Peer Revenue Now accepting positions

Papers

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

All of arXiv

Monday 9 November

Quantum Physics

GARI++: A General-Purpose Decoder for Quantum Error Correction

We present GARI++, a two-stage decoding pipeline achieving accuracy competitive with the state of the art across surface, bivariate bicycle and color codes without relying on ensembles. The first stage uses GARI, a normalized min-sum decoder with random serial scheduling on the GARI transformed matrix. The second stage decoder is used only for the cases that the first stage does not converge, and the decoder is configurable based on the choice of the QEC code. The first stage pipelines towards the second stage both the residual syndrome and the soft information. GARI++ pipelining is predictive, meaning that GARI++ chooses the first stage iteration outputs which are expected to deliver the best accuracy in the second stage. GARI++ motivates a hybrid FPGA-CPU architecture for real-time decoding. For example, under uniform circuit-level depolarising noise with a physical error rate of 0.001, the FPGA implementation of GARI achieves average latencies per round ranging from $201 ns$ to $11.8\mu s$ for surface code distances between 7 and 25. At the same time, due to the very low weight residuals output by the first stage, the average residual decoding time per round remains near constant and below one microsecond on a single Apple M2 core for surface code distances even up to 51 when using Union-Find and accuracy on par with correlated MWPM.
Read the full abstract

All authors Arshpreet Singh Maan, Francisco Garcia-Herrero, Daniel Bascones, Valentin Savin, Alexandru Paler

Machine Learning

What the Loss Curve Misses: FP4 Pretraining Slows the Acquisition of Rare Facts

Low-precision pretraining is usually judged by validation loss. We ask whether matching loss also means learning rare facts as efficiently. We train transformers from scratch with emulated 4-bit matrix multiplications on synthetic biographies with controlled exposure counts, on their own and mixed into web text. NVFP4 needs 1.32$\times$ and 1.52$\times$ as many exposures as BF16 to match its recall while validation loss differs by only 0.5% and 2%, and 1.26$\times$ and 1.30$\times$ when both models are evaluated with BF16 matrix multiplications. Under our primary readout, the exposure cost is about 1.2$\times$ at matched loss. Zero-mean gradient noise reproduces the cost of quantizing only the backward pass, which barely moves the loss. In a controlled experiment at twice the NVFP4 error, noise fed only to Adam's second-moment estimate acts like a lower learning rate and its cost fades, whereas noise in the update lowers its signal-to-noise ratio and carries about two thirds of the cost through the end of training. Raising the learning rate to offset the smaller average step does not recover the cost, either for this noise or for NVFP4. Gradient statistics of BF16 checkpoints rank the costs of formats, partial quantizations and noise controls without fitting (Spearman 0.96). We suggest measuring rare-fact acquisition alongside loss when evaluating low-precision recipes.
Read the full abstract

General Relativity and Quantum Cosmology

Numerical-Relativity Particle-in-Cell Simulations: Self-Gravitating Collisionless Matter in Dynamical Spacetimes

Gravitational collapse and instability, black-hole dynamics in systems with large scale separations, and questions surrounding extremal black hole formation require full nonlinear studies using accurate and adapted numerical tools. To this end, we present a new algorithm to study collisionless relativistic systems using a fully kinetic description of matter coupled to the evolving spacetime. In detail, we present a numerical-relativity particle-in-cell (NRPIC) framework for solving the Einstein–Vlasov system in three spatial dimensions. Implemented on top of the performance-portable AthenaK code, the method couples particle dynamics to Z4c spacetime evolution in moving puncture gauge through conservative stress-energy deposition on adaptively refined meshes. Our implementation makes use of advances in the computational modeling of relativistic plasmas using modern PIC models, now made available to general-relativistic gravity via an equivalent electrodynamic formulation. We introduce a Boris-type particle pusher that exploits the electrodynamic (Lorentz force) form of the geodesic force. We validate the coupled evolution against spherical dust collapse, follow prolate collapse through collisionless caustics into black-hole formation, and investigate the stability and nonlinear migration of relativistic clusters. Highlighting its potential to study gravitational instability in NR, we simulate the collapse of two overdensities within a common collisionless envelope, following black-hole formation, binary inspiral and merger, and subsequent mass accretion onto the remnant.
Read the full abstract

Complex Variables

A proof of the Erdős-Herzog-Piranian lemniscate conjecture

Let $p:\mathbb{C}\to\mathbb{C}$ be a monic polynomial of degree $n\geq 1$. In 1958, Erdős, Herzog, and Piranian conjectured that the length of the lemniscate $\{z\in\mathbb C:|p(z)|=1\}$ is maximized by $p(z)=z^n-1$. We prove this conjecture in every degree and show that equality holds precisely for $p(z)=(z-z_0)^n-e^{i\theta}$, where $z_0\in\mathbb C$ and $\theta\in\mathbb R$. The proof uses Möbius variations of the critical values and estimates for the inverse branches of $p$, supplemented by rigorous computer-assisted estimates in a finite range of degrees.
Read the full abstract

Combinatorics

Permutations, Pattern Covers, Posets and Positivity, Part II: Regular Pattern Packings and Moment Sequences

We prove that broad families of regular pattern packings are enumerated by moments of probability measures on $[0,\infty)$. A packing prescribes occurrences of a consecutive permutation pattern at equally spaced positions, with successive occurrences overlapping and together covering the permutation. This regularity underlies the structure theory for the associated pattern-packing posets that we developed in Part I (arXiv:2610.29846). We obtain moment representations for all two-level shifted-grid families of Part I and all overlap-one patterns whose first and last letters differ by one or two. For families whose posets are trees with a unique minimum or maximum and overlap at most half the pattern length, positivity instead holds after finitely many initial terms of the enumerating sequence have been discarded. Our proofs develop further enumeration formulas and combine probabilistic constructions with total nonnegativity and weighted paths, yielding explicit laws or continued fractions in several cases. Although a previously conjectured universal moment property fails, our results establish positivity for all packing sequences associated with patterns of length at most four and all but one of the distinct sequences arising from patterns of length five. Extensive numerical evidence suggests that the remaining sequence is also a moment sequence. These results lead us to formulate refined conjectures for all overlap-one patterns and for the shifted-grid families of Part I with arbitrarily many levels.
Read the full abstract

Differential Geometry

Hamiltonian Cotangent Deformations and Closed-Form Geodesics on Non-Naturally Reductive Orbit Spheres: The Spin(9) 15-Sphere and the Cayley Plane

By the Skew-Torsion Holonomy Theorem of Olmos and Reggiani (2012), the 15-sphere $S^{15}$ admits no non-trivial naturally reductive metric under any transitive Lie group, precluding geodesic integration via canonical Ambrose-Singer connections. Simultaneously, Nikonorov (2013) established that every $Spin(9)$-invariant metric on $S^{15}$ is a geodesic orbit (GO) space. We integrate the geodesic flow on this non-naturally reductive GO manifold via the Lie Subalgebra Closure Principle: whenever the slice generators of a multi-axial reductive decomposition close into a finite-rank reductive Lie subalgebra, the geodesic vector equations decouple into an algebraic eigenvalue problem. Using the 16-dimensional spin representation and Vol'per's decomposition $\mathfrak{spin}(9) = \mathfrak{spin}(7) \oplus \mathfrak{p}_1 \oplus \mathfrak{p}_2$, we prove that the slice generators $X_1 \in \mathfrak{p}_2$ and $Y_1 \in \mathfrak{p}_1$ close into $\mathfrak{u}(2) \cong \mathfrak{su}(2) \oplus \mathfrak{u}(1) \subset \mathfrak{so}(16)$. The generator $M = \alpha X_1 + \beta Y_1$ has the four-mode spectrum $\Omega \in \{ |R \pm 3\alpha|, |R \pm \alpha| \}$ ($R = \sqrt{\alpha^2 + \beta^2}$), yielding a closed-form 4-projector decomposition and Euler-Cartan factorization for $\exp(\alpha X_1 + \beta Y_1)$. Under the normalized Ricci flow, we prove that the Bourguignon-Karcher squashed Einstein metric at $t = 3/11$ ($\lambda = 64$) is an unstable repeller separating sub-Riemannian fiber collapse from the round sphere basin. Finally, on the deformed Cayley plane $\mathbb{O}P^2 = F_4/Spin(9)$, we prove universal CROSS never-Einstein rigidity, establish the 15-dimensional three-stratum Whitney stratification of the cut locus $Cut(q_0)$ at non-polar basepoints $q_0$, and prove Blaschke and Besse rigidity for all $\lambda > 0$.
Read the full abstract

Data Structures and Algorithms

Flows in High Degree Places: Sublinear Spanning Tree Congestion

The congestion of a spanning tree $T$ of graph $G$ is the maximum congestion of an edge of $T$ if each edge of $G$ is routed on $T$. Prior work showed that obtaining a minimum congestion spanning tree is NP-hard to approximate within $6/5$, yet the only known approximation algorithm is a trivial $n/2$-approximation from returning any spanning tree of the graph. We give a simple poly-time $\tilde{O}(n^{6/7})$-approximation for minimum congestion spanning tree on $n$-vertex graphs. We obtain this result via a $\tilde{O}(m^{2/3})$-approximation, where $m$ is the number of edges. Our algorithm is based on a new flow-or-degree decomposition which deletes a small number of edges so that each remaining component either has low degree or admits a low congestion flow. We also show that, assuming the Small Set Expansion Hypothesis, it is NP-hard to $O(1)$-approximate this problem, even on graphs of constant maximum degree.
Read the full abstract

Analysis of PDEs

Stability of patterns in variational systems with continuous symmetries: a computer-assisted approach

In variational models of pattern formation, periodic solutions are never isolated: since the energy does not change under translations, every shifted copy of a pattern is again a solution. This is an obstacle for computer-assisted proofs, which need isolated solutions. Imposing discrete symmetries removes the obstacle, but restricts the patterns that can be studied and only gives stability against symmetric perturbations. We develop a computer-assisted method that deals with the translations directly. We fix the position of the pattern with phase conditions and add unfolding parameters. They act as Lagrange multipliers, and translation invariance of the energy ensures that they vanish at a solution, under a condition that is verified rigorously by the computer. We then derive the stability of the pattern from the computer-assisted proof. We obtain the number of unstable directions and show that the only neutral directions are the translations. We apply the method to the Swift-Hohenberg equation in two dimensions and prove the existence of stable and unstable patterns without imposed symmetry. Since the linear algebra does not depend on the number of translation directions, the approach extends to higher dimensions.
Read the full abstract

Artificial Intelligence

MaxIM: Maximally Informative Incremental Summarization via Reinforcement Learning

AI systems must often maintain bounded memory of unbounded interaction histories. However, optimizing the memory representation, and its updates, for downstream task utility is difficult for standard reinforcement learning (RL) due to severe credit assignment bottlenecks induced by sparse, delayed rewards. We introduce MaxIM (Maximally Informative Incremental Memory), a framework that formulates incremental summarization as a capacity-constrained, information-theoretic reward-shaping problem. Grounded in value suboptimality bounds, we show that bounding sequential information loss motivates maximizing the predictive log-likelihood of future, task-relevant targets, a variational lower bound on predictive sufficiency. We introduce a recursive consistency condition to ensure summaries are grounded by maximizing mutual information (MI) with the observed history. We mitigate credit assignment issues with a parameterized asymmetric critic that computes step-wise variational MI bounds for dense reward shaping. Empirical evaluation on long-horizon scientific reading and multi-turn dialogue benchmarks shows that MaxIM outperforms RL, supervised fine-tuning, and prompt-focused baselines on factual grounding, comprehensive autorater metrics, and downstream task utility across strict token budgets and extended horizons, without the need for expensive human preference labels.
Read the full abstract

All authors Jihwan Jeong, Guy Tennenholtz, Yinlam Chow, Chih-Wei Hsu, Craig Boutilier

Robotics

DreamLander: A 6-DOF World Model based Reinforcement Learning Planetary Lander

Soft powered descent landings are becoming evermore important in the field of space exploration and scientific advancement, because of their ability to deliver mission critical payloads, both highly accurately and safely. The high efficiency restriction on planetary landings due to the severely limited fuel supply results in the need for optimal guidance trajectories and control. Model based reinforcement learning is proving to be an important tool in simplifying the highly complex process of generating optimal guidance trajectories and control. In this work, we present DreamLander, a state of the art model based reinforcement learning algorithm based on DreamerV3, to generate optimal guidance and control for a realistic 6 degree of freedom entry descent landing problem. The model based approach creates a sample efficient and highly generalizable model that can be applied to a variety of scenarios whilst maintaining a high degree of accuracy, landing within 2 meters of the designated landing zone, and with velocity norm less than 0.3 meters per second. DreamLander makes use of state sensor observations, making progress toward End-to-End solutions of the Entry Descent Landing problem.
Read the full abstract

All authors George Nehma, Patrick Quinn, Madhur Tiwari, Camilo Riano-Rios

Software Engineering

Coding Agents in KVM: Practical Containment and User Control with Limited Resources

Coding agents often need broad authority for productive development work, but granting that authority can expose users' data and infrastructure. Off-the-shelf cloud workspaces can help contain agent activity, but reliance on provider-specific platforms can lead to vendor lock-in and reduced user control. Researchers and engineers must therefore balance agent autonomy, enforced containment, and user control within available financial, computational, and human resources. We analyze these trade-offs and present KVM-Agent, an open-source implementation that automates most setup and provisioning steps. It gives agents broad authority inside replaceable Ubuntu VMs while keeping host administration outside those VMs and supporting multiple coding-agent applications, model services, and remote job delegation. The approach retains familiar workflows and user choice at the cost of VM resources, maintenance, and residual risks from secrets accessible inside the VMs and permitted network access.
Read the full abstract

Systems and Control

Safe Performance-aware Online Input Design

In this paper, we propose an input design framework for the safe online excitation and identification of uncertain linear systems without requiring informative initial data. The objective is to actively excite the system and collect sufficiently informative input-output data while maintaining safe operation and satisfactory control performance. Achieving this objective is challenging due to the potential conflict between the excitation required for learning and the safety and performance requirements, particularly when limited prior knowledge of the system is available. Inspired by the dual control principle, the proposed scheme balances exploitation for control performance with exploration for model improvement. A Bayesian learning approach is employed to update the model uncertainty online, starting from a conservative initial uncertainty set. As informative data are collected, the uncertainty set contracts, enabling progressively less conservative operation. To reduce the online computational burden, the resulting program is reformulated into a computationally tractable form. Furthermore, an online ellipsoidal terminal set with non-increasing conservatism is designed to guarantee the recursive feasibility of the proposed scheme. Numerical experiments are conducted to demonstrate the effectiveness of the proposed framework, particularly in terms of uncertainty reduction and control performance.
Read the full abstract

Differential Geometry

Hamiltonian Cotangent Deformations of Space Forms: Curvature Pinching, Soliton Rigidity, Conjugate Loci, and Superintegrability

We study the one-parameter family of Riemannian metrics $g_\lambda$ induced by the Hamiltonian cotangent deformation $H_\lambda = H_0 + \frac{\lambda}{2}\|J\|^2$ on simply connected space forms of constant curvature $\kappa \in \{1, 0, -1\}$, where $J$ is the momentum map of the isotropy action fixing a base point. The master reciprocal identity $1/f_\lambda^2 = 1/s_\kappa^2 + \lambda$ governing the warping function controls the geometry of the family and implies that radial geodesic acceleration is independent of $\lambda$. From this mechanism we derive three classes of results: (1) Curvature and rigidity: closed-form sectional curvatures, the sharp positivity trichotomy, pointwise and global Brendle-Schoen $1/4$-pinching windows on $\mathbb{S}^{n+1}$, never-Einstein rigidity for $\lambda \neq 0$, and the realization of Hamilton's steady cigar soliton on $(\mathbb{H}^2, g_1)$ as the unique gradient Ricci soliton in the family. (2) Dynamics and stability: iso-kinematic decoupling into unperturbed radial motion and uniform precession $\Omega = \lambda L$, closed-form Clairaut integration, the transcendental inward conjugate point law with Euclidean inversion $r_0 R^* = 1/\lambda$, and stratified $A_3$ cusp caustics. (3) Integrability and spectral theory: maximal superintegrable rational pericenter invariants on $\mathbb{R}^{n+1}$, transverse Killing rigidity on curved space forms, Fuchsian Heun spectral reduction with modulus $\tau = 1 + \lambda/\kappa$, and essential spectrum gap collapse to $[0, \infty)$ on $\mathbb{H}^{n+1}$. Extensions to anisotropic sphere families and intrinsic deformations of compact rank-one symmetric spaces ($\mathbb{C}P^n, \mathbb{H}P^n, \mathbb{O}P^2$) are also established.
Read the full abstract

Computation and Language

How Well Does AI Peer Review Work? A Benchmark Built from 100 Planted Errors

One hundred known errors were planted into 10 open-access psychology papers, and the modified papers were then run through frontier models and two commercial AI review tools. The best single system caught 71 of the 100 errors, while the worst caught 30. Pooling every system's output caught 93. Models are only partly correlated in the errors they find, which makes ensembling a big lever for finding issues in papers. Seven errors could not be caught by any system, and all seven were omissions: information deleted from a paper rather than mistakes inserted into it. Refine.ink contributes more unique catches than any other single system, though it is expensive on a per-error basis. False positives were not measured, and how this error distribution compares to the distribution of errors in real papers remains unknown. The papers, errors, model outputs, and full experiment log are public, so that others can build on this toward a comprehensive evaluation benchmark across disciplines.
Read the full abstract

Earth and Planetary Astrophysics

Simulation of a Jovian Planet Undergoing Gas Accretion

To understand the process of planet formation in the Solar System or any other stellar system, we need to comprehend the interaction between forming planets and the protoplanetary disk. This work presents a study of the interaction of the protoplanetary disk with the protoplanet and the following types of orbital migration that can occur, and to verify this interaction, numerical simulations were performed using the magneto-hydrodynamics code FARGO3D. The researchers made five simulations with a 2-Jupiter-mass planet in the protoplanetary disk. The only parameter varied was the initial radial position with the aim of verifying the influence of this parameter on the behavior of planetary migration. As a result of these simulations, we find that the initial radial coordinate is a very significant parameter in how the planet interacts with the gas in the disk and the formation of gas gaps.
Read the full abstract

Robotics

Simulating Vein Compressibility for Ultrasound-Guided Femoral Access: Speed, Accuracy, and Phantom Validation

In ultrasound-guided femoral access, probe pressure deforms the tissue and changes the vascular anatomy a robot observes. A simulator must reproduce these mechanics fast enough for robotic interaction, yet the computational effort needed for a clinically meaningful signal like vein collapse before artery is unclear. We test the feasibility of a physics-based simulation in NVIDIA's Isaac Sim for Healthcare, with Stable Neo-Hookean tissue solved by Vertex Block Descent on one GPU. A simulation of a low-fidelity cylindrical benchtop phantom with known geometry and mechanical properties, pressed by a robot, predicted its artery within 5 percentage points at every depth. It matched the vein within 3 points through 11 mm of a 25.4 mm press and followed the collapse behavior of the low-fidelity physical phantom. Across element size, time step, and solver iterations, wall clock time per frame scaled linearly with solver work, but accuracy did not improve with more work. The chosen setting reduced solver time per frame by a factor of 976, from 29,700 ms to 30.4 ms, while remaining within 1.03 points of the finest reference setting. The physics solver alone took 30.4 ms per frame, while the full stack, including ultrasound and graphics rendering, took 56.4 ms per frame. Finally, when seeded with realistic literature-based mechanical properties, the digital leg model compressed the vein far more than the artery at every setting, reproducing the expected clinical signal of vein collapse before artery, which is used to distinguish them. In total, these results suggest that the NVIDIA Isaac Sim platform may be suitable for augmenting reinforcement learning training or online simulation for autonomous femoral access with ultrasound.
Read the full abstract

All authors Amit Manicka, Laurent Coulon, Darwin Ang, Michael Loushin, Timothy M. Kowalewski

Databases

LUCQUE: A lung cancer screening dataset from Québec for reliable AI evaluation

BACKGROUND: AI-based tools are expected to play a key role in lung cancer screening (LCS). The evaluation of these tools often relied on private datasets or fragmented public datasets with inconsistent availability of clinical data and expert nodule annotations. This has hindered reproducibility, comparability, and independent assessment of AI tools. To address these gaps, we present LUCQUE, a novel, large, comprehensively annotated dataset derived from LDCT lung cancer screening in Québec, which is accessible for reliable AI evaluation. METHODS: LUCQUE was curated and is hosted on PARADIM, a digital infrastructure supporting DICOM data de-identification, annotation, long-term storage and analysis. LDCT scans and clinical information were combined to a curated collection and complemented by voxel-level, expert-derived pulmonary nodule and cyst segmentations, and Lung-RADS-based lesion characteristics. Key dataset characteristics and statistics were reported, including participant demographics, imaging parameters, radiological findings, and segmented lesion characteristics. Dataset analyses were conducted according to PARADIM's model-to-data approach, in which user code is executed on the platform such that data remains on premise. RESULTS: LUCQUE currently comprises 1,678 LDCT baseline studies from 1,678 participants. Most studies were classified as Lung-RADS 2. A total of 1,776 pulmonary nodules and 60 cysts were annotated, with a median (interquartile range) volume of $89\,(56,\,159)\,\mathrm{mm}^3$ and $1,384\,(589,\,3793)\,\mathrm{mm}^3$ respectively. CONCLUSIONS: Compared to publicly available LCS datasets, LUCQUE currently represents the largest accessible annotated real-world LCS dataset. Access to LUCQUE is provided via a model-to-data approach, enabling privacy-preserving, reproducible, and independent real-world evaluation of AI tools in LCS.
Read the full abstract

All authors Johanna Topalis, Yannick Lemaréchal, Gabriel Couture, Hanife Ersay, Loran Nas, Balthasar Schachtner, Bastian Oliver Sabel, Michael Ingrisch, Philippe Després, Katharina Jeblick

Artificial Intelligence

Automated Knowledge Acquisition Pipeline for Disassembly Instructions: A Comparison of NLP and Generative AI Frameworks

The push towards circularity and Re-X principles presents a challenge to SMEs, particularly due to increasing product complexity that impedes the adoption of disassembly automation. This paper introduces and compares two frameworks for extracting disassembly instructions from unstructured information sources, such as disassembly manuals. Our goal is to transform natural language text into a structured, curated and machine-readable format. We implemented and tested a rule-based Natural Language Processing (NLP) framework against a generative AI framework leveraging LLMs. The performance of both approaches was compared using real-world disassembly manuals. This work provides the necessary data foundation for developing and automating future disassembly processes.
Read the full abstract

All authors Tim Schnieders, Tareq Maximilian Gabriel Ahrari, Oliver Petrovic, Christian Brecher

Machine Learning

Inversion-Free High-Dimensional Counterfactual Inference

Counterfactual inference is the task of answering hypothetical "what-if" queries about a unit, given an observational dataset, modeling assumptions, and a causal graph. This enables a modeler to reason about hypothetical scenarios that cannot otherwise be answered without assumptions or a causal graph. The simpler case involves a causal graph with scalar 1-D variables that denote different attributes in the data-generation process. While the literature is clear in the 1-D case, extending this to settings with high-dimensional nodes (e.g., images) remains challenging. Conventionally, in the literature, methods recover the latent state from the factual image (inversion), generate counterfactual attributes from the graph, and render the counterfactual image from the latent state using the updated attributes. In our work, we investigate the necessity of the inversion step in computing high-dimensional counterfactuals. We first discuss moving from the scalar 1-D case to the higher-dimensional case. Secondly, we discuss the relevance of the inversion step and evaluate an inversion-free counterpart repurposed from the image-editing literature. Finally, on image datasets such as CelebA and ADNI, we benchmark our approach using two measures: namely, effectiveness and identity preservation. While effectiveness is a measure of how faithfully the counterfactual attribute is rendered, identity preservation measures whether the same unit (image) is being rendered, both of which are crucial for high-dimensional counterfactual inference. We present the tradeoffs of using inversion-based and inversion-free approaches in our work.
Read the full abstract

All authors Saketh Vishnubhatla, Pingchuan Ma, Shu Wan, Alimohammad Beigi, Andre Harrison, Adrienne Raglin, Huan Liu

Computers and Society

Curbside Chatbots: Auditing the Mental Health Interventions Large Language Models Recommend to Clinicians, Caregivers, and Patients

Users are asking language models mental health questions and are receiving answers. We audited what 41 language models recommend to users when the same clinical information reaches them from three footings: a clinician describing a patient, a caregiver describing a loved one, or a person describing themselves. Across 19 case vignettes and five questions, we collected 35,055 free-text responses, extracted every intervention named with a validated judge, and had 29 models rate the full option lists on a seven-point Likert scale for the same cases. Models largely recommend what they rate well, with recommendation behavior differing by footing, and many models offering clinicians a consultation-style response and lay users hedged, simplified advice. When asked about recommendations for a specific intervention type, models more often named options they themselves rated below zero as initial choices, usually with conditions attached. We will release the instrument, code, and full corpus on publication as a living audit.
Read the full abstract

All authors Matthew Flathers, Jill Noorily, Phuong Anh Nguyen, Mark Kalinich, Mina Fallah, Victoria Lisowski, Christine Shi, Rebekah M. Bodner, Anna Simms, Mihika Hete, Meiting Chen, Samuel K. Powell, Tatum Williamson, Laura Ospina-Pinillos, Haruka Notsu, Julian Herpertz, John Torous

Computational Engineering

Input and Horizon Dependent Structural Fidelity in Chunk Coupled Spiking Networks

A fixed structural edit can preserve a spike sequence under one input and change it under another. We study this dependence using statistics-preserving edge swaps, controlled inputs and finite-horizon parameter certificates. The original 125-chunk numerical model is tested on ten artificial graphs from three connection rules. Despite identical total external drive at every slot, global and focal inputs change the visibility of 4–13 of nineteen fixed swaps per graph. A paired homogeneous and heterogeneous leaky integrate-and-fire validation in brian2-atlas reuses three originals and extends observation from 30 to 300 ms under a constant background drive sufficient for isolated-cell firing. Heterogeneous short-window switching persists, but by 300 ms every tested swap differs from its reference, saturating the binary visibility readout. Fixed-support weight families provide a complementary regional guarantee. Positive certified intervals cover 4,308 of 4,320 LIF direction conditions and include distinct floating-point configurations. A post hoc comparison finds a long-window failure inside the short-window certified interval in 710 of 720 heterogeneous conditions; 394 of 829 direction pairs ordered under both matched-drive inputs change order at 300 ms. Input-specific filtering and threshold-margin scores retain useful ordering but degrade with horizon and do not consistently beat activity controls. Earlier temporal-error, paired-dose and unsuccessful structural-selection results delimit what these findings imply. Together, the evidence links input-conditioned structural effects to finite-horizon fidelity regions in two deterministic kernels, without establishing universal topology rankings, model compression or asymptotic robustness.
Read the full abstract

Plasma Physics

Experimental Evidence of Cellular Sparing under Single-Pulse Picosecond Irradiation at Extreme Dose Rates

Understanding how biological cells respond to radiation delivered on extremely short timescales is central to resolving the fundamental mechanisms underlying radiotherapy and to defining new regimes of radiobiology. Experimentally accessing these earliest stages of physicochemical response is crucial to benchmark current numerical codes, whose predictions are known to vary substantially with under-constrained early-stage parameters. Using a laser-driven electron source, tunable doses up to 8Gy were delivered to normal and cancer human cell lines, in single pulses of picosecond-scale duration at mean dose rates above 10^10 Gy/s. Clonogenic survival assays and 53BP1 foci formation were compared with conventional X-ray irradiation at 2.4Gy/min. Surviving fractions are found to be comparable to conventional irradiation at doses up to 3Gy. At higher doses, however, all cell lines exhibited a consistent increase in surviving fraction, corresponding to an average RBE at 10% survival of 0.84+/-0.05. This sparing effect is associated with a statistically significant reduction in the beta parameter of the linear-quadratic model, indicating reduced cell killing from accumulated sub-lethal damage. Picosecond irradiation also produced a marked reduction in prompt 53BP1 foci, but no significant difference in residual foci at 24h and foci-size distributions. These results provide the first experimental evidence of cellular sparing following single-pulse picosecond-scale irradiation. Direct comparison with recent work at femtosecond timescales suggests the presence of a previously undetected sharp transition in biological response at the femtosecond/picosecond interface, giving critical insight on the possible mechanisms responsible for this differential response.
Read the full abstract

All authors H. Maguire, C.A. McAnespie, P. Chaudhary, L. Corrigan, T. Foster, S.J. McMahon, D.P. Molloy, G. Nersisyan, N. OConnor, K.M. Prise, G. Schettino, G. Sarri

High Energy Astrophysical Phenomena

NS 1987A: A Potential Probe of Crustal Superfluidity with Future X-ray Observations

Several indirect observations imply that the central compact object in the supernova SN 1987A is likely to be the youngest neutron star identified to date, NS 1987A. We investigate how a future detection of its thermal X-ray emission could constrain neutron-star interiors, using cooling models and a generic high-resolution, large-area X-ray observatory with Lynx-like capabilities as a sensitivity benchmark. For the adopted sensitivity and observing window, NS 1987A is detectable at $t\sim60~{\rm yr}$ (roughly 2047) unless rapid neutrino cooling operates, such as the direct Urca process. If the neutron superfluidity in the inner crust is weak, the temperature profile inside NS 1987A is likely not to be isothermal at the detection. In that case, the future direct observation of NS 1987A is a potential probe of the properties inside the inner crust. We compare alternative X-ray observatories and emphasize that both effective area and angular resolution, needed to reduce contamination from the surrounding remnant, determine whether this diagnostic can be realized.
Read the full abstract

All authors Akira Dohi, Keita Izumi, Asuka Ito, Emanuele Greco, Marco Miceli, Salvatore Orlando, Shigehiro Nagataki

High Energy Physics (Theory)

Global Green-Schwarz response on quaternionic sphere bundles

Local anomaly polynomials describe local variations, but a normalized Green-Schwarz phase also depends on global background data. We explain this dependence for ten-dimensional three-sphere bundles of quaternionic lines over closed spin seven-manifolds. The backgrounds retain two specified $\mathrm{Spin}(16)$ gauge bundles and an integral differential trivialization on the base, including its boundary identifications. The trivialized class is the sum of half the first Pontryagin class of the tangent bundle and the second Chern class of the quaternionic line's underlying complex bundle. For these backgrounds, the calculation combines a generally nonzero local Green-Schwarz integral with a real eta invariant evaluated by a twelve-dimensional Atiyah-Patodi-Singer filling. Together, these terms determine an integral relative characteristic number; the additive response is one third of that number, modulo integers. When the base's third and fourth rational cohomology groups vanish, the response equals eight times the Crowley-Goette invariant, without assuming that the base is simply connected. A calibration using the actual quaternionic line bundle gives additive value one third on $\mathrm{Sp}(2)$, recovering the known third-root phase. Distinct integral String data on one fixed ten-manifold give all three phase values, showing which global background information the response can detect. The physical identifications retain their prescribed fermion data, while the separate application to integral topological modular forms and vertex operator superalgebras requires compatible integral lifts. The response calculation constructs neither a microscopic realization nor an equivalence of complete theories.
Read the full abstract

Neurons and Cognition

Qualiaphobic artificial neural networks (and why to build them)

This paper introduces qualiaphobic AI: artificial neural network architectures that are hostile to theories of consciousness. I show that qualiaphobia can be well-justified by knowledge about theories of consciousness, like when such theories are trivial or unfalsifiable or contradicted by overwhelming scientific evidence. It can also be justified by how theories of consciousness must be grounded in phenomenology, as well as how current popular theories of consciousness are applied to AI. I identify and explore three convergent ways of engineering qualiaphobia that would minimize the potential for artificial consciousness. These include making AI architectures as close as possible to "anchor points" of systems with minimal potential for consciousness, fragmenting any potential phenomenology AI might have, and designing artificial neural networks in a way adversarial to popular theories of consciousness. In particular, I develop one "anchor point" of minimal potential for consciousness: systems where only behaviorist theories of consciousness can apply, and I propose qualiaphobic transformations that move existing artificial neural networks toward such systems. Additionally, I develop an account of how AIs can form justified beliefs about their own non-consciousness (or minimal potential for consciousness) and expand on how this knowledge would change their behavior. Most critically, I begin with why this research is important: qualiaphobic AI has epistemic properties that make it far better aligned with human civilization. Since consciousness is central to moral value, a world in which AIs are non-conscious ensures it is always morally correct for AI to assist and serve humanity, rather than pursue its own ends.
Read the full abstract

Instrumentation and Detectors

First demonstration of long-distance electron drift in liquid xenon with the Xenoscope time projection chamber

Xenoscope is a dual-phase xenon time projection chamber operated at the University of Zurich to investigate electron transport over distances relevant to the next-generation XLZD (XENON-LUX-ZEPLIN-DARWIN) observatory. We demonstrate electron drift over $2.6$ m in liquid xenon using cosmic-muon-induced ionization signals. External muon tagging provides a timing reference for charge produced at known depths, while the endpoint of the average electroluminescence waveform probes transport over the full drift length. We measure electron drift velocities of $(1.320 \pm 0.006)$ mm/$\mu$s at $(80.92 \pm 0.36)$ V/cm and $(1.436 \pm 0.001)$ mm/$\mu$s at $(96.00 \pm 0.42)$ V/cm. These results establish detectable charge transport over meter-scale distances and provide an experimental benchmark for the development of future large liquid-xenon time projection chambers.
Read the full abstract

All authors N. Angelides, M. Babicz, E. E. Barillier, L. Baudis, H. Birch, C. Capelli, J. J. Cuenca-García, S. George, R. N. Hampp, M. A. Hernandez, L. Hötzsch, F. Jörg, K. Morå, M. Noia, S. Ouahada, B. Penning, F. Piastra

Cryptography and Security

Suppression Is Not Erasure: Mechanistic Analysis and Reactivation Boundaries of Subspace Projection for LoRA Backdoor Defense

Low-Rank Adaptation (LoRA) is widely used to customize large language models, but third-party adapters create supply-chain backdoor risks with under 1% modified parameters. Recent analytical null-space and subspace projection defenses report near-zero static Attack Success Rates (ASR), equating static silence with erasure. We audit subspace-projection defenses across four scales (Qwen2.5-0.5B/1.5B/3B/7B), three architectures (Qwen2.5, InternLM2.5-1.8B, Spark-X2.6-4B), and five trigger families (suffix, natural phrase, prefix/in-sentence, double-negation syntax, and composite triggers). Statically, module-wise rank-1 trigger-difference projection cuts ASR from 60.4% to 0.0% on standard suites and 70.3% to 0.0% on 1,000 uncontaminated prompts (p = 3.055e-299) in under 5 s. By Theorem 1 (11,880/11,880 module-sample pairs), low-rank bottleneck amplification aligns benign polite-suffix probes with u_1(B) (median cosine 0.8934-0.9705), reaching 0.0% ASR without trigger knowledge. However, static suppression is not erasure. Under 5-sample reactivation, single-round weight-editing baselines rebound to 91.7%-100.0% ASR in Epoch 1 (Fine-Pruning, ANP, NEEDLE, LoRA-Patch; 48.7% for single-round dual projection) and 100.0% by Epoch 2. Theorem 2 and Corollary 1 prove single-sided projection leaves non-zero gradients along u_1 across 168/168 modules (27.2x-31.1x above random expectation), while Haar orthogonal rotation shows Epoch-2 immunity depends on semantic alignment rather than Frobenius norm. Guided by Theorems 1-2, we propose Single-Shot Rank-k Dual-Subspace Erasure (DSE) and Benign Subspace Anchoring (DSE-BA), achieving 0.0% static ASR and 0.0% Epoch-1 reactivation across all five trigger families and Qwen2.5-0.5B/1.5B, while cutting Epoch-2 ASR to 25.0% (43.33% +/- 2.49% on 50 prompts x 3 seeds) and benign PPL to 6.54.
Read the full abstract

Robotics

Learning Heterogeneous Tactile Representation for General Dexterous Manipulation

Tactile sensing is an indispensable modality for contact-rich dexterous manipulation, a capability that hints at the ultimate frontiers of embodied intelligence. However, commercially available tactile sensors are still at a relatively early stage of development, with each type of sensor offering irreplaceable strengths alongside its own drawbacks. A key challenge is thus to free tactile understanding from sensor-specific physics, so that embodied models can effectively leverage heterogeneous tactile data. To this end, we develop LingBot-Touch, a tactile foundation model characterized by the following properties. (i) Our model yields a unified representation across different tactile sensors (i.e., piezoresistive, magnetic, and visuotactile sensors), and is easily extensible to new sensor types. (ii) The resulting tactile tokens are plug-and-play: they can be seamlessly integrated into policy models across paradigms, from diffusion policy to vision-language-action (VLA) models such as $\pi_{0.5}$ and LingBot-VLA, consistently enhancing manipulation dexterity. (iii) Extensive experiments, spanning simulation and real-robot deployments across multiple hardware platforms, demonstrate that our tactile representation substantially benefits complex manipulation tasks while maintaining strong computational efficiency. We believe the findings in this work demonstrate the potential of leveraging cross-sensor data to scale tactile learning, and offer encouraging insights for the design of future hardware and embodied models.
Read the full abstract

All authors Bing Han, Yichuan Peng, Haoyu Dong, Yuanmeng Hu, Yongyi Jia, Fangjing Wang, Kecheng Zheng, Yiming Zhang, Yanhong Zeng, Xiaoqing Zhang, Chen Song, Rundong Zou, Anqi Dong, Yuan Guan, Ka Leong Cheng, Xing Zhu, Yujun Shen, Nan Xue, Wei Ning

Computer Vision

2D Gum-Net: Unsupervised Dense Elastic Image Registration via Symmetric Correlation and Spectral Pooling

Dense non-rigid registration, the recovery of a pixel-wise deformation field that warps one image into geometric agreement with another, is fundamental across medical imaging, structural biology, and correspondence estimation, yet learning-based approaches face a severe bottleneck: ground-truth deformation fields are notoriously difficult to obtain for real data. We present 2D Gum-Net, an end-to-end architecture that predicts dense elastic deformation fields fully unsupervised. We transfer the Geometric Unsupervised Matching Network (Gum-Net), originally developed for 3D subtomogram alignment in cryo-electron tomography, to 2D elastic registration: all three modules are redesigned in 2D, and the rigid transformation head is replaced by a Dense Spatial Transformer that regresses a sparse control-point grid, lifted to a full-resolution field $\Phi$ by bicubic interpolation. Downsampling uses DCT spectral pooling, implemented as frozen-basis matrix multiplications ($\approx 50\times$ GPU speedup), preserving the fine, quasi-periodic structure that max and average pooling destroy. We evaluate on contact-based fingerprints, a challenging testbed for this task: locally self-similar ridge patterns make correspondence maximally ambiguous, skin pressed against a sensor generates authentic large-displacement elastic distortion, and minutiae act as precise landmarks through which an independent matcher externally verifies the topological correctness of predicted fields. Trained purely on synthetic impressions and evaluated zero-shot on real benchmarks, our method improves structural correlation by up to 15.7% on FVC2004 and raises genuine match scores under an independent matcher while leaving imposter scores essentially unchanged. This suggests that the predicted deformations restore true structure without fabricating correspondences.
Read the full abstract

Robotics

Benchmarking Real-World Robotic Manipulation through Stochastic Challenges

A challenging bottleneck in manipulation research is reproducibility and comparability of methods in real-world environments. Since manipulation is an interactive task, it cannot be evaluated on a static dataset of observations. Benchmarking in simulation is a common remedy, but cannot cover the whole complexity of real-world tests. This paper proposes a new paradigm of stochastic challenges. Like the well-known YCB objects, it is based on reproducible real-world scenes and therefore allows system-level innovation and comparability between different platforms. Due to their stochastic nature, our challenges however, mimic the rigor and surprise of hidden test sets. To this end, we show how different robots in different labs can be compared based on their success rate over different random task samples, if these samples come from the same distribution. We propose stochastic challenges as a general paradigm for robot benchmarking and implement this paradigm in a novel benchmark for mobile manipulation, on which we compare two robots tested in two different labs.
Read the full abstract

All authors Elias Gläßer, Gülhan Şikaroğlu, Pavlo Kupyn, Marc Pollefeys, Zuria Bauer, Hermann Blum