (SNR), and obtains the received signal. The models are validated by comparing the output SNR of a baseline receiver operating on replayed signals with its performance on the original at-sea recordings, demonstrating close agreement across all channels. The library, including all channel impulse responses, noise model parameters, and replay software, is freely available for download as an open-source package.
Read the full abstract

All authors Zhengnan Li, Mandar Chitre, Diego A. Cuji, James Preisig, Andrew C. Singer, Milica Stojanovic, Paul van Walree

Computer Vision

Learning to Zoom Efficiently with a Contrastive Curriculum

Using a zoom-in tool is an important foundational part of modern visual agents, because it allows to efficiently handle tasks involving high-resolution images. Most previous methods need an extensive warm-start supervised fine-tuning phase for teaching models zoom-in. We show that this is not necessary by proposing a new intrinsic reward for learning tool use in MLLMs without the need for additional labels or warm-start SFT. Our InfoNCE-style reward uses a curriculum of increasingly hard negative tool calls as a contrastive training signal. Empirical experiments on $V^*$, HRBench and MME-RealWorld show that our approach is competitive while being more efficient. When used as a drop-in replacement for SFT, we even outperform all baselines. To directly measure the zoom-in ability of models, we further introduce the scalable synthetic Muffin&Chihuahua (M&C) dataset. Each image consists of a grid with every cell either showing a muffin or chihuahua. Leveraging the M&C dataset's unique region of interest labels, we find that recall is the metric that most strongly correlates the zoom-in region with final task performance. Our model and code for reproduction is publicly available under https://github.com/UKPLab/emnlp2026-zoom-in
Read the full abstract

Information Theory

Variational Probabilistic Quantization for Secret Key Generation

Secret key generation from correlated observations at Alice and Bob, in the presence of an eavesdropper Eve, underpins physical-layer security. Classical pipelines quantize by hand, amplify privacy afterwards, and optimize no objective tied to a key rate. We propose Variational Probabilistic Quantization (VPQ): neural encoders that map the correlated sources directly into a discrete key alphabet, trained by a variational adversarial objective whose entropy, mismatch, and leakage terms match the three terms of the one-way secret key rate. A linear code-offset secure sketch then reconciles the encoder outputs into an identical key without a separate privacy amplification step. We prove that the VPQ losses lower-bound the one-way secret key capacity of the induced source, and derive in closed form the optimal worst-case key rate over the source class of a given alphabet size and mismatch probability, attained by finite-field linear sketches. On Gaussian fading channels, VPQ leaks less to a correlated eavesdropper than one classical and two recent learning-based baselines, and Reed–Solomon reconciliation operates within the predicted finite-blocklength gap.
Read the full abstract

All authors Xinyang Li, Vlad C. Andrei, Peter J. Gu, Yiqi Chen, Ullrich J. Mönich, Holger Boche

High Energy Physics (Phenomenology)

Quark Mixing Model with Three Higgs Doublets and $S_3$ Modular Symmetry at $\tau=i$

We study the quark sector of a model with three Higgs electroweak doublets, complemented with a modular $S_3$ symmetry. We show that a proper assignment of the quark and Higgs fields in $S_3$ and their modular weights allows to write a mass matrix with a phenomenologically viable texture. Moreover, by considering the conditions on the Higgs VEVs derived from the minimization of the scalar potential together with evaluation of the Yukawa couplings at the fixed point $\tau=i$, leads to a significant reduction of the free parameters. We present two setups, one with three effective parameters, which is overly constrained, and one with four effective parameters, where the $V_{CKM}$ matrix can be fitted precisely.
Read the full abstract

Sound

VoxReason: Auditing Source-Grounded Speech Plans Before Synthesis

Plan accuracy alone cannot show whether a speech-delivery decision follows its source: a fixed prior may match the original label yet fail to respond appropriately when a cue changes. VoxReason provides a 100-case verifier benchmark that holds each utterance fixed, edits one designated source-label cue, and scores cited evidence, eight plan fields, and the permitted response. On a source-key-disjoint test of 24 cases, a source-emotion prior reaches plan-slot accuracy 0.958, but none of the 24 edited neutral targets appears in its training labels; its required-change accuracy is 0.000. This diagnoses the support boundary of this prior, not its performance on supported edits. In a complementary 32-case emotion-disjoint test, the prior has seen all edited neutral targets but neither original test emotion; its plan-slot accuracy is 0.219 and required-change accuracy is 1.000. The partitions reuse and overlap the same 100 cases, so these deterministic diagnostics are not independent cohorts or learned-planner results. The benchmark evaluates derived labels and structured plans, not audio input, generated speech, or listener judgments.
Read the full abstract

Information Theory

Adaptive Beam Hopping and Power Control for Dual-Layer Over-the-Air Online Federated Learning in LEO Satellite Networks

This paper investigates over-the-air (OTA) computation enabled online federated learning (FL) in low-Earth orbit (LEO) satellite networks. Specifically, we consider a dual-layer OTA aggregation architecture, where ground devices upload analog model updates to serving satellites via uplink OTA aggregation, and satellites forward the aggregated signals to a data processing center through the second round OTA aggregation. Then, we formulate a long-term data-utilization maximization problem in which devices continuously collect new data and untrained samples gradually lose freshness. The problem is subject to the satellite beam budget, transmit-power limit, and global mean squared error (MSE) constraint that governs end-to-end aggregation distortion. This yields a coupled mixed-integer nonlinear programming (MINLP) problem, involving tightly coupled discrete beam-hopping decisions and continuous power control. Due to the combinatorial action space and nonconvex constraints, the problem is NP-hard and computationally intractable. Furthermore, the time-varying satellite topology and dynamic data generation render it a sequential decision-making problem, necessitating adaptive online scheduling. To address these issues, we cast the problem as a Markov decision process and develop a proximal policy optimization (PPO)-based deep reinforcement learning framework that jointly optimizes adaptive beam hopping and power control, using an MSE-aware reward to balance data utilization and aggregation accuracy. Numerical simulation results verify that the proposed algorithm consistently outperforms other benchmark schemes, achieving superior long-term data utilization and faster FL convergence while satisfying the MSE requirement.
Read the full abstract

All authors Zhendong Li, Shaojie Wang, Zhou Su, Zihao Zhang, Haixia Peng, Nan Cheng, Ying Wang, Wen Chen

Computation and Language

MemoryLACE: Memory Lifecycle-Aware Consolidation and Evidence Retrieval

Long-term LLM agents must preserve information across interactions while distinguishing repeated evidence, historical states, updates, and unresolved contradictions. Existing textual memory systems retrieve semantically relevant memories efficiently but often leave these relationships implicit, whereas richer structured approaches model them through global graphs, hierarchical abstractions, or reflection at greater complexity. We introduce MemoryLACE (MemLACE), a lightweight memory framework that explicitly models the lifecycle of textual evidence through sparse merge, supersession, and contradiction relations while preserving atomic natural-language memories and their provenance. Rather than retrieving memories independently, MemLACE reconstructs relation-aware evidence units that expose current, historical, supporting, and conflicting evidence for downstream reasoning. Across BEAM and StructMemEval, using open-weight and proprietary LLM backbones, MemLACE achieves the highest overall performance in same-backbone comparisons while reducing end-to-end runtime on BEAM by 66.6% relative to Hindsight, the strongest reported reflective-memory baseline. Ablation studies identify lifecycle expansion and temporal awareness as the principal contributors to these gains. Together, the results demonstrate that explicitly modeling the local lifecycle of textual evidence is sufficient to substantially improve long-term memory reasoning without requiring comprehensive knowledge graphs or global reflection.
Read the full abstract

All authors Meriem Yacoubi, Pia Schmidt, Nenad Petrovic, Ahmed Frikha, Martin Kirchhoff, Alois Knoll

Functional Analysis

Stability of maximal relative projection constants

For positive integers $n\ge r$, let $\lambda(r,n)$ denote the maximal relative projection constant of $r$-dimensional subspaces of $\ell_\infty^n$ and $\lambda(r)$ denote the maximal absolute projection constant, respectively. It is known that for any fixed $r$, $\lambda(r,n)$ is a non-decreasing sequence with limit $\lambda(r)$ as $n\to \infty$. A natural question is whether $\lambda(r,n)$ stabilizes at $\lambda(r)$ for some $n>r$. We prove that for any fixed $r$, \[\lambda(r,n)=\lambda(r) \qquad \text{for every}\qquad n\ge 2^{r}\binom{r+1}{2}.\] This answers a question of Basso. The technique used is of independent interest.
Read the full abstract

All authors Hitesh Kumar, Bojan Mohar, Seyed Ahmad Mojallal, Shivaramakrishna Pragada

Computer Vision

RoboTok: A Scalable Data Engine for Internet Demonstration Video Retrieval and Dexterous Manipulation Learning

Robot learning increasingly depends on broad and diverse demonstrations, yet collecting robot data remains expensive and difficult to scale across the wide range of real-world tasks. To address this bottleneck, we introduce RoboTok, a scalable data engine that uses a query human manipulation video to retrieve manipulation-relevant internet demonstrations for training dexterous robot policies. Specifically, we learn a latent motion space from 3D hand trajectories expressed in estimated actor-centered reference frames. This representation enables manipulation behaviors to be compared across variations in camera viewpoint, scene appearance, and actor occlusions, while remaining compact enough for efficient indexing and retrieval over internet video collections. We evaluate RoboTok against existing robot-data retrieval approaches using retrieval metrics and downstream robot policy performance, showing that RoboTok retrieves more manipulation-relevant demonstrations and improves downstream task success. In real-world robot experiments, RoboTok-guided policies achieve a mean success improvement of 42.2 percentage points over the strongest baseline for each task, establishing hand-pose trajectory-aware retrieval as a scalable way to leverage continuously growing web video for robot learning.
Read the full abstract

All authors Howard Qian, Yiting Chen, Yunfei Xie, Kejia Ren, Podshara Chanrungmaneekul, Gaotian Wang, Bowen Wen, Chen Wei, Kaiyu Hang

Complex Variables

A Smooth Counterexample to the Trautman Conjecture

The Trautman conjecture asserted that a smooth three-dimensional CR manifold admitting a nowhere-zero closed section of its canonical bundle must be locally embeddable in $\mathbb{C}^2$. We modify a standard construction of nonembeddable smooth strongly pseudoconvex CR 3-manifolds so that the condition of having a nowhere-zero closed section of the canonical bundle is preserved, thus providing a strongly pseudoconvex counterexample to the Trautman conjecture.
Read the full abstract

Mathematical Physics

Computational Algorithms for Invariant Reduction of Variational Forms

Symmetry reductions of partial differential equations (PDEs) inherit more geometric structures than other types of reductions: invariant conservation laws, variational structures, and, under suitable conditions, Hamiltonian-type structures of the original model descend to the reduced model through the mechanism of invariant reduction. This paper develops computational algorithms that carry out such reductions explicitly. We use the interpretation of variational $p$-forms as conservation laws of an enlarged system — the (degree-shifted) tangent system — consisting of the original equations together with their linearizations, in which the perturbation variables are treated as anticommuting. This allows us to formulate the algorithms using well-known concepts from the theory of conservation laws, naturally adapted to the graded-commutative setting. The reduction of conservation laws, variational $1$-forms, and presymplectic structures thereby becomes a single algorithmic procedure. We present (i) a homotopy-based reduction algorithm for systems of evolution equations, implemented in Maple; (ii) a descent reduction algorithm applicable to general $\ell$-normal systems for $p>0$; and (iii) a simple reduction algorithm available for point symmetries under suitable conditions, including the description of reductions in terms of systems involving fewer independent variables. Examples include nonlinear evolution equations in one and two spatial dimensions, the Laplace equation, the incompressible Euler equations, and the cotangent system of Pavlov's equation.
Read the full abstract

Optics

Grammar-based topology search for reconfigurable integrated photonic spectrometers

Reconfigurable integrated photonic spectrometers can generate exponentially increasing measurement states from a small number of switching elements, making them a promising design class for high-performance portable spectroscopy. However, selecting a circuit architecture remains largely intuition driven. We introduce a grammar-based topology optimization framework that composes standard photonic building blocks, canonicalizes and physically prunes candidate circuits, and jointly searches topology and component parameters using a coherent scattering-matrix model. Designs are screened by a decoder-independent architecture objective that combines absolute singular-value noise gain with multi-line conditioning, thereby accounting for response diversity and optical throughput. Direct held-out spectrum reconstruction confirms that this surrogate is a strong predictor of reconstruction quality at high incident signal-to-noise ratio (SNR). Across the nonresonant and resonator-augmented design spaces examined, dFT and its variants offer the best balanced performance; increased component loss favors a shallower dFT variant in which one differential-delay stage is replaced by an analog optical phase shifter. Passive and active rings provide no robust advantage attributable to resonant amplitude coding. The results identify balanced complementary interferometric responses, efficient terminal collection, and limited lossy circuit depth as central design rules, and establish a systematic route for selecting task-optimized photonic circuit topologies under technology-specific constraints.
Read the full abstract

Combinatorics

Lecture hall polytopes and Lakshmibai-Seshadri paths

Using a bijection between the lattice points in a lecture hall polytope and Lakshimibai-Seshadri (L-S) paths, we prove the Koszul property of lecture hall polytopes in complete generality, and give new proofs for their Integral Decomposition Property and for a criterion on Gorenstein property.
Read the full abstract

All authors Rocco Chirivì, Martina Costa Cesari, Xin Fang, Peter Littelmann

Quantum Physics

Experimental validation of a compact fault-tolerant architecture for trapped ions

Quantum error correction (QEC) is beginning to enable logical operations that outperform their unencoded physical counterparts, but useful fault-tolerant computation will require more than low-error quantum memory. An effective architecture must orchestrate efficient logical encoding, low-overhead logical operations, and access to the non-Clifford resources required for universal computation. Here, we introduce and experimentally validate such an architecture based on the $[[20,2,6]]$ $C_4$-Helix code, designed for the early fault-tolerant regime. Using Quantinuum Helios, a 98-qubit trapped-ion quantum processor, we experimentally demonstrate the principal components of this architecture: we perform repeated quantum error correction with an error of $4.6^{+6.2}_{-2.6}\times10^{-5}$ per logical qubit per QEC cycle. We benchmark the complete Clifford group on the two logical qubits of a single codeblock under active error correction, obtaining an error of $2.8^{+1.0}_{-1.6}\times 10^{-4}$ per two-qubit logical Clifford. We further demonstrate a fault-tolerant chain-map interface between $C_4$-Helix and a distance-5 surface code, preparing a heterogeneous three-logical-qubit GHZ state with a fidelity lower bound of $99.925^{+0.068}_{-0.245}\%$. In each case, the encoded implementation outperforms its corresponding unencoded physical baseline without relying on postselection. Circuit-level simulations indicate that improvements in physical fidelity bring the same architecture into the $10^{-6}$-$10^{-8}$ logical-error regime targeted for early fault-tolerant computation. Together, these results establish $C_4$-Helix as a hardware-validated fault-tolerant architecture rather than a bare quantum memory.
Read the full abstract

All authors Noah Berthusen, Ali Lavasani, Asmae Benhemou, M.S. Allman, Joan Dreiling, Brian Estey, Cameron Foltz, Trent Jacobs, Michael Mills, Annie Jihyun Park, Adam P. Reed, David Hayes, Tzvetan S. Metodi, Andrew C. Potter

Statistical Mechanics

Generative Nested Sampling of Atomistic Thermodynamic Landscapes

Nested sampling (NS) resolves the thermodynamics of an atomistic system from a single simulation, but its practical reach is limited by the Markov-chain updates needed to decorrelate walkers within each likelihood-constrained ensemble. Flow-based NS has reduced this bottleneck for gravitational-wave (GW) inference, yet its transfer to atomistic systems is not merely a change of application. Comparing a GW150914-like binary-black-hole likelihood with an eight-particle two-dimensional Lennard-Jones (LJ) system of comparable dimensionality, we show that the two landscapes differ fundamentally: atomistic multimodality is discrete and combinatorial, generated by particle permutations separated by hard collision walls, and its coordinate coupling is dense and collective, whereas the GW posterior exhibits smooth degeneracies and localized parameter coupling. Guided by this diagnosis, we introduce NS-Flows: a single conditional normalizing flow, conditioned on the NS energy bound and trained on a sliding window of recent live sets, that replaces MCMC by direct parallel draws corrected by importance-weighted rejection resampling. Live sets supply the data self-consistently, allowing flow training without structured priors or a pre-existing dataset. For LJ disks under periodic boundary conditions, the algorithm can reduce energy evaluations by two orders of magnitude and wall-clock time by about 30%, an advantage that becomes more favorable as the cost of the potential grows. The flow's generative efficiency further acts as a physical diagnostic: it varies non-monotonically along the annealing trajectory, is lowest in the dense disordered regime, and is quantitatively captured by the constrained ensemble's internal mode complexity together with target drift across the training window, identifying liquid-like ensembles, not prior-target separation, as the hard case for current flow architectures.
Read the full abstract

All authors Alessandro Coretti, Nico Unglert, Sebastian Falkner, Georg K. H. Madsen, Christoph Dellago