Cryptography and Security
CaMeLoT: CaMeL orchestrated with Temporal logic for static verification and liveness
Elia Nikolaou, Magnus Wiik Eckhoff, Robert Flood et al. (+3 more)
LLM-based
agents generate and execute multi-step plans that invoke external tools which can access private data or execute commands. In this setting, security is a property of the entire execution that a plan creates, not just any single step. The plan itself is a critical artefact that captures the tool calls, control flow, and data dependencies. We present CaMeLoT, a complement to CaMeL, an existing defence against
prompt injection in tool-using LLM
agents. CaMeLoT extends CaMeL by adding a static verification layer that checks an
agent's plan before any tool is invoked. CaMeLoT translates a generated plan into a finite-state transition system, labels it with tool calls, provenance and taint information, and checks it against temporal policies expressed in CTL using the nuXmv model checker. Because verification happens before execution, unsafe plans are rejected without using LLM calls or tool calls, saving
tokens that runtime could have cost, as well as the need to unwind changes or teardown temporary sandboxes. When a verification fails, the model checker returns a counterexample to give feedback to the
agent to repair the plan. We evaluate CaMeLoT on policies derived from the AgentDojo
benchmark, SOC workflows, and
prompt-extraction experiments, showing that it verifies a broad class of temporal properties before execution while preserving CaMeL's runtime-checkable coverage.
Read the full abstract ▾
All authors Elia Nikolaou, Magnus Wiik Eckhoff, Robert Flood, Gudmund Grov, Vasileios Mavroeidis, David Aspinall
Sound
Beyond EER: Multi-Dimensional Evaluation of Information Leakage in Speaker De-Identification
Seungmin Seo, Oleg Aulov, P. Jonathon Phillips et al. (+2 more)
Speaker de-identification (SDID) aims to preserve
privacy by concealing speaker identity while maintaining
speech utility. However, current evaluations often reduce
privacy to a single dimension - biometric verification performance - typically measured by Equal Error Rate (EER). This narrow focus ignores critical leakage channels, such as soft biometric inference,
embedding-level re-identification, and structural template similarity, which threaten the unlinkability and irreversibility of biometric references. We propose a holistic evaluation framework across five complementary metrics: (i) EER, (ii) soft biometric leakage score , (iii) cumulative match characteristic re-identification analysis, (iv) canonical correlation analysis and Procrustes
embedding alignment, and (v) intelligibility via word error rate and semantic similarity. Evaluating five SDID systems from the IARPA ARTS program, we demonstrate that these metrics capture independent dimensions of information leakage. Our results indicate that reliance on a single metric can misrepresent the
privacy properties of an SDID system.
Read the full abstract ▾
All authors Seungmin Seo, Oleg Aulov, P. Jonathon Phillips, Kevin Mangold, Jonathan Eskin
Computation and Language
Selection Is Retrieval, Abstention Is Not: On-Device Tool Routing over 70 Korean-English Actions
Janghoon Lee
An AI assistant that calls tools makes two decisions on every request: which tool to invoke, and whether any available tool applies. In the usual design a single
language model makes both, by emitting a call or by declining to emit one. On a device that has to answer without a server, the
language model is what makes that design expensive, dominating both the latency and the
memory of the router. The common alternative is to remove the model completely and rank the catalog of local actions with a retriever instead. That substitution is not symmetric across the two decisions. A retriever returns its highest-scoring candidate for every input and cannot signal that the catalog holds no valid action. Our earlier study found that constraining a decoder to a tool grammar repairs malformed output without improving the choice. What the substitution costs in each decision has not been measured. We evaluate the two decisions separately over 600 Korean and English requests and a catalog of 70 local actions. The router may also ask for a missing slot, reply, or delegate. Half the in-catalog requests reuse catalog vocabulary and half paraphrase it, separating lexical overlap from the action requested. Character 3-gram BM25 selects 162 of 164 lexically matched requests and 85 of 166 paraphrases. Restricting the candidate set to seven raises the paraphrase figure to a mean of 0.825 over five trials. No classifier over its score features separates in-catalog from out-of-catalog above 0.697 area under the curve, where the frozen encoder
multilingual-e5-base reaches 0.806. Using that encoder for abstention alone keeps 376 of the requests local and misroutes 9 of the 150 needing delegation. Abstention, not selection, is where a neural component is required. A neural ranker improves every quality metric and is rejected on latency and
memory rather than accuracy.
Read the full abstract ▾
Quantum Physics
Theory-agnostic nonclassicality certification in an integrated photonic circuit
Vinicius P. Rossi, Emanuele Polino, Beatrice Polacchi et al. (+8 more)
Certifying nonclassicality without assuming any particular underlying physical theory constitutes both a foundational challenge and a requirement for device-independent protocols. Realizing such theory-independent certification in scalable architectures is essential for connecting fundamental tests with emerging
quantum technologies. Integrated photonic
circuits are a leading platform for scalable
quantum information processing, motivating the development of rigorous methods to certify the nonclassical resources underpinning
quantum advantage. In this paper, we report a theory-independent certification of generalized contextuality on a three-mode integrated photonic
circuit. Our approach combines theory-agnostic
tomography with simplex-embeddability certification—a framework that requires no assumptions about the underlying physical theory—and applies it to experimental data from a three-mode photonic
circuit seeded by single
photons. An independently constructed
quantum model of the experiment provides a consistency check on the observed nonclassicality, without entering the certification as a theoretical assumption. The method rules out simplex-embeddability for our experimental data, providing robust, theory-independent evidence that our photonic
circuit exhibits nonclassicality.
Read the full abstract ▾
All authors Vinicius P. Rossi, Emanuele Polino, Beatrice Polacchi, Valeria Cimini, David Schmid, John H. Selby, Giacomo Corrielli, Andrea Crespi, Roberto Osellame, Fabio Sciarrino, Ana Belén Sainz
Analysis of PDEs
Sharp Strichartz estimates on the one-dimensional torus
Yangkendi Deng, Chenjie Fan, Shaoming Guo et al. (+2 more)
In this article, we prove the sharp Strichartz estimates for the Schrödinger equation on the one-dimensional
torus. More precisely, via obtaining a sharp weak $L^{6}$ estimate with no derivative loss, we obtain sharp $L^{6}$ Strichartz estimate, and sharp no loss $L^{p}$ Strichartz estimate for any $2<p<6$. The latter answers a question by Bourgain \cite{Bourgain93}.
Read the full abstract ▾
All authors Yangkendi Deng, Chenjie Fan, Shaoming Guo, Zihua Guo, Yongming Luo
Robotics
M$^3$P-R1: Reinforcement Learning for Large Language Model Guided Multi-Modal Motion Planning via MIP Code Generation
Xingpeng Sun, Zherong Pan, Kai Cheng et al. (+3 more)
Multi-Modal Motion Planning (M$^3$P) requires joint
reasoning over continuous motions and discrete mode transitions, making it difficult to solve efficiently. For instance, a bipedal
robot may walk to a target location and then use its arms to grasp an object. This scenario captures both mode transitions and continuous dynamics, yielding feasible paths that neither purely discrete nor continuous planners can handle. While Mixed-Integer Programming (MIP) offers a principled framework, constructing tractable formulations for non-convex problems is typically manual and domain-specific, especially in the approximate, discretization-based MIP regime needed for non-convex robotic tasks. We propose M$^3$P-R1, a reinforcement learning method that fine-tunes large
language models (LLMs) to decompose M$^3$P tasks into MIP variables, constraints, and objectives. Instead of directly outputting answers, which are often prone to
hallucination, the model generates executable Python code using MIP optimization libraries and constraint interfaces. This enables solver-backed execution for robust and verifiable solutions. Trained with an outcome-driven reward against the solver, M$^3$P-R1 learns to compose modality-level discretization primitives and synthesize cross-modal coupling constraints, producing executable MIP programs for complex M$^3$P tasks.
Read the full abstract ▾
All authors Xingpeng Sun, Zherong Pan, Kai Cheng, Xindi Tang, Syed Talha Bukhari, Aniket Bera
High Energy Physics (Theory)
The geometry of multiloop Feynman Integrals from the Scattering Facet
José Ríos-Sánchez and Germán Rodrigo
The Scattering Facet (SF) of the Cosmological Polytope is a positive geometry defined from a Feynman diagram and encoding its combinatorics. Its canonical differential form reproduces the result of the integration over the energy components of loop momenta in scalar Feynman integrals, and is therefore closely related to the Loop-Tree Duality (LTD). In this article, we present a novel description of the SF as a base-fiber geometry. The fibers are a family of polytopes called deformed graphical zonotopes, which have a natural interpretation in terms of the causal
evolution of the underlying Feynman diagram, and capture interesting information on scattering amplitudes. This decomposition of the SF induces a description of the canonical form leading to the so-called causal representation. Moreover, it provides new mathematical tools that allow us to characterize all its regular triangulations. Canonical forms obtained from these triangulations correspond to spanning tree representations, matching the original LTD amplitudes expressed as a sum over spanning trees. This geometrical viewpoint provides us with new methods to obtain the causal and spanning tree representations, without the need of performing any integration.
Read the full abstract ▾
Computer Vision
Video-Based Markerless Motion Capture for Clinical and Rehabilitation Biomechanics: A PRISMA-ScR Scoping Review of Validated Architectures, Clinical Readiness, and Emerging Methods
Florian Delaplace, Elodie Piche, Frédéric Chorin et al. (+1 more)
Background..
Video-based markerless motion capture promises movement analysis without the cost, space and skin-marker constraints of optoelectronic systems, with particular potential for clinical and rehabilitation settings. Whether validated pipelines yet deliver clinically acceptable biomechanics, and how they relate to the underlying computer-
vision research, remains unclear. Methods. We conducted a scoping review following the PRISMA extension for Scoping Reviews, with a registered protocol and searches of PubMed, Scopus and IEEE Xplore (January 2015 to February 2026; the computer-
vision scan was updated to July 2026). A dual-tier design paired a primary corpus of validated biomechanical studies with a complementary, curated and deliberately non-exhaustive corpus of emerging computer-
vision work, used qualitatively. We charted study characteristics, pipeline architecture, validation methods and joint-angle accuracy. Results. We included 117 studies, most published from 2024 onward and conducted on healthy adults walking in a laboratory. Pipelines formed five architectural families across
monocular and multi-camera modalities; most reported raw joint angles without biomechanical refinement. Sagittal lower-limb agreement clustered around 5 to 6{\textdegree}, generally short of clinical acceptability, while out-of-plane kinematics, kinetics, and pathological or older populations were rarely validated. Emerging computer-
vision building blocks (foundation-model mesh recovery, differentiable inverse kinematics,
video-based kinetics) were almost absent from validated studies. Conclusions.
Video-based markerless capture is not yet interchangeable with marker-based systems for clinical joint kinematics, and it remains barely validated where rehabilitation needs it most: older and pathological populations, out-of-plane kinematics, and kinetics. Mapping this evidence gap onto emerging computer-
vision advances, we propose hypothesis-generating design guidelines, not a validated method, to steer the next generation of pipelines toward accessible, clinically meaningful movement analysis.
Read the full abstract ▾
All authors Florian Delaplace, Elodie Piche, Frédéric Chorin, Raphael Zory
Quantum Physics
Impact of a CSS quantum error correction code in underwater quantum key distribution
Juliette Florin, Nicolas Le Josse, Arnaud Coatanhay et al. (+1 more)
Quantum key distribution (QKD) enables secure
underwater communications essential for maritime infrastructure.
Underwater optical channels introduce substantial
photon loss (erasures) and ambient noise that degrade QKD performance. This paper investigates whether two four-
qubit Calderbank-Shor-Steane (CSS)
quantum error correction codes (QECC) mitigate these impairments in vertical
underwater communication BB84 QKD protocol. After developing a comprehensive
stochastic channel model incorporating
photon loss,
geometric spreading, and
solar noise, we assess the viability of QECC through the analytical study of the
quantum bit error rate (QBER) and the secure key rate (SKR) with and without security depending on the signal-to-noise ratio (SNR) validated against Monte Carlo simulations. The standard four-
qubit CSS code achieves a 3 dB SNR gain at the QBER 11% security threshold; the discard code variant achieves 4.5 dB. However, QECC includes an encoding overhead that reduces the SKR. We demonstrate a crucial relationship between the SKR and the probability of arrival of the sent
photon. Analysis shows that QECC is beneficial exclusively in marginal SNR regimes; at high SNR, raw BB84 dominates. For an ocean Type III Jerlov water scenario with sunlight coming from the sun located at the top of the atmosphere, we identify operational depth-range windows where QECC enables communication otherwise infeasible. This analysis establishes that error correction deployment must be scenario-dependent: extend operational range at the cost of throughput when SNR is marginal, or prioritize key generation rates at high SNR.
Read the full abstract ▾
All authors Juliette Florin, Nicolas Le Josse, Arnaud Coatanhay, Gilles Burel
Number Theory
Stable Trace Formula for Newton strata of Shimura varieties
Dhruva Kelkar
Let $(G,X)$ be a Shimura datum of abelian type satisfying the hypotheses of the main
theorem, and suppose that the associated Shimura varieties have hyperspecial good reduction at $p$. Their special fibres are stratified by the $\sigma$-conjugacy classes $b\in B(G_{\mathbb{Q}_p},\mu_h^{-1})$. For an individual Newton stratum, this paper constructs a stabilized formula for the alternating traces of Frobenius–Hecke correspondences on its compactly supported $\ell$-adic
cohomology, with coefficients in an $\ell$-adic local system. The formula, obtained by modifying the Langlands–Kottwitz method, expresses these Lefschetz numbers as
elliptic stable
geometric distributions on endoscopic groups of $G$, placing the cohomological traces in a form suitable for comparison with automorphic spectral data. We also give a group-theoretic criterion determining which
elliptic endoscopic groups can contribute to a given Newton stratum. For certain
unitary Shimura varieties, we exhibit an intermediate Newton stratum with no non-trivial endoscopic contribution, although there are non-trivial endoscopic contributions in the
cohomology of the ambient Shimura variety.
Read the full abstract ▾
Differential Geometry
Mathias Braun and Xue-Mei Li
We introduce a codimension one construction of coarse
Ricci curvature in Lorentzian geometry. Using the $1$-Lorentz-Wasserstein distance $\ell_1$, we compare probability measures supported on small spacelike hypersurfaces through nearby events and recover, in a precise asymptotic regime, the ambient
Ricci curvature in future-directed unit timelike directions, with a universal dimensional prefactor reflecting this codimension one construction. The construction echoes the Raychaudhuri equation, which relates the
evolution of spatial volume expansion along timelike geodesics to
Ricci curvature. Our quantitative estimates rely on transport maps with controlled displacement, constructed through a Moser-type flow. At leading order, the slice construction is insensitive to smooth weights. By smearing the spacelike slices into thin timelike tubes and calibrating the temporal and spatial scales, we recover the timelike Bakry-Émery tensor $\mathrm{Ric}+\mathrm{Hess}\,V$ associated with the weighted reference measure $\smash{\mathfrak{m}=\mathrm{e}^{-V}\,\mathrm{vol}_g}$.
Read the full abstract ▾
Robotics
VLA-ULAP: Interleaving Cloud VLA Calls with Ultra-Lightweight Local Action Prediction at the Edge
Deyu Cao, Ryuji Oi, Kosuke Matsushima et al. (+4 more)
Billion-parameter
vision-
language-action (VLA) policies run either onboard, consuming substantial power, or on remote servers, adding communication latency. To address these drawbacks and better balance latency and onboard energy consumption, we propose VLA-ULAP. It partitions inference across decision times, interleaving remote VLA calls with predictions from an Ultra-Lightweight Local Action Predictor (ULAP). A single ULAP has $\sim$7.4M parameters including the frozen
vision encoder. It combines current views, proprioception, and executed action history to predict chunks in one pass. Trained independently, it requires no VLA hidden states, online verification, or server round trips. On Jetson Orin Nano, ULAP takes 20.7 ms and 0.122 J of idle-subtracted energy per inference, versus 289.3 ms and 40.46 J for GR00T on RTX A6000. Across four simulated base-policy/
benchmark pairs, VLA-ULAP removes 45.3-77.3% of VLA calls while retaining 95.0-98.5% of baseline success rates at selected operating points. On VLA-JEPA, VLA-ULAP also surpasses local acceleration alternatives, using an estimated 49.4% less inference time and 51.5% less GPU energy per successful episode than ACT, and 77.5% less time and 80.7% less energy than SP-VLA at higher success rates. In physical SO-101 trials, it similarly removes 70.3-71.9% of VLA calls without observed success-rate loss at seen or held-out placements. Measured device costs imply 64.6-66.4% less inference time and 70.1-71.7% less idle-subtracted energy per successful episode at these call counts. Beyond these savings, faster responses help VLA-ULAP exceed $\pi_{0.5}$'s success rate by 11.0 and 15.5 percentage points (pp) on two tasks in latency-aware LIBERO-Safety simulation while approximately halving VLA calls.
Read the full abstract ▾
All authors Deyu Cao, Ryuji Oi, Kosuke Matsushima, Yuxuan Pan, Ziheng Wang, Daichi Fujiki, Atsutake Kosuge
Hardware Architecture
Automated Instruction Encoding Synthesis for Modern GPU ISA Compression
Mingyuan Ma and Hu He
Modern GPU kernels increasingly stress the instruction supply path, while fixed instruction containers can leave substantial footprint slack. This paper presents an automated encoding-synthesis framework that treats instruction layout as a constrained slot-assignment problem over a validated instruction-form field specification. The formulation separates semantic field identity from physical bit positions and supports tied, pinned, and free placement constraints, making it applicable when recurring decoded fields are not frozen by a public format contract. We instantiate the framework for NVIDIA SASS: raw public encoding text is normalized into a machine-readable specification, a SASS disassembler is validated against nvdisasm on 3.78M instructions, released as an open
benchmark, and CP-SAT synthesis is used for fixed-length and variable-length encodings. On 142 Blackwell kernel inputs, variable-length synthesis reduces instruction footprint by 33%, with comparable reductions after re-synthesis on Ampere and Hopper; fixed-length synthesis on the same specification reduces decoder area by 16% against a decoder generated from the NVIDIA-observed 128-bit layout by the same generator and flow. Generated fetch/decode RTL meets 1.5 GHz in TSMC 22 nm with a replicated area delta of 0.12% of a GA100-class die; a same-node SRAM comparison shows the footprint reduction corresponds to about 9x this added logic in instruction-SRAM bit-
cell area.
Read the full abstract ▾
Mathematical Physics
Contact canonoid maps and their conserved and dissipated quantities
R. Azuaje
We study conserved and dissipated quantities associated with contact canonoid transformations and their extension to possibly singular smooth maps. The defining condition is imposed on the pulled-back contact one-form, without requiring it to remain contact. A
polynomial construction produces first integrals from one-forms having the same dissipation rate as the original contact form. Two normalizations extend this construction to maps whose associated
Hamiltonians have different dissipation rates. On the set where the associated
Hamiltonian is nonzero, we construct a smooth invariant endomorphism whose traces are conserved. This construction includes zeros of the original
Hamiltonian, where the endomorphism vanishes. We relate the tensor traces to the coefficients of the
Hamiltonian-normalized
polynomial and obtain an upper bound of n functionally independent trace invariants in dimension 2n + 1. A five-dimensional singular map attains this bound. Examples include singular maps for the damped free particle, a transformation with unequal dissipation rates, the damped harmonic oscillator, and motion under
gravity with linear friction.
Read the full abstract ▾
Analysis of PDEs
Polyconvexity for incompressible inversion-symmetric energies of Valanis-Landel type
Ionel-Dumitrel Ghiba, Maximilian P. Wollner and Patrizio Neff
{Let $\lambda_i = \nu_i(F)$ denote the three singular values of the deformation gradient $F \in {\rm GL}^+(3)$.} We consider the family of incompressible isotropic energies $ W_\psi(F)=\sum_i \psi\left(|\!\log\lambda_i|\right)$ with $\psi:[0,\infty)\mapsto\mathbb{R}$. Set $g(s)=\psi\left({\rm arcosh}\frac{s}{2}\right)$ for all $s\geq2$. If $g$ has a convex and non-decreasing extension $\bar g$ to $[0,\infty)$, then $W_\psi$ is the restriction to ${\rm SL}(3)$ of the explicit polyconvex function \ {$ F\mapsto\sum_i \bar g\left(\nu_i(F+{\rm Cof} F)\right). $} The proof uses the identity $\nu_i(F+{\rm Cof} F)=\lambda_i+\lambda_i^{-1}$, up to permutation, on ${\rm SL}(3)$ and Ball's convexity
theorem for functions of the singular values. We also give a direct proof along rank-one lines contained in ${\rm SL}(3)$ and derive a convenient one-dimensional differential sufficient condition. In particular, $ F\mapsto\sum_i e^{\log^2\!\lambda_i} $ is rank-one convex on ${\rm SL}(3)$ and possesses the stated polyconvex extension. A simple-shear computation shows that scalar convexity in $\log \lambda_i$ alone is insufficient; the quadratic Hencky energy $\sum_i \log^2\lambda_i$ is not rank-one convex on ${\rm SL}(3)$.
Read the full abstract ▾
Instrumentation and Methods for Astrophysics
The Nancy Grace Roman Space Telescope Coronagraph Community Participation Program
Dmitry Savransky, Vanessa P. Bailey, Schuyler G. Wolff et al. (+53 more)
In preparation for the operational phase of the Nancy Grace Roman Space
Telescope, NASA has created the Coronagraph Community Participation Program (CPP) to prepare for and execute Coronagraph Instrument technology demonstration observations. The CPP is composed of 7 small, US-based teams, selected competitively via the Nancy Grace Roman Space
Telescope Research and Support Participation Opportunity, members of the Roman Project Team, and international partner teams from ESA, JAXA, CNES, and the Max Planck Institute for Astronomy. The primary goals of the CPP are to prepare simulation tools, target databases, and data reduction software for the execution of the Coronagraph Instrument observation phase. Here, we present the current status of the CPP and its working groups, along with plans for future CPP activities up through Roman's launch. We also discuss plans to potentially enable future commissioning of currently-unsupported modes.
Read the full abstract ▾
All authors Dmitry Savransky, Vanessa P. Bailey, Schuyler G. Wolff, Maxwell A. Millar-Blanchaer, Jason Wang, Lisa Altinier, Ramya Anche. Pierre Baudoz, Beth Biller, Sarah Blunt, Wolfgang Brandner, Marah Brinjikji, Oscar Carrión-González, Amanda Chavez, Elodie Choquet, David Doelman, Julien H. Girard, Alexandra Z. Greenbaum, Samantha N. Hasler. Justin Hom, James G. Ingalls, Stephen R. Kane, N. Jeremy Kasdin, Oliver Krause, Masayuki Kuzuhara, Alexis Lau, Zhexing Li, John Livingston, Patrick J. Lowrance, Kevin Ludwick, Bruce Macintosh, Eric Mamajek, Mark Marley, Johan Mazoyer, Bertrand Mennesson, Toshiyuki Mizuki, Sarah E. Moran, Naoshi Murakami, Jun Nishikawa, Malachi Noel, Laurent Pueyo, Sergi Hildebrandt Rafels, Jason Rhodes, Tyler Robinson, Robert J. De Rosa, Matthias Samland, Nicholas Schragal, Jürgen Schreiber, Jennifer Sobeck, Karl Stapelfeldt, Motohide Tamura, Taichi Uyajma, Arthur Vigan, Michele Woodland, Marie Ygouf, Kenta Yoneta, Robert T. Zellem, Neil T. Zimmerman
Cryptography and Security
A Security Risk Assessment Framework for AI-Powered Development Tools
Salem AlJanah
AI-powered development tools are now widely used to generate code and assist developers with routine programming tasks. Although existing work has identified vulnerabilities in AI-generated code, security-oriented work is often focused on vulnerability detection rather than risk assessment. To address this gap, this paper presents a Security Risk Assessment Framework (SRF) to evaluate the security risks of AI-generated code. SRF combines threat modeling, security analysis, and a quantitative risk evaluation approach based on vulnerability criticality. The framework is applied to a set of security-relevant programming tasks, where code generated by multiple AI-powered development tools is analyzed using Bandit and Semgrep. The results show that AI-generated code can introduce security vulnerabilities across all evaluated tools. They also show that risk levels vary by task type, as input processing and file handling tasks showed higher risk, while simpler tasks remained low-risk. Differences between tools exist but are smaller than differences across task categories. Overall, SRF enables reproducible evaluation of AI-generated code and provides a practical framework for assessing its security implications.
Read the full abstract ▾
Numerical Analysis
Pressure-robustness by commuting interpolation operators for Stokes discretizations with continuous pressures
Philip L. Lederer and Theresa Vock
Common finite element discretizations for the incompressible Stokes equations with a continuous pressure approximation – the MINI element, the Taylor–Hood element, or stabilized equal-order elements – are not pressure-robust: the velocity error is polluted by the pressure best-approximation error, multiplied by the inverse viscosity. The established remedy replaces the test function in the momentum equation by a divergence-preserving reconstruction operator; for continuous pressure approximations this construction is cumbersome, requiring the solution of local problems (Lederer, Linke, Merdon & Schöberl, SIAM J. Numer. Anal., 2017). We propose a simpler alternative: interpolating the force by a Nédélec interpolation operator of suitable order instead of reconstructing the test function. Exploiting the commuting diagram property linking Nédélec and Lagrange interpolation operators, we show that the resulting method is pressure-robust for any $H^1$-conforming, continuous-pressure discretization. We give the complete analysis for the MINI element, and then for stabilized equal-order $P^kP^k$ elements of arbitrary degree $k$, proving that the interpolation-induced consistency error is pressure-robust and converges of optimal order. We validate the theory by several numerical examples.
Read the full abstract ▾
Machine Learning
Revisiting Distributed Sign-Based Variance Reduction
Wei Jiang, Zechao Li and Lijun Zhang
Sign-based methods reduce communication costs in distributed environments, but aggregating local signs can introduce bias when data are heterogeneous. As a result, existing sign-based variance reduction methods fail to obtain the optimal convergence rates. In this paper, we solve this problem and obtain optimal rates for both nonconvex
stochastic and finite-sum optimization. We first give a counterexample showing that majority voting can fail to approach stationary points even with exact local gradients. Motivated by this limitation, we propose tracking the global gradient at the server through unbiased compression of recursive gradient increments. As a result, we can obtain the convergence rates of $O(\sqrt{d/K}+\sqrt d (a/(nK))^{1/3})$ for the $\ell_1$-norm and $O(\sqrt{a/K}+\sqrt a/(nK)^{1/3})$ for the $\ell_2$-norm. Here, $K$ is the iteration number, $n$ is the number of workers, $d$ is the dimension, and $a=1+\omega$, with $\omega$ denoting the compressor's relative variance. For finite-sum problems with $M$ components, we combine periodic exact gradient refreshes with compressed component-gradient differences. The resulting total sample complexities are $O(M+d\sqrt{aM}\epsilon^{-2})$ and $O(M+a\sqrt M\ epsilon^{-2})$ for $\ell_1$ and $\ell_2$ gradient norms at most $\epsilon$, matching the corresponding bounds in centralized settings.
Read the full abstract ▾
Machine Learning
Learning to Program Adaptive Non-Local Observables for Machine Learning
Yu-Ting Lee, Samuel Yen-Chi Chen and Huan-Hsin Tseng
Quantum neural networks (QNNs) are typically built from variational
quantum circuits (VQCs), which are limited by local measurements. Adaptive non-local observables (ANO) address this by jointly optimizing
circuit parameters and multi-
qubit measurements. However, existing ANO-based VQCs learn only a single static observable that remains invariant across all inputs. We propose QFWP-ANO, a novel architecture which employs a classical hypernetwork to dynamically program VQC parameters and/or non-local observables conditioned on each input. On multivariate time-series
forecasting across four ETT
datasets, QFWP-ANO achieves the lowest MSE in 16 of 20 settings and second-lowest in the remaining four, surpassing ANO-based and other strong baselines. On reinforcement learning tasks, QFWP-ANO consistently surpasses ANO-VQCs. Our results establish input-conditioned ANO as an effective approach for enhancing QNNs.
Read the full abstract ▾
Chemical Physics
Variational computation of anharmonic ground and excited vibrational eigenstates using bound Quartic Force Fields: Application with MCTDH and ElVibRot
Ali Al-Jaaidi, David Lauvergnat and Daniel Peláez
In this work we introduce the use of Quartic Force fields (QFF) potential expansions in the context of variational calculations. Such potentials are commonly employed in molecular Vibrational Second-Order Perturbation Theory (VPT2) studies, for which equations explicitly dependent on the QFF parameters exist. However, QFF are unbound potentials for more or less large displacements from the reference point and, most of the time, this prevents their use in conjunction with variational wavepacket-based calculations. In this work, we propose a general correction to QFFs and introduce a fully automated numerical approach to avoid their unbound character. Our corrected potentials, bound QFF (bQFF), do not exhibit appreciable modification of the local topography around the region of interest for infrared spectroscopy. As a consequence of this, we can affirm that the vibrational eigenstructure (
eigenvalues, eigenstates) remains essentially unaltered by our correction. To illustrate their numerical stability, we have interfaced our bQFF routines in combination with to two well-established
quantum simulation software packages MCTDH and ElVibRot which feature variational approaches. More specifically, our bQFFs are separable and hence directly expressible as MCTDH operators. Furthermore, concerning the size of our bQFF expansion, we show that it is possible to tensor-decompose our bQFF in Canonical Polyadic form (CP-bQFF). We use the Monte Carlo Canonical Polyadic decomposition
algorithm for this. CP-bQFF results are virtually identical to uncompressed bQFF, but the computational efficiency is largely improved. Our approach paves the way for the automated variational study of anharmonic eigenstates in molecular systems within the reach of QFF-based potentials, using either time-dependent or time-independent schemes.
Read the full abstract ▾
Materials Science
Adaptive Substrate Support Based on Thin-Film Piezoelectric Actuators
Ertuğ Şimşek, Bas Jansen, Marcelo Ackermann et al. (+1 more)
Advanced lithography scanners require extreme substrate flatness in the range of nanometers to prevent focus errors. Such a flatness is challenging to reach and maintain using static substrate supports. Active correction concepts using bulk piezoelectric or linear actuators become prohibitive due to wiring and volume limitations, considering the thousands of pillars required. In this work, we present to our knowledge the first ever demonstration of an active substrate support concept based on thin-film piezoelectric actuators. The working principle, fabrication steps, and characterization of a proof-of-principle device, supported by finite element modeling (FEM), are presented in detail. The fabricated device shows a displacement above 9 nm, which is sufficient to correct some of the focus errors in lithography systems. We foresee that this new concept may open the way towards smaller chip dimensions and improved yield.
Read the full abstract ▾
All authors Ertuğ Şimşek, Bas Jansen, Marcelo Ackermann, Muharrem Bayraktar
Numerical Analysis
A Reynolds-Semi-Robust, Globally Divergence-Free HDG Method for the Smagorinsky Model
Shuaijun Liu and Xiaoping Xie
We develop and analyze a fully discrete, globally divergence-free hybridizable discontinuous Galerkin (HDG) method for a gradient-based Smagorinsky model. The method combines backward Euler time stepping, interior-penalty discretizations of molecular and nonlinear eddy
diffusion, and an upwind convective flux. The discrete velocity is $H(\operatorname{div})$-conforming and pointwise divergence-free, which yields pressure robustness. For sufficiently large penalty parameters, we prove unconditional energy stability and existence of a discrete solution, and establish uniqueness under additional smallness conditions. A velocity error estimate is derived without explicit inverse powers of the molecular viscosity. The nonlinear facet residuals are controlled using local trace-approximation estimates and a viscosity-independent facet penalty. We retain the dependence of the discrete Gronwall factor on the filter scale and the mesh size; a mesh-uniform bound follows under suitable solution regularity, a fixed time-step margin, and the scaling $\delta=O(h)$ on quasi-uniform meshes. The reported manufactured-solution results are consistent with the resulting pre-asymptotic error bounds. Further flow examples illustrate the dissipative behavior of the method and are distinguished from the boundary conditions and parameter range covered by the analysis.
Read the full abstract ▾
Robotics
FIERCE: From Generalist Robot Policies to Fast Specialists via Progress-Failure Feedback
Runjia Tan, Yuang Tu, Yujie Yan et al. (+3 more)
Generalist
robot policies offer useful initialization, but refining compact specialists through limited physical interaction requires informative learning feedback. We present FIERCE, a generalist-initialized reinforcement learning framework centered on a unified, task-adaptive progress-failure evaluator. Its architecture shares an observation-
language representation between an observed-progress head and an action-conditioned latent predictor whose past and current predictions feed a causal sequence head for task-failure estimation. Joint supervision from progress and preference labels, synchronized commands and observations, and terminal outcomes trains the evaluator; target-task rollouts support adaptation and calibration. Fixed evaluator snapshots provide progress shaping and failure-risk penalties alongside independently verified terminal rewards, while evaluator and policy updates alternate as new experience is collected. Refinement requires neither continued generalist action queries nor a dedicated target-task simulator or manually annotated dense rewards. Only the compact specialist is retained at deployment. The evaluation separates feedback quality, policy-learning efficiency, and deployment cost across simulation and two contact-rich real tasks. Code, model weights, and data-restoration tools are released at https://github.com/ar-mine/FIERCE.
Read the full abstract ▾
All authors Runjia Tan, Yuang Tu, Yujie Yan, Lan Yu, Xuesong Tian, Chen Lv
Robotics
From Gameplay to Policy: Towards Scalable Robot Data Collection via Gamified Robot-Free Interaction
Zheng Li, Liang Zhu, Junzhe Wang et al. (+13 more)
Learning generalizable
robot manipulation policies requires large-scale and diverse interaction data, yet collecting real-world demonstrations remains costly and difficult to scale. Existing approaches to data collection are either dependent on specific
robot hardware that limits crowdsourcing and transferability, or suffer from incomplete annotation and limited behavioral diversity. Inspired by how games sustain long-term human engagement, we explore an alternative paradigm that turns data collection into an engaging gameplay experience and transfers the resulting human manipulation experience to real
robots. We present Project Kitchen, a VR-based gamified
egocentric data collection platform that elicits diverse, goal-directed manipulation while remaining independent of specific
robot embodiments and hardware, making it applicable to broader and potentially large-scale deployment. To bridge the game-to-real gap, we further introduce Game2Policy, which extracts embodiment-invariant affordance cues, including contact points and sub-goal states, from gameplay trajectories. An affordance model is pre-trained on game-collected data and then jointly fine-tuned with downstream policies using only a handful of real-
robot demonstrations. Experiments show that Game2Policy improves average success rates by 10.0 points in simulation and 18.3 points on real
robots in the few-shot setting. User studies and quantitative analyses further show that Project Kitchen promotes diverse manipulation behaviors and provides an engaging data collection experience. These results demonstrate the potential of gamified virtual environments as a scalable source of manipulation knowledge. The platform and code will be released upon acceptance.
Read the full abstract ▾
All authors Zheng Li, Liang Zhu, Junzhe Wang, Huayuan Chen, Ziyun Liu, Jiahang Cao, Xinyu Sheng, Pei Qu, Yufei Jia, Ximeng Zhang, Jiarui Xie, Zizhao Yuan, Haoang Li, Yi Cai, Jinni Zhou, Jun Ma
Computation and Language
DyMT-ESB: Dynamic Multi-Turn Evaluation of Social Bias in User-LLM Interactions
Rem Hida, Masahiro Kaneko, Daisuke Oba et al. (+2 more)
Warning: This paper contains examples of stereotypes and social bias. LLMs are increasingly used in interactive settings by the general public, making the evaluation of model behavior in multi-turn conversational scenarios important for safety, including stereotyping-related harms. However, existing multi-turn social bias evaluations often rely on pre-specified or template-based user inputs that do not adapt to model responses and typically assume a fixed dialogue length in advance. In this paper, we study social bias dynamics in response-conditioned multi-turn interactions using a controlled evaluation protocol that generates follow-up user queries from the evolving dialogue history and allows evaluation over variable numbers of turns. Experimental results show that LLMs exhibit social bias even in coherent, response-conditioned multi-turn interactions, revealing late-emerging bias, non-monotonic bias patterns, and bias re-emergence. These results motivate evaluations that extend beyond fixed-turn, pre-scripted protocols. Our findings highlight the importance of analyzing social bias as a turn-level dynamic phenomenon.
Read the full abstract ▾
All authors Rem Hida, Masahiro Kaneko, Daisuke Oba, Danushka Bollegala, Naoaki Okazaki
Combinatorics
Sim-Width, Induced Matching Treewidth, and Tree-Independence Number in Induced $K_{t,t}$-Free Graphs
Mengyuan Niu and Xiumei Wang
The tree-independence number $tree\text{-}\alpha(G)$, the induced matching treewidth $tree\text{-}\mu(G)$, and the sim-width $simw(G)$ are graph parameters defined in terms of tree or branch decompositions. We establish two
polynomial bounds for the tree-independence number of induced $K_{t,t}$-free graphs, one in terms of sim-width and the other in terms of induced matching treewidth. Abrishami et al. (SIDMA, 2025) and Brettell et al. (EJC, 2025) asked whether bounded sim-width, together with the exclusion of an induced $K_{t,t}$, implies bounded tree-independence number. We answer this question by proving that, for integers $t\geq 2$ and $s\geq 1$, every induced $K_{t,t}$-free graph $G$ with $simw(G)\leq s$ satisfies $tree\text{-}\alpha(G)=O_t\left((s+1)^{2t^2-2t}\right)$. This also proves a
polynomial strengthening of a
conjecture of Bešter Štorgel et al. (arXiv, 2026) concerning induced $K_{1,t}$-free graphs and improves a
theorem of Alon et al. (arXiv, 2025) by reducing the exponent from $3t^2+1$ to $2t^2-2t$. Alon et al. (arXiv, 2025) asked whether, for fixed induced matching treewidth, the tree-independence number is polynomially bounded in $t$. Using a VC-dimension argument, we answer this question affirmatively by showing that, for integers $\mu\geq 1$ and $t\geq 2$, every induced $K_{t,t}$-free graph $G$ with $tree\text{-}\mu(G)\leq\mu$ satisfies $tree\text{-}\alpha(G)=t^{O_\mu(1)}$.
Read the full abstract ▾
Materials Science
Low Temperature Halide Assisted HVPE Growth of Single Crystal AlN Films
K. Udwary, M. Smeaton, J.S. Mangum et al. (+6 more)
Low defect, single polarity aluminum nitride layers grown by halide vapor phase epitaxy (HVPE) methods have classically needed growth temperatures well exceeding 1100{\deg}C with many of the best results being grown in the range of 1400{\deg}C. These high temperatures have typically been required to obtain Al-polar AlN films with smooth morphologies and high crystalline quality. However, these elevated temperatures make the use of quartz reactor chambers difficult, as quartz begins to soften at 1200{\deg}C, and requires specially designed reactors. This study examines the use of anhydrous hydrogen chloride (HCl) gas, injected directly to the growth surface, to suppress inverted polarity grains that form at low growth temperatures. This method allows for high quality, single
crystal AlN layers with a growth temperature of just 1000{\deg}C in a quartz reactor chamber. With this method, AlN layers of 5 um thickness were grown on c-plane (0 0 0 1) sapphire, with a growth rate of 20 um/hour, and a resulting full-width at half-maximum (FWHM) of symmetric (0 0 0 2) rocking curve reflection under 400 arcsec and asymmetric (1 0 -1 2) reflection under 1000 arcsec. Electron backscatter diffraction measurements showed the suppression of inversion domains with increasing free HCl, resulting in smoother and higher quality films.
Read the full abstract ▾
All authors K. Udwary, M. Smeaton, J.S. Mangum, R. Gannon, G. Dodson, B. Tellekamp, K.L. Schulte, J.H. Leach, J. Simon
Solar and Stellar Astrophysics
A homogeneous analysis of archival and new low-resolution spectra of YZ Cancri
Zhibin Dai, Lihuan Yu, Jiao Li et al. (+2 more)
YZ Cancri is an SU UMa-type
dwarf nova with sparse and heterogeneous optical spectroscopy. We present a homogeneous analysis of archival outburst-related spectra and seven new low-resolution quiescent spectra. We consistently characterize the Balmer, He I, Fe II, O I, and occasional He II and Bowen features, and assess which state-dependent differences are robustly supported. We applied a common continuum-normalization and line-measurement procedure, measuring equivalent widths, integrated line intensities, and selected profile parameters, and comparing line morphologies and intensity ratios between quiescent and outburst-related states. Because the spectra have modest and nonuniform resolving power and are not orbital-phase resolved, profile widths and centroid shifts are used only as relative diagnostics. Outburst-related spectra are dominated by broad Balmer absorption or composite profiles, whereas quiescent spectra show strong single-peaked Balmer and He I emission, recurrent Fe II {\lambda}5169, and weak O I {\lambda}7772 emission. He II {\lambda}4686 is detected in only two quiescent exposures, and the Bowen blend in one. The quiescent Balmer decrement is relatively flat, while the He I ratios show substantial exposure-to-exposure scatter. The measured Fe II ratios show no clear systematic departure from optically thin reference values, although blending limits their interpretation. These spectra establish an empirical state-dependent description of YZ Cnc and provide a reference
dataset for future phase-resolved, higher-resolution spectroscopy.
Read the full abstract ▾
All authors Zhibin Dai, Lihuan Yu, Jiao Li, Xuefei Chen, Zhanwen Han
Solar and Stellar Astrophysics
The emergence of a [WC] star in Sakurai's object
W. Marcolino, P. A. M. van Hoof, V. C. Maria et al. (+9 more)
Sakurai's object provides a rare opportunity to observe
stellar evolution on human timescales. Since its born-again event and detection in 1996, its
evolution has been extensively monitored, and recent optical spectroscopy has suggested the emergence of [WR]-type emission features. In this Letter, we present a secure spectroscopic identification of its central
star based on new VLT/FORS2 observations and NLTE expanding atmosphere models. Several observed emission lines arise in a [WR]-type
stellar wind, establishing the central
star as a [WR] object. The strongest
stellar features are due to C II-III and He I. We derive the
stellar and wind parameters of Sakurai's object and discuss its evolutionary status. The [WR]
star is much cooler (Teff ~ 30.5 kK) than the born-again
star V605 Aql (Teff} ~ 95 kK), which erupted about 80 years earlier. The inferred temperature is lower than predicted by models suppressing convective mixing efficiency during the VLTP, but is consistent with higher time-resolution calculations involving slightly lower-mass (<~ 0.6Msun) remnants. Continued spectroscopic monitoring and atmospheric modelling will be crucial for tracing its reheating curve and constraining the physics of this rapid evolutionary phase.
Read the full abstract ▾
All authors W. Marcolino, P. A. M. van Hoof, V. C. Maria, H. Todt, G. C. Van de Steene, J. A. Toalá, S. Kimeswenger, M. Hajduk, J.-C. Bouret, D. Tafoya, D. Barría, A. A. Zijlstra