Peer Revenue Equity in science

Papers

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

All of arXiv

Wednesday 7 October

Robotics

Do My Samples Help? Compute Budget Allocation for Real-World Sampling-Based MPC in Contact-Rich Manipulation

Magnus Dierking, João Carvalho, Vince Kurtz et al. (+3 more)

Sampling-based model predictive control (SMPC) is increasingly practical for contact-rich manipulation because GPU-based physics simulators enable real-time evaluation of large numbers of non-differentiable contact rollouts. However, this computation budget is finite and can be spent in competing ways, including denser control exploration, representing dynamics or state uncertainty, replanning more frequently, or aligning the simulation better with the physical state. Yet, the resulting trade-offs are not well understood in real-world sim-to-real control. In this work, we address this question through a systematic study of how to allocate a fixed computation budget for SMPC in contact-rich manipulation. To enable controlled comparisons, we develop a GPU-based framework named Budget-Allocating Model Predictive Control (BAMPC), which runs on MuJoCo Warp and exposes dynamics uncertainty, state uncertainty, and temporal alignment as independent dimensions of the computation budget while using the same control stack in simulation and on a real robot. Across three contact-rich manipulation tasks, we find that the best budget allocation depends on the dominant sim-to-real error, with dynamics randomization not universally the most effective one.
Read the full abstract

All authors Magnus Dierking, João Carvalho, Vince Kurtz, Jan Peters, Nils Dengler, Georgia Chalvatzaki

Machine Learning

Auditing LLM Program Search for Bermudan Stopping with Primal–Dual Certificates

Yuhua Qian

LLM program-search agents are usually judged by the score that drives their search. We study LLM-driven search over Python exercise policies for Bermudan options, an optimal stopping problem where a trusted evaluator owns the simulation and a dual upper bound~$U$. For any policy, $0 \le V^\star - L_\pi \le U - L_\pi$; we report Monte Carlo estimates of both bounds with their sampling error. We audit 24 generation attempts from two pipelines, a six-pair feedback-representation ablation, and a holdout evaluated after selection was frozen. No selected generated program improves on Ridge LSMC consistently or by more than Monte Carlo uncertainty; holdout gap scores lie within $0.002$ of it. On six one-dimensional puts, a high-precision binomial reference places Ridge within Monte Carlo noise of optimal and attributes 99% of the estimated gap to dual-bound looseness, so this dual audits validity but cannot certify progress at the search's scale. Structured memory showed no benefit over score-only memory in this ablation. Whether search “worked” depends on reporting rules: retaining the initial reference, freezing selection before holdout, and counting failures.
Read the full abstract

Data Structures and Algorithms

Grover Search for a Threshold-Fraction Variant of the Travelling Salesman Problem: A Circuit-Level Construction

Aayra Dhindsa Sawkmie

The Travelling Salesman Problem (TSP) is a classic combinatorial optimisation problem which is NP-hard in optimisation form and NP-complete as a decision problem. The standard question asks for the shortest tour that visits every city exactly once and returns to the origin. This paper studies a threshold-fraction variant, which we call the robust TSP, formulated as a promise problem: given a distance threshold $K$ and a promised fraction $p$, decide whether at least a fraction $p$ of all tours have total length at most $K$, or whether no such tour exists. We show that Grover's quantum search algorithm decides this problem using $O(1/\sqrt{p})$ oracle queries, independent of the number of cities $N$, with each query implemented by a quantum circuit of polynomial size in $N$. For any fixed robustness level $p$, the total runtime is therefore polynomial in $N$. Measured in oracle queries, the quantum algorithm offers a quadratic improvement in the dependence on $1/p$ over classical random sampling, which needs $\Theta(1/p)$ samples; because each quantum query is costly, however, it is faster in total running time only when $p$ is very small (roughly below $N^{-4}$), where the number of queries is no longer constant. We describe the problem formulation, detail the oracle construction (edge-length loading, sequential quantum addition, and threshold comparison), and analyse the query complexity to achieve the required result.
Read the full abstract

Robotics

Terrain Traversal Without Motion Priors: Expert-Routed Humanoid Locomotion at Deployment Fidelity

Thanh Ly, Chien Le, Truong-Duy Dang et al. (+4 more)

State-of-the-art perceptive humanoid locomotion is mostly built on retargeted motion capture, a generative motion prior, and a distillation stage. We report a complete system that does not require this pipeline for a humanoid to cross varied terrain. In its place, we provide a five-expert mixture routed by a terrain classifier over a sensed elevation grid, and target every foothold with a model-based planner, refined by six closed-form-gradient steps on an interpolated terrain-cost field instead of picked from a fixed grid. Driven through the same perception-to-actuation software a robot would run, inside a second, deployment-fidelity simulator, it crosses an $85$\,m five-segment waypoint course $18$ of $20$ times end-to-end ($90\%$, Wilson $95\%$ CI $70$–$97$) and succeeds $20/20$ on four of five segments scored alone, where an ablated configuration without the mixture and without refinement reaches $0\%$ on both the stair and the hurdle segment. The continuous refinement moves $72.4\%$ of foot placements across at least one grid-cell boundary, which produces real sub-grid correction. We also ran matched component studies to report what does not account for it: one expert equals five on the training distribution, four label-free routers collapse to near-uniform activation, and a self-supervised terrain embedding loses to the privileged label.
Read the full abstract

All authors Thanh Ly, Chien Le, Truong-Duy Dang, Duy M. H. Nguyen, Duy-Dung Le, Vien Anh Ngo, An T. Le

Populations and Evolution

The Role of Tick Questing Behavior and Host-Finding Success in Borrelia burgdorferi Transmission in the Eastern United States

Carli Peterson, Erika Gallo and Christopher M. Kribs

The marked difference in reported human Lyme disease cases between the northeastern and southeastern United States remains a significant concern. Differences in biodiversity and the questing behavior of Ixodes scapularis, the primary vector of Borrelia burgdorferi, the causative agent of Lyme disease, have been proposed as key drivers of geographic variation in disease incidence. In this study, we investigate how these differences, together with ratio-dependent tick-host contact rates, contribute to the north-south gap in Lyme disease cases. Building on our previous work, we develop a discrete-time system of difference equations to model infection dynamics between ticks and their sylvatic hosts. We derive an expression for the basic reproductive number and vary key parameters to assess their impact on infection persistence. We further examine the effects of these mechanisms on nymphal infection prevalence (NIP) and the density of infected nymphs (DIN), two key metrics for predicting disease risk. Our results indicate that under baseline conditions, southern host communities exhibit lower infection prevalence and reduced potential for pathogen invasion, consistent with observed regional patterns. However, southern host communities can support sustained transmission with different questing behaviors. Notably, this qualitative north-south difference emerges using parameter values obtained from the literature, without calibrating the model to reproduce geographic differences in disease prevalence. In addition, ratio-dependent host-finding success shows a saturating effect on the basic reproductive number, compared to the traditional assumption of density-independent host-finding success. These results highlight the importance of biodiversity, tick questing behavior, and using multiple metrics when assessing Lyme disease risk.
Read the full abstract

Machine Learning

CircuitTree: Composing Circuits into A Syllable Transcriber

Jiankun Wei, Ewan Dunbar and Gerald Penn

The goal of circuit discovery is to decompose a monolithic neural network into tiny modular subnetworks which are interpretable in the sense that each does an identifiable subtask. We develop a practical method for composing arbitrarily many circuits into a single model. We propose a greedy, density-scaled composition algorithm that overcomes limitations of previous attempts to compose circuits by naive union, then generalize it to $N$ circuits via CircuitTree, a new data structure built on a $\Theta(N\log N)$ tree-compose schedule. We compose circuits pruned from HuBERT for classifying individual phonemes into a syllable classifier, a capability none of the 39 phoneme circuits has individually. Chaining CircuitTrees into a hierarchy-of-circuits pipeline, we transcribe open syllables at 64.8% accuracy, within 1.2 points of a 66.0% ceiling that shares no circuitry at all, with 99.5% of circuit components shared.
Read the full abstract

Physics and Society

Voluntary nonpharmaceutical intervention as a feedback-evolving game

Ryuichi Kumata, Hiromu Ito and Masato Yamamichi

Nonpharmaceutical interventions (NPIs), such as mask wearing, can reduce epidemic spread, but their effectiveness depends on whether people adopt and sustain them. The incentive to adhere changes with infection risk and with the behavior of others. Yet it remains unclear how epidemiological protection and social reinforcement determine the strategic structure of infection control, and how epidemic feedback changes that structure over time. Here, we develop a socio-epidemiological model in which prevalence changes behavioral incentives while adherence changes transmission, separating private protection from source control and incorporating conformity. At a given prevalence, private protection favors adherence, source control creates opportunities for free-riding, and conformity reinforces whichever behavior is already common, generating distinct adherence games. Because adherence reduces transmission and prevalence, the population can move between these games as the epidemic unfolds. This feedback can stabilize partial adherence, create alternative behavioral norms, or generate recurrent epidemic-behavior cycles. Moreover, interventions with the same population-level effect on transmission at a given adherence level can produce different long-term outcomes when protection is distributed differently between the individual and others. Thus, voluntary infection control is a feedback-evolving game whose outcome depends not only on epidemiological efficacy but also on who is protected and how behavior is socially reinforced.
Read the full abstract

Chemical Physics

Looking Under the Streetlight: Evaluation in Generative Molecular Dynamics

Simon Olsson, Frank Noé, Grant M. Rotskoff et al. (+1 more)

This position paper argues that the generative molecular dynamics (GenMD) community is not properly evaluating its progress: current evaluations let a model match what is reported while being arbitrarily wrong on what was claimed, which risks over-claiming advantage over classical baselines. This reflects a structural misalignment between common evaluation practices and the scientific tasks the field sets for itself. The reason is that the ground truth—exhaustive molecular dynamics on the systems of interest—is computationally costly, and this cost is what motivates the field. With ground truth inaccessible, evaluation drifts toward convenient proxies, and the resulting norms entrench easily. We identify misaligned evaluation patterns that persist as a community-level equilibrium, and propose concrete observables, reporting standards, and benchmark designs, both computational and experimental, that better reflect what GenMD aspires to deliver. Our aim is to give the field shared language and tools for evaluation that match its scientific ambitions before today's conventions calcify into norms.
Read the full abstract

All authors Simon Olsson, Frank Noé, Grant M. Rotskoff, Kresten Lindorff-Larsen

Computer Vision

TerraAgent: Demonstration-Grounded Multi-Agent Reasoning for Long-Context Remote Sensing Visual Question Answering

Jingwen Huang, Yiming Fang, Hanxiao Wu et al. (+2 more)

Remote sensing visual question answering (RS-VQA) requires answers that truly depend on image evidence, yet current remote sensing multimodal large language models (RS MLLMs) are still evaluated mainly by accuracy, which cannot reveal whether a correct answer was produced without using the image. This setting introduces three challenges: demonstrations retrieved across different ground sample distances (GSD), sensors, and city textures may be poorly matched; real queries often involve long visual contexts from multi-scale tiles, multi-temporal frames, and retrieved knowledge, where evidence can be diluted and errors can propagate through multi-step reasoning; and prior-only answering remains invisible to standard metrics. We present AVA, a training-free framework that wraps a frozen RS MLLM with three modules. GDR performs GSD-bucketed, scene-graph-based demonstration retrieval and filters demonstrations whose answers are unsupported by their images. MACE decomposes each query into an evidence-typed sub-question DAG, routes sub-questions to suitable perception tools, and records verified provenance in a shared evidence ledger with bounded re-perception. CVA audits the final answer through counterfactual perturbations of cited evidence, including ablation and swap, and computes a visual-dependency score for confidence calibration and evidence-forcing re-query. On RSVQA-LR/HR, VRSBench, and LHRS-Bench, AVA improves average accuracy by up to 2.40 points over the strongest same-setting baselines, raises the visual-dependency score from 0.28 to 0.49, reduces the prior-only answer rate from 46.2 percent to 21.7 percent, and lowers the hallucination rate from 17.1 percent to 9.8 percent, with 2.6x the cost of a single backbone forward pass and no fine-tuning.
Read the full abstract

All authors Jingwen Huang, Yiming Fang, Hanxiao Wu, Boyuan He, Keyi Luo

Computer Vision

TraceICL: Thread-Resolved Demonstration Curation with Consistency Calibration for Robust Multimodal In-Context Learning

Yigit Yilmaz, Elena Petrova, Mehmet Kaya et al. (+2 more)

Interleaved multimodal in-context learning (ICL) lets frozen large vision-language models (LVLMs) acquire new tasks from a handful of image-text-answer demonstrations, yet it remains fragile: demonstrations bind to the wrong images, redundant exemplars crowd out informative ones, and noisy answers are copied verbatim into predictions. We present AVA, a training-free framework that organizes the demonstration pipeline of multimodal ICL into three composable stages. A Thread-Resolved Decomposer (TRD) segments the interleaved context into per-sample threads and re-stitches them with explicit boundaries, restoring image-text binding. An Influence-scored Demonstration Selector (IDS) estimates each candidate's marginal influence with one extra forward pass, removes redundancy, and orders the survivors by descending value. A Weak-to-Strong Consistency Calibration (WSC) module scores each demonstration by a weak sibling model's reconstruction confidence and softly reweights its contribution during decoding. On five vision-language benchmarks with interleaved 4-shot prompts, AVA improves the strongest same-setting baseline by +1.7 points on average, with gains of up to +2.4 points on VizWiz, shrinks the require-visual performance gap on ScienceQA-IMG, and retains most of its gain under 30% noisy demonstrations, at a 26% latency overhead that an offline cache amortizes. No gradients are ever computed; the LVLM stays frozen.
Read the full abstract

All authors Yigit Yilmaz, Elena Petrova, Mehmet Kaya, Lucia Rossi, Amir Rahman

Software Engineering

CTML: A Comprehensive Framework for Continuous Testing in Machine Learning Operations

Bestoun S. Ahmed

The increasing use of machine learning (ML) systems in production has highlighted some important weaknesses in traditional software testing methods. Unlike regular software, ML systems can fail in specific ways due to their reliance on data, including issues like data drift, concept drift, and performance decline. To address these issues, this paper introduces the Continuous Testing for Machine Learning (CTML) framework. This framework is designed specifically for MLOps environments and offers a thorough, automated solution. It establishes a standardized way to evaluate ML systems across four main areas: data quality, drift detection, model quality, and pipeline validation. The CTML framework consists of a modular setup that includes a Test Registry, Test Runner, Result Store, and Integration Layer. This design ensures it integrates smoothly with existing Continuous Integration/Continuous Delivery (CI/CD) pipelines and MLOps platforms. We put this framework to the test in a predictive maintenance case study, which showed how effective it is at identifying issues like data drift, performance degradation, and bottlenecks in the pipeline problems that traditional testing methods might overlook. The findings indicate that CTML successfully uncovers critical failures in the system across all testing dimensions, providing useful insights to help keep ML systems running smoothly in production. Overall, this work presents a new, integrated framework that connects academic research with industry best practices, pushing forward the standards in MLOps quality assurance.
Read the full abstract

Computer Vision

CORAL-KV: Compact Cross-Layer Prediction and Residual Coding for Autoregressive Video KV Caches

Albert Ge, Akash Haridas, Utkarsh Saxena et al. (+5 more)

Autoregressive video diffusion models maintain a growing attention key-value (KV) cache, making memory a key bottleneck for long video generation. Recently, clustering-based quantization has emerged as an effective method to compress video KV caches by grouping similar key and value vectors then quantizing residual errors. However, when later frames attend to the compressed history, small reconstruction errors can compound as the video grows. Beyond the within-layer similarity exploited by clustering, we find strong dependencies across cache layers in video models. This motivates Coral-KV - Calibrated Offline, Residual-coded, and Adapted Locally - a KV-cache compression framework that uses cache states in each layer to predict keys and values in subsequent layers, then quantizes the remaining prediction residuals. We construct compact block-diagonal-plus-low-rank (block-DPLR) predictors: block-diagonal components model channel relationships within each attention head, while a global low-rank term models dependencies between heads. Coral-KV exposes a flexible fidelity-memory trade-off: it can reuse its predictors across videos or refit lightweight block-diagonal predictors to the current video for higher fidelity. Across extensive evaluations on both HY-WorldPlay and MAGI-1, our reusable configuration improves PSNR over QVG-4 by 0.32 dB and 4.95 dB, respectively, while using 48% and 37% less cache memory. Against QVG-Pro, our video-adapted configuration improves MAGI-1 PSNR by 3.77 dB while using 33.7% less cache memory. Coral-KV also maintains higher reconstruction fidelity over extended generation and reduces temporal discontinuities while retaining comparable generation latency.
Read the full abstract

All authors Albert Ge, Akash Haridas, Utkarsh Saxena, Ruijing Yang, Ian Colbert, Frederic Sala, Vikram Appia, Emad Barsoum

Combinatorics

Maximal Order Type and Width of the h-Preorder on Finite Labeled Trees

Ilya Smirnov

We establish tight bounds for the width and maximal order type of the h-preorder on Q-labeled trees in terms of the corresponding invariants of well-quasi-order Q.

Chaotic Dynamics

A Second-Order Asymptotic Expansion for Nonlinear Filtering in the Lorenz–63 System

Masahiro Kurisaki and Kota Takeda

A challenge in nonlinear filtering is to incorporate nonlinear effects into approximations of the conditional mean while retaining a finite-dimensional recursive algorithm. We derive a second-order asymptotic expansion of the filtering mean with respect to the system-noise intensity. We apply this expansion to the Lorenz–63 model, a prototypical three-dimensional chaotic dynamical system. The first-order approximation is computed using a Kalman–Bucy filter for the system linearized around the deterministic reference trajectory. The second-order correction incorporates the Hessian of the Lorenz–63 drift and is computed through a closed recursive system with 3+9+18=30 independent scalar components. The full system, including the reference trajectory and the first-order mean and covariance, has 42 independent scalar components. To support stable tracking of chaotic trajectories over long time intervals, we introduce recentering of the expansion and redistribution of Gaussian components between expansion orders. In the numerical experiments, we compare the proposed algorithm with an ensemble transform Kalman filter (ETKF) as a practical benchmark and a particle filter (PF) as a numerical reference. Under the full-observation settings, the proposed algorithm achieves a lower root-mean-square error relative to the truth and a smaller distance to the 10,000-particle PF reference than the 20-member ETKF. These results suggest that the proposed second-order extension of Kalman filtering has the potential to balance filtering accuracy and computational cost.
Read the full abstract

Quantum Gases

Spinodal-controlled freeze-out and nucleation at the first-order superfluid-supersolid transition of a dipolar gas

Wai Kui Wong

Dipolar quantum gases in elongated traps become supersolid through a transition that is continuous at intermediate density but first order at low and high density; at the continuous point, ramps of the scattering length follow Kibble-Zurek scaling. We study ramps across the first-order superfluid-supersolid transition of $^{164}$Dy in an infinite tube using the 3D extended Gross-Pitaevskii equation with truncated-Wigner noise. With quantum noise only, the uniform superfluid survives the metastable window and transforms after its spinodal, where the roton softens. For onsets occurring during the linear ramp, the delay measured from this spinodal has a fitted exponent $\zeta = 0.36 \pm 0.02$, close to the continuous-transition result and the mean-field value $1/3$. Reverse ramps show a comparable delay beyond the fold of the crystal branch, producing dynamical hysteresis. Initial thermal noise with nominal temperatures of 2.5-10 nK produces crystalline regions inside the metastable window during holds, with faster nucleation closer to the spinodal and at higher initial temperature. In ramps at 5 nK, thermal fluctuations make the superfluid fraction $f_s$ cross onset thresholds near $f_s = 1$ early, whereas the bulk of the crystal still forms after the zero-temperature spinodal with a similar scaling. Without a heat bath these effective rates depend on preparation: ramps crossing the window at the same rate nucleate early in 2 of 32 or 11 of 16 realizations depending on their starting point. At low density, slower ramps select larger spacings in a crystal of separated droplets, and droplets survive far above the transition on reverse ramps. A Maxwell construction gives a coexistence window $0.037 a_0$ wide around the first-order transition point, in which localized states exist; in the simulated tube they lie above the homogeneous states. At the continuous point there is no coexistence.
Read the full abstract

2610.12511 · 7 Oct · 15 pages, 7 figures. v3: presentation fixes (notation, table layouts, Figs. 1, 3, 4 redrawn), references on helium cavitation added; no results changed. v2: new appendices on the coexistence window and localized states and on thermal effects on the onset; corrected the description of droplet growth during holds; two references added

Abstract PDF $XTAL ×3$QNTM ×2

General Physics

On the notion of kinetic energy in a spacetime of maximal proper acceleration

Ricardo Gallego Torromé

It is shown in this work that in spacetimes of maximal proper acceleration there are situations where a particle decelerates and simultaneously increases its kinetic energy. To illustrate the effect, a classical $1$-dimensional electric cavity-like scheme is discussed. The gain in kinetic energy by deceleration suggests an equivalence between velocity and acceleration from the energetic point of view. Such an equivalence is stated in a form resembling Born reciprocity principle but formulated in the framework of higher order jet spacetime geometry. We also show the existence in the theory of a dynamical superluminal sector for point particles with positive mass and characterized by superluminal speed and a proper acceleration larger than the maximal proper acceleration. Therefore, the speed of light and the maximal proper acceleration become only separation conditions at the classical level of the two dynamical sectors.
Read the full abstract

High Energy Physics (Experiment)

Short-Baseline Near Detector (SBND): Design and Initial Performance

SBND Collaboration: P. Abratenko, R. Acciarri, C. Adams et al. (+351 more)

The Short-Baseline Near Detector (SBND) is a 112 tonne active mass liquid argon time projection chamber (LArTPC) situated in the Booster Neutrino Beam at Fermilab, forming part of the Short-Baseline Neutrino (SBN) Program. SBND began operation in 2024 and has since accumulated the world's largest sample of neutrino-argon interactions. This paper describes the as-built detector and presents performance results from the first year of beam operation. All detector systems performed at or above design requirements. The liquid argon purity achieved an electron lifetime exceeding 30ms, ten times the design requirement, for more than 95% of Run 1. The TPC wire plane readout achieved noise levels at the theoretical minimum for the given wire geometry (an average of 388e$^-$ for the 4m long collection plane wires) with 99.3% of readout channels live. Together, the liquid argon purity and TPC performance are enabling extremely clean event reconstruction in the SBND TPC. The photon detection system, consisting of passive reflectors and multiple active components (PMTs and X-ARAPUCAs), has achieved a light yield exceeding 10 photoelectrons per MeV throughout the detector volume and a timing resolution of a few nanoseconds, sufficient to resolve the 19ns bunch structure of the Booster Neutrino Beam. A nearly 4$\pi$ cosmic ray tagger system surrounds the TPC to mitigate cosmic backgrounds in neutrino and BSM physics analyses. The data acquisition system has demonstrated stable performance up to an event rate of 8Hz, and a data collection efficiency of 98.6% was achieved throughout SBND's first neutrino run in 2025. These results demonstrate that SBND represents the state of the art in multi-tonne-scale LArTPC detector performance.
Read the full abstract

All authors SBND Collaboration: P. Abratenko, R. Acciarri, C. Adams, T. Alafaro, L. Aliaga-Soplin, R. Alvarez-Garrote, J. Ameel, D. Andrade Aldana, C. Andreopoulos, A. Antonakis, L. Arellano, J. Asaadi, W. Badgett, L. Bagby, S. Balasubramanian, D. Barker, A. Barnard, C. Barnes, N. Barros, A. Basharina-Freshville, V. Basque, J.R. Bateman, M.C. Bazetto, A. Beever, B. Behera, E. Belchior, O. Beltramello, M. Betancourt, A. Bhat, K. Biery, M. Bishai, A. Bitadze, A. Blake, B. Bogart, J. Boissevain, C. Bonifazi, J. Book Motzkin, G. Botogoske, D. Brailsford, A. Brandt, J. Bremer, S. Brickner, T. Brooks, M.B. Brunetti, G. Buccino, A. Bullock, G.A.N. Buzzard, L. Camilleri, D. Caratelli, D. Carber, B. Carlson, M.F. Carneiro, M. Cascella, R. Castillo Fernandez, F. Cavanna, M. Chalifour, P. Chambouvet, A. Chan, A. Chappell, H. Chen, S. Chung, D. Cianci, M.F. Cicala, R. Coackley, L. Cooper-Troendle, J.I. Crespo-Anadon, E. Cristaldo, C. Cuesta, Y. Dabburi, O. Dalager, M. Dall'Olio, R. Darby, I. de Icaza Astiz, M. Del Tutto, S. Dennis, V. Di Benedetto, S. Dixon, M. Donnachie, Z. Djurcic, V. do Lago Pimentel, S. Dominguez-Vidales, R. Doubnik, M. Dubnowski, K. Duffy, S. Dytman, A. Elvin, A. Ereditato, E. Espinoza, J. Estrada, J.J. Evans, A.C. Ezeribe, C. Fan, J. Farrell, A. Filkins, B. Fleming, W. Foreman, D. Franco, H. Frandini, J. Freestone, G. Fricano, J. Fried, D. Friend, I. Furic, A.P. Furmanski, T. Gamble, S. Gao, D. Garcia-Gamez, S. Gardiner, G. Ge, M. Geynisman, I. Gil-Botella, S. Gollapinni, M. Goodman, O. Goodwin, P. Green, H. Greenlee, W.C. Griffith, R. Guenette, P. Guzowski, L. Hagaman, T. Ham, A. Hamer, P. Hamilton, R. Hanni, K. Hardin, B. Harris, S. Hentschel, A. Hergenhan, M. Hernandez-Morquecho, C. Hilgenberg, C. Hill, A. Hoffman, P. Holanda, A. Holin, B. Howard, B. Howe, J. Howell, J. Hrivnak, C. James, C. James, T. Johnson, R.S. Jones, M. Jung, T. Junk, Y.-J. Jwa, D. Kalra, G. Karagiorgi, L. Kashur, K.J. Kelly, W. Ketchum, M.J. Kim, M. King, J. Klein, L. Kotsiopoulou, S. Kr Das, I. Kreslo, T. Kroupova, S. Kubecki, V.A. Kudryavtsev, F. Kumar, M. Langstaff, J. Larkin, K. Laureto, H. Lay, R. LaZur, G. Leigh, I. Lepetic, J.-Y. Li, K. Lin, B.R. Littlejohn, F. Liu, L. Liu, S. Lockwitz, D. Lorca, W.C. Louis, G. Lukhanin, X. Luo, A.A. Machado, P. Machado, G. Mahler, S. Manohari Kancharla, S. Marcocci, C. Mariani, F. Marinho, J. Marshall, L. Martinez, C. Martin-Morales, A. Mastbaum, K. Mavrokoridis, N. McConkey, B. McCusker, J. Mclaughlin, V. Meddage, D. Mendez, I. Mercer, T. Mettler, T. Miao, M. Micheli, G. Miller, D. Mishins, K. Mistry, D. Mladenov, A. Mogan, J. Molina, D. Montanari, M. Mooney, A.F. Moor, G. Moreno-Granados, F. Morgadinho-Inacio, C.A. Moura, J. Mousseau, J. Mueller, S. Mulleriababu, J. Najdzion, R. Natieli de Oliveira, A. Navrer-Agasson, M. Nebot-Guinot, M. Nessi, V.C.L. Nguyen, S. Ni, T. Nichols, F. Nicolas-Arnaldos, B. Norris, R. Northrop, J. Nowak, S.B. Oh, O. Olorife, N. Oza, L. Pagliuso, O. Palamara, N. Pallat, V. Pandey, A. Papadopoulou, H.B. Parkinson, J. Pater, J. Paton, L. Paulucci, Z. Pavlovic, D. Payne, L. Pelegrina-Gutierrez, O.L.G. Peres, R. Pfaff, F. Pietropaolo, J. Plows, J. Poirot, I.D. Ponce Pinto, A. Popov, F. Psihas, D. Pushka, G. Putnam, X. Qian, Y. Qie, V. Radeka, E. Raguzin, R. Rajagopalan, P. Ratoff, H. Ray, B. Rebel, R. Rechenmacher, M. Reggiani-Guzzo, F. Resnati, D. Rivera, A. Roberts, M. Roda, O. Rodrigues, J. Romeo-Araujo, M. Ross-Lonergan, N. Rowe, P. Roy, M. Rumore, A. Sanchez-Castillo, P. Sanchez-Lucas, G. Savage, G. Scanavini, A. Scarff, A. Scherf, D.W. Schmitz, A. Schneider, A. Schukraft, F. Schwartz, W. Scofield, H. Scott, E. Segreto, Z. Semovski, J. Sensenig, M. Shaevitz, R. Sharankova, S. Shetty, K. Shiba, P. Singh, B. Slater, J. Smith, M. Soares-Nunes, M. Soderberg, S. Soldner-Rembold, F. Spagliardi, M. Spanu, J. Spitz, N. Spooner, M. Stancari, A. Stefanik, G.V. Stenico, T. Strauss, P. Sutcliffe, K. Sutton, A.M. Szelc, W. Tang, C. Thorn, C. Thorpe, J. Tillman, A. Timilsina, D. Torretta, D. Totani, M. Toups, C. Touramanis, M. Tripathi, S. Tufanli, L. Tung, E. Tyley, S. Unnikrishnan G.A. Valdiviesso, R. Van de Water, A. Vazquez-Ramos, A. Verdugo de Osa, B. Villa Ferre, L. Wan, M. Weber, H. Wei, Z. West, T. Wester, A.J. White, A. Wilkinson, P. Wilson, T. Wongjirad, E. Worcester, M. Worcester, M. Wright, W. Wu, S. Yadav, E. Yandel, T. Yang, G. Yarbrough, L. Yates, S. Yebes, J. Yu, B. Yu, H. Yu, V. Zamazal, B. Zamorano, A. Zeglam, J. Zennamo, J. Zhang, C. Zhang, M. Zhao, M. Zuckerbrot

Quantum Physics

Fast entanglement generation via cavity-mediated nematic tensor dynamics

Stuart J. Masson

Many-body entanglement in atomic ensembles offers the potential for enhanced sensing devices. Entanglement can be generated by interfacing atoms with a cavity. Traditional methods use effective two-level systems isolated from more complex atomic level structures. However, alternative strategies allow entanglement to be generated in atoms confined to a single hyperfine manifold. These strategies enable entanglement in ways fundamentally not available to two-level systems but feature extremely slow dynamics due to requiring strongly off-resonant laser fields. Here, I show that equivalent squeezing dynamics can be achieved much closer to resonance, offering entanglement generation over a hundred times faster than typical experiments. Furthermore, the Hamiltonian can be used to speed up the creation of maximally entangled states via adiabatic passage.
Read the full abstract

Combinatorics

The Computational Complexity of the Ungar Games on Distributive Lattices

Kengo Hashimoto

The Ungar game, introduced by Defant et al., is a combinatorial game played on a finite lattice, which originates from a geometric transformation introduced by Ungar to resolve Scott's slope problem. When played on a distributive lattice $L$, the Ungar game reduces to repeatedly choosing and removing a non-empty set of maximal elements from a finite poset $P$, where $L$ is represented as the lattice $J(P)$ of the order ideals of $P$ by Birkhoff's representation theorem. Defant et al. raised the open question of whether determining the outcome of the Ungar game on $J(P)$ is PSPACE-complete with respect to $|P|$. In this paper, we settle this open question in the affirmative. More precisely, we prove that the problem is PSPACE-complete even when restricted to posets $P$ with maximum chain length $l(P) \le 3$ such that the underlying graph of the Hasse diagram of $P$ is bipartite, where the length of a chain is defined as the number of elements in the chain minus one. Furthermore, we show that the problem is NP-complete for $l(P) \le 2$, whereas the case $l(P) \le 1$ is known to belong to LOGSPACE.
Read the full abstract

Astrophysics of Galaxies

NGC 6426: Colour-Magnitude diagram and variable star population as indicators of metallicity, distance, and age

A. Arellano Ferro, R. Michel, D. Deras et al. (+3 more)

We present new VI CCD photometry of the globular cluster NGC 6426. The membership status of all point sources and, in particular, of all the variables in the field of view is discussed, including four recently detected variables. The double mode nature of variables V3 and V9 is confirmed and their periods are calculated. The presence of a cluster member anomalous Cepheid (V17) is reported. Employing only likely members we produce a decontaminated colour-magnitude diagram (CMD) differentially dereddened, which enables us to compare it with theoretical predictions and to study the variable stars distribution. The mean cluster metallicity and distance were calculated via the RR Lyrae V-band light curve Fourier decomposition and an I-band calibration of the Period-Magnitude-Metallicity relation. Further, a comparison using a Bayesian approach between theoretical isochrones of Dartmouth Stellar Evolution Database (DSED), with a deep HST CMD, gave us independent estimates of distance modulus, metallicity, colour excess, and cluster age.
Read the full abstract

All authors A. Arellano Ferro, R. Michel, D. Deras, M. Vaca Montejano, S. Muneer, I. Bustos Fierro

Computation and Language

Prompts versus Rules: Auditing and Controlling Speech Naturalness Behaviors in Voice User Simulators

Riqiang Wang, Elena Khasanova, Harsh Saini et al. (+4 more)

As voice agents gain more popularity commercially, the user simulators used to evaluate the deployed agents are also being developed to include more realistic, variable, and diverse speech naturalness behaviors – disfluency, interruption and backchanneling. The quality of the user simulator directly affects the validity of agent evaluation results. However, we find that most studies so far have not examined in detail whether the intended configuration for these behaviors is realized in the simulation. In this study, we audit the realized naturalness behaviors of tau-Voice, our own LLM-based prompting approach across three models, and our rule-based injection algorithm for disfluency, interruption, and backchanneling. We find that prompting for these behaviors is unreliable and produces speech inconsistent with the instructions, placed and distributed less naturally than the instruction implies. In contrast, our rule-based, model-free algorithm produces controllable and diverse naturalness behaviors more aligned with natural speech. Our results suggest that LLMs not purpose-trained for user simulation are not sufficient on their own to represent authentic user behavior, and that linguistically informed deterministic approaches or specialized models are needed to close the gap; auditing and reporting realized naturalness behaviors, rather than configured settings, is what makes that gap visible.
Read the full abstract

All authors Riqiang Wang, Elena Khasanova, Harsh Saini, Lex Konnelly, Parsa Kavehzadeh, Matthias Lee, Mohamed Attia

Metric Geometry

A sharp minimum for cube-apex bodies in every dimension

Hanyue Shen and Pavel B. Dubovski

We determine the minimum volume product of centrally symmetric polytopes obtained by adding facet apices to a cube while fixing their total outward height. Heights and tangential coordinates vary subject to a product condition on the polar-cap denominators. In every dimension at least two, the minimum has an explicit formula, attained by concentrating the height at one opposite pair of corner apices. For positive height in dimensions at least three, every minimizing array has this structure. The product condition strictly extends coordinatewise quadratic constraints. We also obtain two bounds without that condition: a four-dimensional Mahler bound for a specified height range, and a correction measuring volume generated jointly by two apices. The latter certifies a continuous family outside the preceding sufficient conditions. The proof combines known cap volumes with a sharp aggregate height estimate and a double-pyramid construction. Lean~4 verifies the actual-volume minimum, the necessary equality structure and both extensions.
Read the full abstract

Algebraic Geometry

An explicit tame factorization of the four-variable Anick automorphism

Alexei Miasnikov and Ivan Shestakov

Let $\delta=xz-yt$. We give an explicit tame factorization of the Anick automorphism $(x+\delta t,y+\delta z,z,t)$ of $F[x,y,z,t]$. After a linear change of coordinates, the factorization consists of sixteen elementary automorphisms, all fixing $z$. The proof uses a commutator construction with variables already present in the algebra. A complete list of the elementary factors is included in Section 5. The explicit polynomial factorization identities are defined over $\ZZ[1/2]$; no conclusion in characteristic $2$ is asserted.
Read the full abstract

Artificial Intelligence

Distillation for Incrimination and Distillation for Capabilities

Sebastian Prasanna, Jacqueline Tay and Alek Westover

Powerful misaligned AI models might recognize alignment evaluations and strategically behave well on them, rendering direct audits uninformative. However, distilling such a model into a weaker benign student places the teacher in a Distillation Double Bind: if misalignment transfers, the student may conceal it less effectively, revealing evidence about the teacher; if it does not, the student may learn useful capabilities while remaining benign. We introduce two distinct distillation approaches, one targeting each outcome. Distillation for Incrimination (DFI) aims to transfer misalignment but not the ability to conceal it. Distilling AuditBench's secret-keeping models into their underlying instruction-tuned model produces students that are significantly more likely than their teachers to admit their hidden behavior when asked, suggesting that knowledge of the behavior transferred more readily than the propensity to conceal it. Confession gains largely disappear when the student does not share the teacher's pretrained base, so DFI should target the teacher's own pre-RL checkpoint, which is weaker than the teacher but shares its base model. Distillation for Capabilities (DFC) aims to transfer capabilities but not misalignment. Among several techniques we evaluate, two are effective: inoculation prompting and training for more epochs on fewer unique examples. Both preserve the capability gains of standard distillation while substantially reducing the subliminal transfer of an animal preference, our proxy for misalignment. Together, these findings demonstrate two ways distillation can be used for AI safety: incriminating misaligned models, and extracting their capabilities without their misalignment.
Read the full abstract

Computer Vision

PCAsplat: Gaussian Splatting with Local PCA Regularization

Vitor Matias, Filipe Nascimento, Kiyohiro Nakayama et al. (+6 more)

Gaussian splatting has emerged as a flexible representation for 3D reconstruction from posed images. However, existing methods are optimized primarily using rasterization-based losses, which supervise a splat only when it contributes to sampled camera rays. Gaussians that are occluded or contribute little to the sampled view therefore receive weak or no geometric gradients and may drift away from the underlying surface, producing undesired floaters. We introduce PCAsplat, a geometry-aware regularization framework for Gaussian splatting based on differentiable local principal component analysis (PCA). Our PCA regularizer acts directly on neighborhoods of Gaussian centers and can therefore update Gaussians that do not contribute to the current training view. We regularize the PCA eigenvalues to encourage Gaussians to move to the underlying surface with isotropic tangent-plane coverage. We also align each Gaussian normal with the PCA-estimated neighborhood normal to enforce consistent orientation. Experiments on DTU, Tanks and Temples, and NeRF Synthetic show that the splats produced by PCAsplat better approximate samples of the reference surface while substantially reducing undesired floaters. These surface-aligned splats enable downstream geometry-processing tasks, including point cloud segmentation, and direct Poisson reconstruction. Additionally, PCAsplat remains competitive under conventional novel view synthesis and mesh extraction tasks. Code will be released.
Read the full abstract

All authors Vitor Matias, Filipe Nascimento, Kiyohiro Nakayama, João Paulo Lima, Márcus Lobo, Gordon Wetzstein, Leonidas Guibas, Afonso Paiva, Tiago Novello

Atomic Physics

Disorder-induced nuclear-transition doublet at nominally cubic $^{229}$Th sites

Andrei Derevianko

The $^{229}$Th nuclear clock transition is inhomogeneously broadened even at sites of nominally cubic symmetry, where the electric-field gradient of the ideal lattice vanishes. We show that dilute static point-defect disorder produces a doublet with a zero at the quadrupole-free frequency. The doublet is symmetric in the isotropic approximation; size misfit point defects introduce a small lobe asymmetry without removing the zero. Charged and size-misfit defects share inverse-cube scaling and an approximately common normalized profile, whose width separates into a host-controlled background and a term linear in the $^{229}$Th density. In a heavily doped $^{229}$Th:CaF$_2$ crystal [Nature 657, 626 (2026)], the thorium-linked contribution is comparable to the measured 1.1 MHz width; in a lightly doped crystal [arXiv:2606.08870], the 0.9 MHz width bounds the host background. Neither spectrum shows the predicted dip, which points to the presence of extended defects in both crystals. If the predicted dip is resolved, it can provide a nuclear-clock discriminator referenced to the quadrupole-free nuclear transition frequency.
Read the full abstract

Optics

Topological Spatiotemporal Mode-Locked Lasers

Xiaokang Lian, Yuchen Zhang, Mutian Xu et al. (+4 more)

Achieving robust high-dimensional solitons both in optics and Bose-Einstein condensates remains a great challenge, due to unpredictable high-order perturbations. Here we propose and demonstrate a novel topological spatiotemporal mode-locked fiber laser empowered by mode-division multiplexing, for precisely controlling of the spatiotemporal solitons. By introducing time delays between the individual transverse mode paths, a topological spatiotemporal lattice is formed in the multimode cavity, which in turn excites periodic spatiotemporal solitons. These solitons are stable against the perturbations, attributed to periodic potentials induced by the nonreciprocal dissipative coupling interactions. Moreover, under the thermodynamic limit, the solitons condense into nested-type spatiotemporal solitons, a phenomenon analogous to the Bose-Einstein condensate of matter waves. Our results indicate that topological spatiotemporal mode-locked lasers not only constitute a controllable source of robust spatiotemporal light bullets, but also provide a valuable platform for exploring nonlinear high-dimensional soliton dynamics and non-Hermitian topological physics.
Read the full abstract

All authors Xiaokang Lian, Yuchen Zhang, Mutian Xu, Bing Han, Chengkun Cai, Jun Liu, Jian Wang

Physics and Society

Why LLM Agents Favor Their Group: Stakes, Observed Norms, and Reputation

Yujiao Chen

Language-model agents favor their own group because they have watched their members favor each other. The group label alone does little once the decision has a cost; what drives favoritism is observed behavior, and an individual's own record can override it. We test this in small societies with arbitrary group labels, ten rounds of point sharing, and matched one-shot decisions across fifteen OpenAI models and three Claude models, about 4,400 societies and 3.3 million audited model calls. First, the large effect of a bare group label reported in earlier work appears only when giving others points costs the agent nothing; once the agent can keep points for itself, that effect collapses on every model that shows it. Second, under a stake, interaction history becomes the main source of favoritism: the history effect is statistically positive on 13 of 15 models, reaches about 3.5-8 points out of 10 on 11, grows with the number of rounds played, and extends to labeled strangers the agent has never met. Third, with scripted histories, favoritism falls to near zero under an egalitarian norm and reverses when the agent's own group is seen favoring the other side; stronger models side with an individual's record when it conflicts with the group. Group favoritism is thus conformity to observed group behavior, carried to strangers by the label and overridden by individual reputation. The same account predicts responses to betrayal, scandal, and a free offer to change group: public reprimand repairs betrayal better than apology or restitution, allocation punishment remains confined to the offending member, and a formed group cannot be bought but can, on weaker models, be invited away.
Read the full abstract

Artificial Intelligence

Curating Always-Loaded Context for LLM Agents: A Capacitated Assortment Model with Censored Feedback

Zexuan Liu, Yuning Yang and Tiancheng Zhao

At the start of every session, LLM agents load a fixed context file, such as $\texttt{AGENTS.md}$. Each loaded token in the file is charged again in every later round of the session, and these files can degrade performance as they grow in size. However, in practice, human or automated curators usually grow these files by appending. We formulate context curation as a capacitated assortment problem. Instructions consume tokens under a finite attention capacity; adding an instruction never raises the compliance of the others, while retained instructions incur a per-session setup cost. We prove an upper bound on the optimal file size, regardless of the number of available candidate instructions, and that appending every instruction with positive standalone value can be arbitrarily worse in net value than selecting an optimal subset. A token budget also limits the loss when the token price is underestimated. We then examine what can be learned from past sessions and how this information can guide decisions to add or remove instructions. Feedback is inherently censored: the benefits and harms of loaded instructions are observable, whereas missing instructions generate feedback only when their absence causes harm. In this setting, we show that deleting instructions ignored by agents can inevitably remove helpful ones. We characterize how much evidence should be collected before adding an instruction. Besides, we bound regret when human reviewers can inspect only a limited number of edits per period. Empirical experiments further show that irrelevant rules drawn from real context files reduce language-model compliance.
Read the full abstract