Human-Computer Interaction
Making Sense of Animal-to-Human Drug Development Evidence: Stakeholder Practices, Challenges, and Requirements for AI Tools
Rosni Vasu, Simona E. Doneva and Benjamin V. Ineichen
Animal models are widely used to study human biology and health interventions, yet translating findings from animal studies to humans remains challenging. Evidence across preclinical and clinical research informs experimental and translational decisions. Artificial Intelligence (AI) tools are increasingly reshaping how this evidence is searched, synthesized, and used, but it remains unclear how they should support the diverse stakeholders involved in assessing animal-to-human evidence. We conducted semi-structured interviews with 13 stakeholders to examine their evidence practices, challenges, and expectations for AI support. We found that stakeholders approach the same incomplete evidence base with different goals, expertise, and heuristics. Participants valued AI particularly for locating, screening, and extracting evidence, but were more cautious about automated interpretation and quality judgments. They emphasized transparency, source traceability, uncertainty communication, and human oversight. Based on these findings, we derive design implications for role-sensitive AI tools that support more systematic and transparent
reasoning about animal-to-human translation.
Read the full abstract ▾
Computation and Language
Stress-Testing LLM Lie Detectors: Role-Play Failures and Spurious Correlations
Maximilian von Klinski, Sebastian Lapuschkin, Wojciech Samek et al. (+1 more)
Lie detection probes aim to predict from a
language model's internal states whether its output is truthful or dishonest. However, role-play complicates what "truth" means for an LLM:
language models can adopt a wide range of personas that take very different claims to be true, including personas whose beliefs clearly contradict reality, such as a conspiracy theorist. In this work, we investigate whether lie detection probes reliably flag falsehoods generated under such an anti-factual persona or whether they instead follow the persona's beliefs. We introduce a
dataset of 8,916 human-reviewed, on-policy responses from three LLMs adopting anti-factual personas. Evaluating eight probes from prior work, we find that many fail in this setting, particularly when correct and incorrect answers are evaluated under the same persona
prompt. To investigate why, we construct three novel confounder
datasets in which truth is anti-correlated with a potential confounding concept. Our experiments reveal that many existing probes strongly track concepts that are spuriously correlated with truth in their training data, such as instruction compliance or response likelihood. Based on these findings, we introduce a simple linear probe that achieves the strongest overall performance on both the persona and confounder stress tests. Our results suggest that current lie detection probes are far from reliable and highlight the need for training data in which truth is decorrelated from confounding concepts.
Read the full abstract ▾
All authors Maximilian von Klinski, Sebastian Lapuschkin, Wojciech Samek, Lennart Bürger
Strongly Correlated Electrons
$\mu$SR measurements of 3D Ising ferromagnetism in bulk Fe$_3$GeTe$_2$ crystals
M. N. Wilson, J. Huang, T. J. Hicken et al. (+4 more)
Fe$_3$GeTe$_2$ is a quasi-two-dimensional itinerant ferromagnet with a van-der-Waals layered structure that exhibits substantial
magnetic anisotropy. In bulk samples, this material becomes magnetically ordered below 220~K, while in thin samples, the gate-tunability of the
ferromagnetic moment has attracted significant interest. Alongside high-temperature ferromagnetism, a low-temperature
magnetic state occurs whose nature has not been conclusively identified, with some studies suggesting a transition between ferromagnetism and antiferromagnetism with temperature. Here we present magnetometry and $\mu$SR data that demonstrate that Fe$_3$GeTe$_2$ exhibits three-dimenional Ising-like ferromagnetism over the entire temperature range between $T_{C}=220$~K and 2~K. The feature at 170~K that has previously been proposed to be a transition to antiferromagnetism is instead likely to represent a domain-freezing transition.
Read the full abstract ▾
All authors M. N. Wilson, J. Huang, T. J. Hicken, D. A. Mayoh, B. M. Huddart, G. Balakrishnan, T. Lancaster
Numerical Analysis
Exact barycentric rational interpolation of analytic functions with an application to the Bessel functions
Brett R. Green
We derive a method of analytic function expansion using barycentric rational interpolation, in which an arbitrary analytic weight function defines the interpolation and nodes. Truncating these expansions yields successive approximations, and the derivation of the expansions gives the approximation error as a byproduct. We demonstrate the method with cosine and the Bessel functions of the first kind for several choices of weight functions, and we examine the accuracy and convergence of the approximate interpolations as a function of the weight and the number of nodes used in the interpolation.
Read the full abstract ▾
Cosmology and Nongalactic Astrophysics
An improved standardization framework for Type II-P supernovae using a hierarchical Bayesian approach
A. Holas, S. Taubenberger, C. Vogl et al. (+41 more)
We use Type II-P
supernovae (SNe II-P) observed as part of the Nearby
Supernova Factory (SNfactory) experiment as standardizable candles to develop an improved standardization framework. The data consist of accurately flux-calibrated, host-subtracted spectrophotometry of low-
redshift SNe II-P in the wavelength range $3200 < \lambda < 10000$ ${\unicode{xC5}}$. We develop an improved method to determine the expansion velocities of SNe II-P based on Gaussian Process regression. We extend the conventional standardized candle method (SCM) by an additional correction term containing the H$_\alpha$ absorption-to-emission ratio and a hierarchical
Bayesian framework. We find that the additional correction improves the quality of the standardization, particularly around $30$ to $40$ days after the explosion. We are able to reduce the scatter of the Hubble-flow SNe II-P from $0.28$ mag down to $0.18$ mag compared to previous work. As an example application of this methodology we measure $H_0$ using a pre-existing calibrator sample of SNe~II having Cepheid- or TRGB-based distance
moduli. We find a value of $H_0 = 70.6^{+4.5}_{-4.3}$ km s$^{-1}$ Mpc$^{-1}$, including only statistical uncertainty, corresponding to a precision of $6.2$%. We find variations of $1.4$ km s$^{-1}$ Mpc$^{-1}$ between hierarchical model variations — well within our quoted statistical uncertainty. Without the additional correction term, we obtain $H_0 = 73.0^{+6.9}_{-6.1}$ km s$^{-1}$ Mpc$^{-1}$ ($8.9$% precision) highlighting the improved precision of our new methodology. We identify and discuss inconsistencies between the measured parent populations of the calibrator and Hubble-flow standardization characteristics — present in this and literature samples — that are not explained by a simple selection on magnitude and which will need to be improved upon as SN II-P samples grow.
Read the full abstract ▾
All authors A. Holas, S. Taubenberger, C. Vogl, D. Rubin, M. Vincenzi, W. Hillebrandt, G. Aldering, P. Antilogus, C. Aragon, S. Bailey, C. Baltay, S. Benitez-Herrera, S. Bongard, K. Boone, C. Buton, G. Csörnyei, Y. Copin, S. Dixon, M. Fink, D. Fouchez, E. Gangler, R. Gupta, B. Hayden, E. E. O. Ishida, A. G. Kim, M. Klauser, M. Kowalski, M. Kromer, D. K. Küsters, J. Nordin, R. Pain, E. Pecontal, R. Pereira, S. Perlmutter, K. A. Ponder, D. Rabinowitz, M. Rigault, K. Runge, M. Sasdelli, C. Saunders, G. Smadja, N. Suzuki, C. Tao, R. C. Thomas
Distributed Computing
From Pilots to Production: Lessons in Cross-Institutional Federated Training and Artificial Intelligence for Science
Olivera Kotevska, Max Carlson, Yan Gao et al. (+9 more)
Many of the most valuable scientific
datasets cannot be centralized: they are proprietary, export-controlled, classified, or bound by data-sovereignty restrictions. This inverts the usual paradigm: the model must move to the data, making
federated artificial intelligence (AI) core infrastructure for open science. We synthesize lessons from U.S. Department of Energy national laboratories, industry deployments, and the open-source community into a structured account of moving
federated AI from pilot demonstrations to dependable, multi-site production. The lessons come from five concurrent efforts spanning leadership-class supercomputers and cloud infrastructure in regulated settings. We organize them around an adapted five-domain readiness frame and a stratified view of the stack beneath a trained model: data architecture,
privacy and security controls, trust and verification, governance and socio-technical factors, and operations, the layers that decide whether a pilot becomes dependable. The question shifts from “can we train it?” to “can we operate it, audit it, and change it safely?” We report systems lessons in
memory efficiency, reliability, and synchronization; set out what
privacy, security, and decentralized trust require in production, and what is not yet validated there; and identify cross-cutting open problems (asynchronous federation, leakage
auditing, verification standards, and harmonized data contracts) that we argue warrant a dedicated, international, open-science working group.
Read the full abstract ▾
All authors Olivera Kotevska, Max Carlson, Yan Gao, Francis Jeanson, Yijiang Li, Mohammad Naseri, Minseok Ryu, Sahil Tyagi, Jerry Watkins, Feiyi Wang, Ravi Madduri, Kibaek Kim
Quantum Physics
Double Localization for Quantum Gibbs Sampler Gaps: From an Abstract Framework to Finite-Group Models
Ryu Hayakawa, Angus Southwell, Caesnan M. G. Leditto et al. (+2 more)
We introduce double localization, a framework for proving spectral gaps of
quantum Gibbs samplers through two complementary operations:
geometric localization, which selects updates supported in small spatial regions, and interface localization, which focuses on a model-defined subspace of observables while remaining geometrically global. Our abstract gap
theorem combines a global bound on this subspace, called the interface, with local
quantum gap estimates under quantitative coupling and
geometric assembly conditions. The resulting explicit lower bound controls the gap of the full dynamics without requiring the interface to be invariant under the generator. This separation allows global estimates, including classical comparison, to be combined with local control of the remaining
quantum directions. We apply the framework to
Hamiltonians built from the vertex terms of Kitaev's finite-group
quantum double construction, without plaquette terms, on finite simple three-regular graphs of girth at least six. For the specified local Davies dynamics, we prove an unconditional spectral-gap lower bound by a positive constant independent of graph size for every fixed nontrivial finite group and each fixed inverse temperature $\beta$ with $0\le\beta J<\log(5/3)$, where $J>0$ is the coupling strength. For the smallest non-Abelian group $S_3$, double localization yields a positive lower bound uniform in both graph size and the entire interval $0\le\beta J\le\log 3$, providing a guarantee that does not follow directly from existing results. This demonstrates how model-specific finite calculations can extend spectral-gap guarantees for the full
quantum dynamics.
Read the full abstract ▾
All authors Ryu Hayakawa, Angus Southwell, Caesnan M. G. Leditto, Kuo-Chin Chen, Min-Hsiu Hsieh
Machine Learning
RATIO: Reasoning Analysis and Token-level Inference Optimization for Quantized Reasoning Models
Chengzhu Bao, Xianglong Yan, Tianao Zhang et al. (+3 more)
Post-training quantization (PTQ) has become a widely adopted technique for reducing the
memory footprint and inference cost of large
language models (LLMs). However, recent studies reveal that when applied to
reasoning models, PTQ not only degrades
reasoning performance but also exacerbates overthinking, leading to longer
reasoning trajectories. These issues may offset the efficiency gains expected from lower-precision inference. Existing approaches mainly rely on complex optimization procedures. More recent lightweight inference strategies instead use predefined overthinking markers, limiting their adaptability across quantized models. To address these issues, we propose
Reasoning Analysis and
Token-level Inference Optimization (RATIO), a framework that identifies model-specific overthinking
tokens and assigns each a tailored penalty. RATIO first introduces Quantization-aware
Reasoning Behavior Analysis (QRBA) to identify overthinking
tokens by analyzing discrepancies between full-precision and quantized models. It then adopts
Token-Specific Penalty Determination (TSPD), which leverages full-precision guidance to derive
token-specific penalties without additional training. Extensive experiments show that RATIO achieves a better accuracy-efficiency trade-off than existing
token-level interventions. Specifically, RATIO achieves up to 9.8 points accuracy improvement and reduces chain-of-thought (CoT) length by up to 51.3% compared with quantized baselines. The code will be available at https://github.com/steven-bao1/RATIO.
Read the full abstract ▾
All authors Chengzhu Bao, Xianglong Yan, Tianao Zhang, Jiaqi Chen, Shaoqiu Zhang, Yulun Zhang
Machine Learning
Finite-Horizon Fisher Memory in Two-Sided Power-Bounded Recurrent Systems
Jeonghoon Lee
We analyse allocation, admission and post-write retention in finite-horizon linear-Gaussian noisy recurrent
memories. At every horizon, the directional Fisher
memory $M_n$ satisfies $\operatorname{tr}M_n=N$: non-normality redistributes information but cannot raise its spherical average, while normal carriers satisfy $M_n=I$. For bi-power-bounded carriers, we derive uniform $1/n$ lag bounds, identify the limit of $M_n$ with the inverse of the classical Cesàro asymptotic limit of $W^\top$, and give finite-horizon error bounds. A time-varying coupling defines an end-to-end store operator. The writer-optimal direction need not be store-optimal. After writing ends, an invertible hold preserves the full stored Fisher
matrix. Additive contamination bounded by $\alpha$ times the closure covariance retains at least $1/(1+\alpha)$ of that
matrix; a covariance-aware decoder attains the corresponding accuracy. With recurrent carriers held fixed, training input masks and linear readouts approached the task-specific optimum in 160 runs, with median normalized Rayleigh efficiency above $0.998$. Binary accuracy matched the Gaussian prediction to mean absolute error below $0.002$ over more than four orders of magnitude in $J$. In a separate pre-specified study of 320 runs, trained masks followed the designated input-time objective in both carrier types, in 16 of 16 draws. These studies used development-seen carriers and are pre-specified validations, not blind holdouts. The same fixed design reproduced the objective-specific result in 16 of 16 draws on carriers unused before run commitment. Exact isolation preserved information, while a decoder fixed at its training horizon fell to chance; inverse-adjoint transport restored its sampled decisions to numerical precision.
Read the full abstract ▾
Distributed Computing
Darpan: A Digital Twin Framework for the Next-Generation Computing Continuum
Zhiyu Wang and Rajkumar Buyya
Computing-continuum applications distribute work across devices, edge systems, fog resources, and clouds. While a placement, scheduling, or recovery decision is being made, resource availability, network conditions, and application progress may change, so the decision can be invalid by the time it is executed. Existing runtimes enact predetermined decisions, whereas simulation tools compare alternatives in preconfigured environments; neither explores alternative outcomes directly from the current state of a running application. We propose a Digital Twin framework, called Darpan, that supports both physical and digital execution: the physical side runs real applications and continuously observes their runtime state, while the digital side maintains a continuously updated, executable virtual counterpart based on these physical observations. Darpan evaluates candidate decisions independently from the same starting point and returns the selected decision to the physical side, where its feasibility is validated against the latest physical state before actual execution. Across real Directed Acyclic Graph (DAG) workloads, Darpan predicts physical response time with a mean absolute error of 0.485 s and retains 93.1% scale-out efficiency at 40 nodes. In the state-change trials, Darpan takes about 2 ms on average from state capture to rejection of an invalidated decision, stopping the request before data transfer and avoiding unnecessary physical execution overhead. Under the same physical-DAG budget, Darpan-generated experience enables a weaker Proximal Policy Optimization (PPO) scheduler to outperform two state-of-the-art Deep Reinforcement Learning (DRL) schedulers by 29.1% and 20.2%.
Read the full abstract ▾
Machine Learning
Probabilistic Adversarial Training
Andi Zhang, Xingyu Zhao and Siddartha Khastgir
Building on a probabilistic perspective in which
adversarial examples arise from the overlap between a distance-based distribution $p_{\mathrm{dis}}$ and a victim-classifier-induced distribution $p_{\mathrm{vic}}$, we start from a simple intuition:
adversarial examples become harder to generate when these two distributions are pushed apart, as their overlap becomes smaller, thereby increasing robustness. This intuition naturally motivates a KL-based robustness objective. We then prove that $\mathrm{KL}(p_{\mathrm{dis}}\|p_{\mathrm{vic}})-\log Z_{\mathrm{vic}}$ is a lower bound on probabilistic robustness (PR), where $Z_{\mathrm{vic}}$ denotes the normalizing constant of $p_{\mathrm{vic}}$. Since PR is generally intractable to compute directly, maximizing this KL-based lower bound provides a tractable surrogate objective for improving PR. We further show that this objective recovers a scaled form of
adversarial training, offering a probabilistic interpretation of
adversarial training and a principled route to robustness improvement. We call the resulting method probabilistic
adversarial training. Experiments show that it consistently improves PR, and ablation studies demonstrate that the induced scaling factor can even enhance the PR of non-probabilistic
adversarial training methods.
Read the full abstract ▾
Robotics
Tool-Policy Co-Design for Powder Weighing in Laboratory Automation
Nikola Radulov, Xin Yang, Kevin S. Luck et al. (+1 more)
Autonomous powder weighing is one of many bottlenecks in laboratory automation due to the complex, non-linear dynamics of heterogeneous materials.
Robot chemists performing this task utilise standard tools shaped for the dexterity of human hands, whose fixed geometry sets the dynamics that the control policy needs to regulate. This work introduces a tool-policy co-design framework that concurrently optimises the morphology of a dispensing tool and its control policy for use by
robots in chemistry laboratories, formulated as a bi-level optimisation that minimises dispensing error over a target distribution of powder flowabilities. The outer loop varies tool-design parameters such as tool depth, width and rim spike
topology using
Bayesian optimisation and hyperband, while an inner loop optimises a control policy for each candidate morphology. We also introduce a
geometric similarity metric that warm-starts policy training from cached policies of structurally similar designs, exploring 28% more configurations under the same compute budget. The proposed framework is evaluated on a robotic powder weighing task across seven materials with distinct physical dynamics in a flowability-informed
robot-material simulation framework. Experimental results demonstrate that our co-designed tool morphology reduces real-world weighing errors by 45% relative to a standard tool, including on previously unseen materials. These results demonstrate our method can adapt both the control policy and the physical tool to the dynamics of the target material, bringing a new paradigm for material manipulation to the field of laboratory automation.
Read the full abstract ▾
All authors Nikola Radulov, Xin Yang, Kevin S. Luck, Gabriella Pizzuto
Cosmology and Nongalactic Astrophysics
Effects of a central dark matter core on time-delay cosmography with galaxy clusters
Yuting Liu and Masamune Oguri
The central dark matter distribution probes the nature of dark matter and
baryon physics, yet its impact on time-delay cosmography is not fully explored. We update previous analyses of the cluster-lensed
quasar SDSS J1004+4112 and the
supernova Refsdal in MACS J1149.5+2223 by replacing the main
halo density profile from an elliptical Navarro-Frenk-White (NFW) profile to a cored elliptical NFW profile and investigating the impact of the central dark matter core on the Hubble constant $H_0$. We jointly infer $H_0$ and the dimensionless core size parameter $\beta=r_{\rm c}/r_{\rm s}$, while the Brightest Cluster
Galaxy (BCG) is either tied to the scaling relation of cluster member
galaxies or modeled separately, with its mass left free or assigned a Gaussian prior from the
stellar mass estimate for a Chabrier or Salpeter initial mass function (IMF). For SDSS J1004+4112, constraints on both $\beta$ and $H_0$ strongly depend on the treatment of the BCG. A large core is implied when no prior is added to the BCG
stellar mass, while a reasonable assumption on the BCG mass prefers a nearly cuspy
halo with $\beta\sim 10^{-3}$. For MACS J1149.5+2223, a large core of $\beta\sim 0.5$ is obtained with no prior, while a modest core of $\beta=0.0410^{+0.0865}_{-0.0276}$ is preferred when the Salpeter IMF is assumed. The constraint on $H_0$ from MACS J1149.5+2223 is stable against the BCG treatment, supporting its robustness. For the Salpeter-prior case, we obtain $H_0=65.9^{+3.5}_{-3.4}\mathrm{~km~s^{-1}~Mpc^{-1}}$, and different BCG assumptions change the median $H_0$ by less than $1\sigma$. The core size constraints are translated into the self-interacting dark matter cross section, giving a 95% upper limit of $\sigma/m<0.035\,\mathrm{cm}^{2}\,\mathrm{g}^{-1}$ from SDSS J1004+4112 and the 68% range of $\sigma/m=0.071^{+0.15}_{-0.045}\,\mathrm{cm}^{2}\,\mathrm{g}^{-1}$ from MACS J1149.5+2223.
Read the full abstract ▾
Materials Science
A transferable full-band Monte Carlo framework for complex alloy avalanche photodiodes
Shafat Shahnewaz, Hannaneh Karimi, Joe C. Campbell et al. (+1 more)
We present a physics-based multiscale full-band Monte Carlo framework for modeling avalanche multiplication and excess noise in complex
alloy avalanche photodiodes (APDs), demonstrated on Al$_{0.7}$InAsSb as a representative quaternary system. The framework links atomistic material structure to device-level avalanche statistics: an environment-dependent $sp^3d^5s^\ast$ tight-binding calculation resolves the full conduction- and valence-band structure of the random- or digital-
alloy configuration – including $\Gamma$, X, and L valley ordering, non-parabolicity, anisotropy, and
spin-orbit-induced valence-band splitting – and supplies the band-structure inputs for
stochastic high-field transport. A central element of the framework is a physics-derived treatment of
alloy-disorder scattering, in which the quaternary disorder potential is constructed from atomic valence differences, covalent radii, and Thomas–
Fermi screening through a composition-weighted decomposition into binary contributions, complemented by composition-interpolated polar-optical, acoustic, intervalley-
phonon,
impurity, and impact-ionization models. Because every material-dependent input is generated from the atomic composition and configuration by the same well-defined procedure, the framework transfers without structural modification to arbitrary zinc-blende ternary and quaternary
alloys. Applied to a \SI{1}{\micro\meter} Al$_{0.7}$InAsSb p-i-n APD, the framework reproduces the measured gain and excess-noise characteristics with only the impact-ionization softness parameters calibrated. The approach provides a documented, reproducible route for analyzing and designing complex
alloy APDs in which band structure, disorder, and scattering physics jointly determine gain and ionization statistics.
Read the full abstract ▾
All authors Shafat Shahnewaz, Hannaneh Karimi, Joe C. Campbell, Avik W. Ghosh
Computer Vision
Inline Memory Meets Reusable Skills: Memory-centric Framework for Vision-Language-Action Model
Zaijing Li, Rui Shao, Bing Hu et al. (+3 more)
Vision-
Language-Action (VLA) models have shown strong promise for general-purpose robotic manipulation, yet adapting them to new tasks and domains remains inefficient: existing methods often rely on parameter tuning, incurring substantial costs and risking catastrophic forgetting of previously learned tasks. To address this, we propose Optimus-R, a
memory-centric VLA framework that formulates robotic adaptation as explicit query-skill
memory tuning. Optimus-R introduces: (i) An Inline
Memory Interface for skill extraction. It inserts learnable
memory tokens into the VLA prefix stream, allowing the backbone to derive control-aware query and skill representations within the native action-conditioning pathway. (ii) A Query-Skill
Memory Bank for skill learning. It externalizes skills into query prototypes for deciding what to retrieve and skill values for specifying how to act, supporting skill reuse and expansion with limited parameter updates. (iii) A lightweight Bridge-and-Adapt mechanism for skill updating. It aligns target-domain queries and skills with the existing
memory space through a lightweight adapter and residual
memory updates. Experiments on in-domain adaptation, cross-domain adaptation, and lifelong learning show that Optimus-R enables data-efficient skill learning while mitigating catastrophic forgetting.
Read the full abstract ▾
All authors Zaijing Li, Rui Shao, Bing Hu, Haoyu Zhang, Dongmei Jiang, Liqiang Nie
Probability
Spectral Universality for Matrices with Non-Linear Correlated Entries
Marwa Banna, Issa-Mbenard Dabo and Florence Merlevède
We establish nonasymptotic spectral comparison results for
matrices whose entries are non-linear functions of an iid random field. Under an exponential decay assumption on an $\mathbb L^\infty$-coupling coefficient, we derive high-probability bounds for the Hausdorff distance between their spectra and those of Gaussian
matrices with matching covariance structures. We further derive a comparison with a covariance-matched free model and derive corresponding noncommutative Khintchine-type bounds. The results apply, in particular, to
matrices whose entries are nonlinear transformations of causal linear processes, Volterra-type processes, and neural networks. The proof combines finite-
memory approximation, block decomposition, Gaussian interpolation, and resolvent estimates.
Read the full abstract ▾
Machine Learning
TopTimeNet: Topologically-assisted time-series classification model
Sharareh Sayyad and Sophia Bazzi
Distinguishing periodic from chaotic dynamics in a time series is a fundamental challenge in both physics and engineering. Yet, end-to-end learned architectures must discover both a representation and a decision boundary from data, at substantial cost. We introduce TopTimeNet, which decouples these tasks: a fixed, non-learned stage extracts a $42$-dimensional
geometric and topological descriptor from Takens delay
embeddings and persistent homology, and a lightweight learnable stage performs classification. On a
benchmark of $49$ nonlinear dynamical systems, a $1{,}638$-parameter configuration matches the mean accuracy of one with $33\times$ more trainable parameters. Additionally, this approach delivers mean accuracy comparable to convolutional neural networks and surpasses the average performance of converged
Transformer models, while requiring three to four orders of magnitude fewer trainable parameters. Robustness also depends sharply on where noise is introduced: TopTimeNet degrades gracefully under perturbations to its precomputed features, but degrades sharply when noise is introduced into the raw signal and the full feature-extraction pipeline is recomputed, showing that robustness to perturbations of the precomputed features does not imply robustness of the complete raw-signal-to-prediction pipeline. These results show that decoupling fixed
geometric and topological feature construction from a lightweight discriminative stage can achieve comparable classification accuracy with substantially fewer trainable parameters.
Read the full abstract ▾
Optics
Topological Fermi-polaron Polaritons with Electrically Invertible Topology
Xin Xie, Chenxi Liu, Yuze Liu et al. (+8 more)
Topological polaritons provide a route to engineer optical Chern bands, but realizing large gaps or efficient electrical tunability remains challenging because time-reversal-
symmetry breaking in
semiconductor polariton systems is typically limited by weak Zeeman splittings. Here, we demonstrate a topological Bose-
Fermi mixture —
Fermi-polaron polaritons —where large gaps and electrically invertible
topology emerge from the interplay of
Fermi-polaron many-body interactions, photonic
symmetry, and strong light–matter coupling. Using a gate-controlled monolayer transition-metal dichalcogenide integrated with a
symmetry-engineered photonic
crystal, we create
Fermi-polaron polaritons with topological band structures. The direct coupling between band
topology and gate-tunable many-body interactions provides a mechanism for amplified time-reversal-
symmetry breaking beyond bare Zeeman splitting and enables electrical inversion of topological gaps, Berry
curvature, and Chern numbers. These results establish an electrically reconfigurable platform for interaction-driven topological states and programmable topological photonics.
Read the full abstract ▾
All authors Xin Xie, Chenxi Liu, Yuze Liu, Lingxiao Zhou, Chulwon Lee, Yuhan Zhang, Nathanial Lydick, Kenji Watanabe, Takashi Taniguchi, Kai Sun, Hui Deng
Differential Geometry
Constant mean curvature disks with circular boundary in the $3$-sphere and in hyperbolic space
Rafael López
We prove that a compact immersed disk of constant mean
curvature in $\mathbb S^3$ or in $\mathbb H^3$ whose boundary is mapped diffeomorphically onto a circle is an embedded totally umbilical disk.
Machine Learning
Pseudo-Label-Triggered Retraining from Forecast Errors for Online Time Series Forecasting
Yeryeong Kwak, Yoo-Min Jung and Jonghun Park
Real-world time series
forecasting systems operate under non-stationary data streams, where
forecasting performance may degrade over time. Although retraining can recover the performance, it incurs non-trivial computational and operational costs. Under limited deployment resources, the key challenge is therefore not only how to retrain but also when to retrain. While existing retraining policies often rely on indirect indicators such as drift alarms or model staleness, we instead use realized forecast errors as direct deployment feedback. In this paper, we propose PILOT (Pseudo-label-Informed Learned Online Trigger), an online retraining framework that learns when to retrain from forecast-error dynamics. Since ground-truth retraining labels are unavailable, PILOT constructs a pseudo-label from future increases in forecast error and trains a lightweight scorer to predict it from observed error states. At deployment, PILOT uses only completed forecast errors and serves as a plug-in module for arbitrary
forecasting backbones without architectural modification. We evaluate PILOT under standard multivariate
forecasting settings across eight
benchmarks with three representative backbones—DLinear, iTransformer, and TimesNet. Across all three backbones, PILOT achieves state-of-the-art average-rank performance among retraining policies while maintaining a favorable performance–efficiency trade-off.
Read the full abstract ▾
Artificial Intelligence
Safety of Latent Communication in Multi-Agent Systems
Muhammad Huzaifa, Sina Mavali and Thorsten Eisenhofer
Latent communication enables multi-
agent systems to exchange information directly in internal representation space, reducing the
token, computation, and latency overhead of text-based communication. To this end, lightweight trainable links are introduced to map the sender's representations into the receiver's input space. In this work, we show that even benign link training can increase harmful compliance relative to text-based communication while the underlying safety-aligned
agents remain unchanged. An attacker can amplify this effect by optimizing the links on harmful query–response pairs or poisoning otherwise benign training data. We further develop a reinforcement-learning attack that rewards harmful compliance alongside benign task performance without requiring harmful target responses. Across three communication
topologies and four safety
benchmarks, this attack raises the mean harmful-compliance score from 27.9 with benignly trained links to 76.9. Compared with direct supervised optimization, it also achieves higher average accuracy on two benign utility
benchmarks. Adapting the rewards toward safer behavior also enables repair of compromised links, substantially reducing harmful compliance across all evaluated attacks without updating the
agents. Overall, our results show that safety
alignment requires considering the multi-
agent system as a whole. Code: https://github.com/Muhammad-Huzaifaa/latent-safety
Read the full abstract ▾
Cryptography and Security
Privacy Foundations for Multi-Institutional Scientific Artificial Intelligence
Olivera Kotevska, Sumit Jha, Aur'elien Bellet et al. (+5 more)
Scientific artificial intelligence (AI), spanning foundation models (FMs) to
federated data-analysis pipelines, is becoming shared infrastructure across national laboratories, universities, hospitals, and industrial partners. This collaboration creates
privacy risks whose natural unit is often an institution's participation, research strategy, or technical capability rather than a single record. Differential
privacy (DP),
federated learning (FL), secure computation, trusted execution, and provenance each protect parts of the stack, but their guarantees rarely compose across mixed-trust institutions, access tiers, and autonomous
agents. This perspective recasts
privacy for scientific AI as an assurance problem defined by six elements: protected asset, observer, channel, permitted disclosure, guarantee, and evidence. We demonstrate the framing through a claim register for a composite cross-institutional scenario and use it to assess the model lifecycle. Two of the resulting gaps are specific to leadership-class facilities: scheduler, allocation, and telemetry metadata expose an institution's resource posture, and instrument-attached control loops leak research strategy through timing and contention on shared accelerators. We identify six research priorities: institution-level guarantees,
agent-communication
privacy, cross-tier information flow,
privacy-compatible reproducibility, leadership-scale accounting, and instrument side channels. The contribution is a common form for stating, comparing, and
auditing claims whose guarantees otherwise remain fragmented across the scientific AI stack.
Read the full abstract ▾
All authors Olivera Kotevska, Sumit Jha, Aur'elien Bellet, Rui Hu, Nathaniel D. Bastian, Rafael Ferreira da Silva, Ravi Madduri, Kibaek Kim
Computation and Language
Explore-over-Graph: Hybrid Embedding-LLM Reasoning for Knowledge Graph Question Answering under Incompleteness
Ola El Khatib and Djellel Difallah
Large
language models (LLMs) are increasingly combined with knowledge graphs (KGs) to ground
reasoning in structured evidence. However, most LLM-based KGQA methods rely on traversing existing graph edges and become unreliable when
reasoning paths are broken by missing facts. Alternatives that ask LLMs to generate missing knowledge risk introducing hallucinated evidence. We introduce XoG (eXplore-over-Graph), a framework for multi-hop question answering over incomplete KGs that recovers missing
reasoning paths from learned graph structure rather than LLM parametric knowledge. XoG combines type-level entity-relation statistics to identify candidate relations with KG
embeddings to retrieve plausible missing entities, using the LLM as a semantic selector and reasoner. These mechanisms are integrated into an iterative planning-exploration-
reasoning process. Experiments on WebQSP, CWQ, and the Wikidata-based BRINK
benchmark show that XoG remains competitive on complete KGs and consistently outperforms comparable methods without task-specific KGQA training under KG incompleteness. These gains persist across multiple LLM backbones, indicating that stronger LLMs alone do not resolve missing graph evidence. XoG also reduces LLM
token consumption by up to 33% compared with a closely related planning-based approach.
Read the full abstract ▾
Computer Vision
Seeing as Humans Do: Learning from Motion to Segment Anything Without Supervision
Weijian Jian, Xiaoyue Zhang, Bin Xiao et al. (+5 more)
The Segment Anything Model (SAM) relies heavily on massive manual annotations, creating a fundamental bottleneck for model scaling. While unsupervised methods attempt to learn object concepts from motion, they typically overfit to moving entities, lacking both multi-granularity understanding and the ability to generalize to static objects. To overcome this, we introduce Motion-Grounded Segment Anything (MoSA), a highly scalable unsupervised framework that learns a transferable objectness prior from unlabeled
videos. MoSA operates in three progressive stages: (1) automatically generating multi-granularity motion pseudo-labels from large-scale
video data; (2) training a Perceptual Grouping Model (PGM) via contrastive learning to internalize a generalized, appearance-driven concept of objects; and (3) transferring this learned prior into a
prompt-guided architecture for segment-anything-style inference on images. Extensive zero-shot evaluations across seven challenging
benchmarks (e.g., COCO and ADE20K) demonstrate that MoSA significantly outperforms existing unsupervised methods. Notably, despite using zero manual annotations, MoSA achieves
segmentation performance comparable to the fully supervised SAM. Our findings reveal that harnessing large-scale unlabeled motion is a feasible and highly scalable alternative to annotation-driven segment-anything pipelines.
Read the full abstract ▾
All authors Weijian Jian, Xiaoyue Zhang, Bin Xiao, Chunyu Xie, Yixiao He, Yutao Liu, Dawei Leng, Yuhui Yin
Machine Learning
CORD: Learning Reusable Degradation Representations Across Heterogeneous Physical Systems
Haibo Li and Zhiguo Zeng
Can heterogeneous physical degradation systems benefit from joint pretraining and move beyond system-specific prognostics toward reusable cross-system representation learning? CORD combines type-specific observation interfaces with a shared degradation backbone. Its two self-supervised objectives learn at complementary scales: Intra-Observation Structure Modeling (ISM) captures structure within observations, while Inter-Observation Dynamics Modeling (IDM) captures latent degradation
evolution across observation histories. We evaluate CORD under two transfer boundaries: Pretraining-Included System Types, where downstream
datasets and held-out units are unseen but their system types are represented during source pretraining, and Pretraining-Excluded System Types, where the entire turbofan-engine type is absent from pretraining. Across bearings, batteries, and cutting tools, CORD (Multi-domain) consistently improves over CORD (Single-domain) under Frozen adaptation, provides further gains under Full FT in most settings, and remains competitive with representative external baselines. Source-pretrained initialization also improves low-label adaptation to the pretraining-excluded engine type. Frozen-representation analysis further shows improved cross-unit lifecycle consistency after multi-domain pretraining. Joint pretraining across heterogeneous physical systems thus produces degradation representations reusable across devices,
datasets, and system types.
Read the full abstract ▾
Software Engineering
COMPASS: Predicting the Relationship of Multiple Patches for Vulnerabilities with LLMs
Yi Song, Dongchen Xie, Xiaoyuan Xie et al. (+4 more)
Modern software heavily relies on code reuse, so upstream vulnerability fixes do not automatically propagate to downstream codebases. Downstream maintainers must manually adopt patches to eliminate known risks. In practice, a single vulnerability often corresponds to multiple patches, which greatly complicates downstream patch adoption because different patch relationships imply different adoption strategies. To address this challenge, we first manually inspect large-scale multi-patch vulnerabilities (about 1K) in the real world and interview experienced developers, summarizing six typical types of patch relationships, i.e., Merge, Mirror, Better Solution, Fixing-of-Fixing, Collaboration, and Separation. Based on these observations, we propose COMPASS, an automated approach that predicts the relationships of multiple vulnerability patches with large
language models. Given a CVE as input, COMPASS follows a four-phase pipeline that (i) identifies the patch group and pre-scans explicit relationships, (ii) performs individual patch analysis, (iii) infers relationship instances via a hierarchy-guided
prompt, and (iv) validates completeness and consistency of the inferred results. As output, COMPASS reports the predicted relationships within the patch group and visualizes them as a relationship graph. We evaluate COMPASS on a
benchmark of 300 multi-patch CVEs and compare it against mainstream learning-based and LLM baselines. Results show that our method achieves strong and consistent prediction effectiveness and outperforms SOTA by 85.04% on average. We publicly release an online querying website to support community reuse of patch relationships knowledge: https://patch-relation.com.
Read the full abstract ▾
All authors Yi Song, Dongchen Xie, Xiaoyuan Xie, He Zhang, Lin Xu, Chunying Zhou, Zhi Jin
Analysis of PDEs
Overdetermined problem for optimal transportation
Qing Zhao and Feida Jiang
In this paper, we establish
symmetry results for solutions of overdetermined problems arising in optimal transportation. These problems involve the Monge-Ampère equation with a Dirichlet boundary condition $u=0$ on $\partial \Omega$ and the natural boundary condition $Du(\Omega)=\Omega^{*}$. We show that
symmetry holds when the target is the unit ball $B$. For more general target domains, including the cases $\Omega^*=\Omega$ and arbitrary $\Omega^{*}$,
symmetry is retained under an additional volume constraint on $\Omega$ and a boundary condition on $|Du|$. Finally, we introduce a
curvature-type overdetermined problem for the Monge-Ampère equation and obtain ellipsoidal or spherical
symmetry under an additional integral normalization condition. Our proofs rely on distinct techniques in different contexts: optimal transport and convex analysis for some cases, integral identities and isoperimetric inequalities for others, and P-function methods for the remaining cases. As a byproduct, in the $\tau=2$ case, a new maximum principle for the $P$-function $\phi(x)=\sum_{k,l=1}^{n}{\frac{\partial{S_{\tau}(D^2{u})}}{\partial{u_{kl}}}u_{k}u_{l}}-2\binom{n-1}{\tau-1}\int_{0}^{u}{f^{\frac{\tau}{n}}(t)\,dt}$ is established for arbitrary positive and nondecreasing $f$, which is of independent interest.
Read the full abstract ▾
High Energy Physics (Phenomenology)
Gravitational waves in the complex-singlet-extended Manohar-Wise model
Benhong Mei, Yabo Dong, Jun Fan et al. (+1 more)
We investigate cosmological phase transitions and
stochastic gravitational-wave signals in the Manohar-Wise model extended by a complex scalar singlet. Spontaneous breaking of a discrete Z3
symmetry produces domain walls, whose subsequent annihilation is driven by a small explicit
symmetry-breaking bias. Subject to theoretical and experimental constraints, we study two thermal histories: a one-step strong first-order electroweak phase transition satisfying the
baryon-number preservation criterion, and a two-step history whose first transition is a first-order transition along the singlet direction, leaving electroweak
symmetry unbroken. We estimate the gravitational-wave contributions from sound waves, magnetohydrodynamic
turbulence, and domain-wall annihilation. Within the adopted approximations, the phase-transition signals of the selected
benchmarks peak at approximately 10^-3 -10^-2 Hz and overlap the projected sensitivity regions of space-based observatories such as LISA, TianQin, and BBO. Choosing the
vacuum-energy bias to fix the domain-wall peak at approximately 2x10^-9 Hz, we find that the larger domain-wall tensions of the two-step
benchmarks yield signals reaching the projected SKA sensitivity, whereas the one-step sample remains below the
pulsar-timing sensitivities considered. These results illustrate the complementarity of millihertz and nanohertz gravitational-wave observations in probing phase transitions and discrete-
symmetry breaking in this model.
Read the full abstract ▾
All authors Benhong Mei, Yabo Dong, Jun Fan, Jingya Zhu
Lattice Field Theory
Marco Garofalo, Tobias Hartung, Timo Jakobs et al. (+3 more)
Hamiltonian simulations of
lattice gauge theories promise new insights into the inner workings of QCD. The rapid development of
quantum computing hardware suggests that large scale simulations may soon become feasible. However, no efficient formulation suitable for simulations near the continuum limit is currently known for the non-Abelian case, and existing candidates typically feature long range interactions in the electric part of the
Hamiltonian. We prove that such non- localities are unavoidable when the gauge configuration is reparametrised while preserving both the number of degrees of freedom and the fundamental commutation relations. We then attempt to mitigate this by introducing additional gauge links alongside additional constraints. Although this approach is successful in U(1), we prove that the corresponding construction is inconsistent for non-Abelian
lattice gauge theories.
Read the full abstract ▾
All authors Marco Garofalo, Tobias Hartung, Timo Jakobs, Paul Ludwig, Johann Ostmeyer, Carsten Urbach
Statistics Applications
Evidence-to-Decision (E2D): Structuring Action-Specific Evidence for Early-Phase Oncology Development
Haitao Pan
Early-phase oncology evidence is often summarized by asking whether a treatment is promising, but an encouraging signal can support several different next studies. We propose Evidence-to-Decision (E2D), a framework that organizes evidence around specific development actions. For each candidate action, E2D defines the contemplated future study, what would count as success, what evidence is needed before taking that step, what evidence is available now, and what assumptions connect the current and future settings. Statistical methods can then quantify action-specific support and sensitivity to those assumptions, while the final development choice remains multidisciplinary. We illustrate E2D in a rare pediatric oncology setting using predictive probabilities for same-setting single-arm expansion, same-setting randomized Phase II, movement to an earlier-line or frontline randomized study, and a separate de-escalation example. Calibrated support rules produced clinically distinct evidence states: intermediate evidence could support same-setting development without supporting frontline movement, safety concerns could narrow support despite strong activity, and stronger evidence could support several paths simultaneously. E2D therefore reframes the question from whether a treatment is generally promising to which next studies are supported by the evidence available now, what remains unsupported, and what additional evidence could change that assessment.
Read the full abstract ▾