Mesoscale and Nanoscale Physics
Inertial Dynamics of a Skymeron
Mona Bhukta, Duc Minh Tran, Kilian Leutner et al. (+19 more)
Topological
spin textures in antiferromagnets inherit the compensated
magnetic order of the host material, resulting for skyrmions in the suppression of the skyrmion
Hall effect and enabling ultrafast dynamics that make them attractive for low-power spintronic devices. In intrinsic antiferromagnets, the two sublattices are identical, but synthetic antiferromagnets offer additional control mechanisms by enabling independent tuning of the properties of two
ferromagnetic layers. Analogous to a ferrimagnet, the total
magnetic moment can be adjusted by compensating the two-layer moments. Here, we demonstrate that tuning the effective
magnetic anisotropy difference between the two layers drives a
spin-flop transition where one layer reorients and thus stabilizes an orthogonal configuration with one layer oriented along the out-of-plane direction and the other in-plane. In this transition, an
antiferromagnetic skyrmion undergoes a homotopic reconfiguration into a complex
spin texture comprising a skyrmion coupled to an in-plane bimeron. Element-specific X-ray microscopy resolves this texture layer by layer: one hosts an out-of-plane skyrmion coupled to an in-plane bimeron in the other. We refer to this previously unexplored three-dimensional
spin texture as a skymeron. By using time-resolved pump-probe X-ray microscopy, we discover unique polarity ,dependent dynamics during current pulses: a short post-pulse inertia-like continuation of the motion, and a slower return towards the pinned initial state. Micromagnetic simulations reveal that the inertia-like propagation originates from the finite-time relaxation of reorientation of the skymeron. Our results establish layer-selective anisotropy engineering as a route to uncharted composite
spin textures with internal dynamical degrees of freedom, not possible in conventional antiferromagnets with identical sublattices.
Read the full abstract ▾
All authors Mona Bhukta, Duc Minh Tran, Kilian Leutner, Takaaki Dohi, Nikolai S. Kiselev, Filipp N. Rybakov, Olle Eriksson, Fabian Kammerbauer, Sebastian Wintz, Markus Weigand, Hendrik Ohldag, Maria-Andromachi Syskaki, Robin Tietgen, Edoardo Mangini, Sabrina Kerber, Thibaud Denneulin, Joseph Vimal Vas, Rafal E. Dunin-Borkowski, Bastian Pfau, Oleg A. Tretiakov, Robert Frömter, Mathias Kläui
Rings and Algebras
Rational linear forms of linearizable ordinary differential equations
Dmitry Lyakhov
Every scalar ordinary differential equation of order at least three with rational right-hand side that is locally linearizable by a point transformation admits a linear form with rational coefficients over the same coefficient field. We prove this by restricting the derived
symmetry algebra to a coordinate line and recovering a scalar differential operator from rational
symmetry-jet data. The construction uses differential elimination and linear
algebra; it does not require the
symmetry generators or a linearizing transformation to be solved for. A single integer parameter suffices to choose the line.
Read the full abstract ▾
Strongly Correlated Electrons
Robustness of the Verwey transition against remanent strain-induced defects in magnetite
M. A. Gala, K. J. Grzywa, K. Komędera et al. (+16 more)
The Verwey transition in magnetite is a
benchmark electronically driven phase transition that is highly sensitive to
lattice imperfections and external perturbations. Doping, deviations from oxygen stoichiometry, and irradiation-induced point defects are known to lower the Verwey transition temperature, $T_{\rm V}$. By contrast, the role of remanent strain fields and extended defect structures generated by uniaxial stress has remained largely unexplored. Here we combine dark-field x-ray microscopy (DFXM) and ac
magnetic susceptibility to determine how strain-induced defects affect the Verwey transition in stoichiometric single-crystalline magnetite after uniaxial compression and unloading. The DFXM measurements were performed on two samples compressed along the same $[011]$ direction: loading to $200~\mathrm{MPa}$ generated stable line-like defects and remanent strain fields, whereas loading beyond the fracture threshold produced denser defect networks and mechanical fracture. To test the effect of a different loading geometry, we also measured the ac susceptibility of a sample fractured by compression along $[001]$. Real-space DFXM imaging,
lattice-orientation mapping, and residual strain mapping show substantial remanent structural disorder after compression. Nevertheless, neither the transition temperature $T_{\rm V}$ nor the sharpness of the transition changes measurably in any of the compressed or fractured samples. This behavior contrasts with the strong suppression of $T_{\rm V}$ by point-like disorder. Our results show that extended line-like defects and remanent strain fields produced by plastic deformation do not measurably perturb the macroscopic electronic ordering responsible for the Verwey transition, thereby distinguishing the effects of point defects from those of extended slip-related defects in magnetite.
Read the full abstract ▾
All authors M. A. Gala, K. J. Grzywa, K. Komędera, R. Zalecki, A. Baczmański, Z. Kąkol, A. Pacanowska, C. M. N. Kumar, R. Rodriguez Lamas, C. Detlefs, C. Yildirim, C. Marin, J. Debray, G. Beutier, L. Ortega, A. Kozłowski, N. Barišić, J. E. Lorenzo, W. Tabiś
Computational Geometry
Thomas Bläsius, Lennart Großkreutz and Jean-Pierre von der Heydt
The sphericity of a graph is the minimum dimension d such that the graph has an intersection representation of d-dimensional balls of equal radius. While sphericity has been studied in Euclidean space, we initiate the study of
hyperbolic sphericity. The
hyperbolic sphericity of a graph can be significantly smaller than its Euclidean counterpart, but, contrary to the Euclidean setting, depends strongly on the radius of the balls. We show that, if the radius of the balls can be chosen depending on the graph, the
hyperbolic sphericity is upper bounded by the Euclidean sphericity. This extends a previous result for 2-dimensional
hyperbolic space, i.e., uniform disk graphs, to arbitrary dimensions. Moreover, our proof is significantly simpler. If we fix the radius, i.e., do not make it dependent on the graph, we show that
hyperbolic sphericity can be larger than Euclidean sphericity, but by at most 1. Additionally, we study how
hyperbolic sphericity changes with the ball radius. We show that choosing a larger radius can substantially decrease the sphericity while increasing it by at most 1. We also provide a construction of a graph where the sphericity oscillates between different values as the radius increases. Besides being theoretically interesting, we note that these results are relevant for graph
embeddings in machine learning, where one is interested in low-dimensional numeric representations of symbolic data like graphs.
Read the full abstract ▾
Analysis of PDEs
The Allen-Cahn equation and general minimal cones
Oscar Ivan Agudelo Rico and Matteo Rizzi
We construct solutions to the Allen-Cahn equation $\Delta u+u-u^3=0$ in $\mathbb{R}^{N+1}$, with $N\ge 3$, whose zero level set is asymptotic (at infinity) to a given minimal cone. Our construction is quite general, since we only use a suitable nondegeneracy assumption; neither stability nor
symmetry of the underlying cone is required. Moreover, we calculate the morse index of such solutions. In particular, our main result generalizes a result by Pacard and Wei (2013) dealing with stable solutions to the Allen-Cahn equation in high dimensions. More precisely, we give a positive answer to a remark raised in the paper by Pacard and Wei about the possibility to relax the assumption for the underlying cone is area-minimizing or at least stable. As a corollary, we also obtain solutions with infinite morse index and connected zero level set.
Read the full abstract ▾
Symplectic Geometry
Hong-kwon Jo and Jungsoo Kang
Let $(M,\omega)$ be a closed monotone integral
symplectic manifold with monotonicity constant $\tau>0$, and let $D\subset M$ be a
symplectic hyperplane section of degree $0<k<\tau$. By Biran's decomposition
theorem, $M$ is the union of a
symplectic disk bundle over $D$ and the skeleton of the complement $M\setminus D$ equipped with its Liouville structure. In this paper, we compute the relative
symplectic cohomology of circle subbundles of this disk bundle over $\mathbb{Z}$ in terms of the
quantum cohomology of $D$. This relative
symplectic cohomology depends on the radius of the circle bundle relative to a certain critical radius. For a broad class of pairs $(M,D)$, we prove that the relative
symplectic cohomology over $\mathbb{Z}$ does not vanish for supercritical radii. Therefore, circle bundles of supercritical radii are heavy subsets, which yields infinite dimensional quasi-flats in the
Hamiltonian diffeomorphism group with respect to both the Hofer metric and the spectral metric over $\mathbb{Z}$. We also establish that if the degree $k$ of $D$ is greater than one, the skeleton of $M\setminus D$ is SH-full over $\mathbb{Z}/k\mathbb{Z}$, and hence non-displaceable in $M$ by any symplectomorphism.
Read the full abstract ▾
Theoretical Economics
Stochastic Choice, Limited Attention, and Aggregation over Attributes
Arkarup Basu Mallik, Mihir Bhattacharya and Anuj Bhowmik
We study an attribute-based model of
stochastic choice in which
attention to each attribute is limited, and characterize two choice rules within it. The Multiplicative
Attention Rule (MAR) rewards an alternative only for the attributes on which it ranks first, combining
attention to these multiplicatively across attributes; it captures choice deferral driven by the cognitive load of attending to `too many' attributes. The Additive
Attention Rule (AAR) instead averages
attention across attributes and credits an alternative by its full ordinal rank within the menu. This difference in aggregation is what separates the two rules behaviourally: AAR reproduces the Attraction Effect while holding the relevant attributes fixed across menus, something MAR can do only when the attribute set itself changes, though AAR, in turn, cannot accommodate the Compromise Effect. Both rules are characterized by axioms on observable choice data, and in both cases the underlying
attention parameters and attribute-based preferences are uniquely identified from that data.
Read the full abstract ▾
Quantum Physics
Achieving Identical Stored Energy in Cascaded Collisional Quantum Battery Charging: Analytical Result
Longyan Gong, Jing Zhang and Yongtao Li
Charging uniformity across distinct trials is an essential requirement for reliable
quantum batteries. Nevertheless, intrinsic
quantum stochasticity poses a fundamental challenge to this goal. This work investigates a two-
qubit quantum battery within the framework of cascaded collision models. We propose an adaptive measurement on the charger after each collision, yielding two conditional battery states of equal expected stored energy. For chargers prepared in the excited state, we analytically prove that complete suppression of trajectory-dependent stored-energy fluctuations is attainable; for chargers prepared in general superposition states, our theory reveals that full suppression is fundamentally prohibited owing to battery-phase-dependent recursive dynamics. Numerical simulations are in excellent quantitative agreement with theoretical predictions. These findings provide a feasible pathway toward high-uniformity
quantum-battery charging.
Read the full abstract ▾
Computer Vision
CLASP: Continual Low-rank Adapters for Spatially Placed Concepts from One Hypernetwork
Wojciech Gromski, Patryk Krukowski, Jan Miksa et al. (+2 more)
Continual personalization of text-to-image
diffusion models requires sequentially acquiring new concepts while retaining previously learned ones. However, existing methods either suffer from catastrophic forgetting or rely on storing additional concept-specific parameters and spatial components, causing their parameter footprint to grow with the concept stream. This limits their ability to scale to long sequences of personalization tasks. We propose a rehearsal-free approach that uses a single fixed-size hypernetwork to continually personalize a frozen
diffusion model. Instead of expanding the model as new concepts are acquired, the hypernetwork dynamically produces the concept-specific adaptations required for personalization while preserving previously learned concepts. Our framework further integrates spatial control into the personalization process, allowing users to specify where a personalized concept should appear without introducing additional per-concept components. This formulation enables continual personalization with a parameter footprint that remains independent of the number of learned concepts, aside from compact concept representations. Experiments demonstrate strong retention of previously learned concepts and reliable spatial grounding, matching or improving upon existing methods while scaling effectively to long streams of personalization tasks.
Read the full abstract ▾
All authors Wojciech Gromski, Patryk Krukowski, Jan Miksa, Maciej Zieba, Przemysław Spurek
Medical Physics
Quiet, rapid 3D multiparametric mapping using magnetization-prepared zero echo time MRI
Alireza Samadifardheris, Shishuai Wang, Ana Beatriz Solana et al. (+10 more)
Purpose: To introduce and evaluate MuPa-ZTE, a quiet, rapid 3D framework combining native and magnetization-prepared zero echo time (ZTE) acquisitions for multiparametric mapping. Methods: MuPa-ZTE combines steady-state native ZTE with transient-state magnetization-prepared ZTE. Two implementations were evaluated: T2T1-ZTE for apparent
proton density, T1, and T2 mapping, and T1-ZTE for apparent
proton density and T1 mapping. Both were assessed in an ISMRM/NIST system phantom and two healthy volunteers; T2T1-ZTE was also demonstrated in a patient with
brain metastases. The 10-minute phantom and 4.5-minute in vivo acquisitions were retrospectively truncated to 3, 2, and 1 minute and reconstructed with and without deep learning-based denoising. Evaluations included phantom agreement, precision, short-term repeatability, apparent SNR, edge sharpness, and consistency with full-duration in vivo maps. Results: T1 estimates remained close to nominal phantom values across implementations, durations, and reconstructions. T2 accuracy was maintained down to 2 minutes over the
brain-relevant range, with limited sensitivity to longer T2 values. Denoising generally reduced variability and improved short-term repeatability. In vivo, 1.1-mm isotropic whole-
brain maps were obtained in 4.5 minutes; denoising increased apparent SNR while preserving edge sharpness and yielded promising image quality after retrospective truncation to 2 minutes. Conclusion: MuPa-ZTE enables quiet, isotropic 3D multiparametric mapping within 4.5 minutes, supporting robust T1 mapping and T2 mapping over a
brain-relevant range, with promising acceleration toward 2 minutes using deep learning-based denoising.
Read the full abstract ▾
All authors Alireza Samadifardheris, Shishuai Wang, Ana Beatriz Solana, Jose de Arcos, Noemi Sgambelluri, Emil Langensee Ljungberg, Sagar Mandava, Mika Vogel, Steven C. R. Williams, Stefan Klein, Juan Antonio Hernandez-Tamames, Dirk H. J. Poot, Florian Wiesinger
Group Theory
Finite bases for full power semirings of finite nilpotent semigroups
Lili Wang, Qingrui Yin and Aifa Wang
We investigate finite equational bases for full power semirings of finite
semigroups, including the empty set, in the constant-free signature with addition and multiplication. For a nontrivial finite nilpotent
semigroup, we associate a finite relational structure recording the ordered products that are nonzero. We prove that the full power semiring is finitely based if and only if this structure has finite duality, and relate this condition to first-order definability and dismantling of the square of its core. Quantitative bounds connect obstruction size with the number of variables required in an identity basis. In the commutative case of nilpotency index $d$, the criterion reduces to the existence of an element with nonzero $(d-1)$st power. We also establish a nonfinite-basis obstruction for
semigroups with a two-element group ideal and prove a finite lifting
theorem. These results yield a direct-product criterion and a five-element counterexample to sufficiency of the identity-fibre condition. The arguments use equational logic, finite relational duality, and explicit algebraic constructions.
Read the full abstract ▾
Number Theory
Sharp $p$-adic Extrema and Least Extremal Rows for Restricted Binomial GCDs
John Fairfax-Ball
For integers $m\ge 2$ and rows $N>m$ divisible by $m$, consider the restricted binomial greatest common divisor $G(N;m)=\gcd\{\binom Nk:0<k<N,\ m\mid k\}$. Fix a
prime $p$ with $p\nmid m$, and let $r_p(m)$ be the least positive integer $r$ such that $m<p^r$. We prove that the largest possible value of $v_p(G(N;m))$, as $N$ ranges over all admissible rows, is exactly $r_p(m)$, and we give a constructive equality row. Attainment is established before the least extremal row $T_p(m)$ is defined. For the special family $m=p^a+1$ with $a\ge2$, we determine that least row exactly: $T_p(p^a+1)=p^{3a}+1$. The proof combines Kummer's carry
theorem with an explicit leading-digit witness for the universal upper bound, a multiplicative-order construction for equality, and a separate strict-minimality argument below $p^{3a}+1$ with a fallback witness for the unique worst leading-digit pattern. The selected-GCD family itself is known in the literature; the relation to earlier results and the limits of the documented literature search are stated explicitly.
Read the full abstract ▾
Quantum Physics
Preservability of Measurement Incompatibility: Purification, Activation, and a No-Go Theorem
Chao-Hsien Wu, Franco Nori and Huan-Yu Ku
Measurement incompatibility is a fundamental
quantum resource that enables advantages in many
quantum information tasks, including cryptography and communication. However, unavoidable interactions between a system and its environment can degrade or even completely destroy measurement incompatibility; such a process is referred to as a measurement-incompatibility-annihilating channel. Recently, the capability of noisy
quantum dynamics to preserve measurement incompatibility has been characterized within the resource theory of measurement incompatibility preservability. This motivates studying how to purify the preservation of measurement incompatibility and how to activate it from a measurement-incompatibility-annihilating channel. To this end, we first introduce our figures of merit as robustness-based resource monotones within this resource theory. We demonstrate that while pre-filtering operations can strengthen measurement incompatibility preservability, they cannot activate it from an incompatibility-annihilating channel, establishing a no-go
theorem. Furthermore, we explicitly demonstrate that an incompatibility-annihilating channel can be stochastically activated via post-filtering operations. Our results provide a practical framework for exploiting measurement incompatibility in
quantum information processing.
Read the full abstract ▾
Artificial Intelligence
An ontology for cross-sectoral crisis management: core and public health modules
Aldo Gangemi, Rita T. Sousa, Luigi Asprino et al. (+6 more)
This paper presents the European Crisis Management Ontology (ECMO), a modular OWL-based ontology intended as a cross-sectoral reference for disaster risk reduction and response. ECMO is designed to be organised as a network of ontological modules. Among the modules, ECMO-CORE captures fundamental crisis management concepts such as hazard, event, exposure, impact, and response measure and uses ontology design patterns and the OWL2 punning technique to resolve ambiguities between hazard types and event manifestations. In addition, domain-specific modules are defined as in the case of the public health module aligned with SNOMED CT and ICD-11. To demonstrate the resource's utility, we used ECMO to represent the data of the Epidemic Intelligence from Open Sources system of the Joint Research Centre to generate an end-to-end pipeline that populates an ECMO-compliant knowledge graph from unstructured epidemiological news. Initial results demonstrate that ECMO provides the formal guardrails necessary for consistent and unified knowledge representation and integration. The ontology is publicly available at https://doi.org/10.5281/zenodo.20070268 and is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.
Read the full abstract ▾
All authors Aldo Gangemi, Rita T. Sousa, Luigi Asprino, Giorgia Lodi, Andrea G. Nuzzolese, Valentina Presutti, Johannes Gysen, Diana F. Sousa, Luigi Spagnolo
Artificial Intelligence
PPO-HRAP: Proximal Policy Optimization with a Hybrid Regime-Aware Policy for Risk-Controlled Trading
Duong Hien Chi Kien and Thanh Trung Huynh
Reinforcement learning for trading often struggles to balance upside participation with drawdown control. Profit-only policies can collapse toward passive long exposure on upward-drifting assets, while aggressively risk-penalized rewards can become too defensive during volatile periods. This paper proposes PPO-HRAP, a hybrid regime-aware policy that combines Proximal Policy Optimization with an interpretable regime prior. The
agent observes both
market features and portfolio-state variables, receives a reward combining portfolio log return, VIX-conditioned drawdown-increase penalty, target-exposure deviation, and turnover cost, and executes a blended action between the PPO actor output and a regime-derived target exposure. On the held-out 2020-2022 SPY test window, PPO-HRAP achieves 27.62% total return, 8.48% annualized return, 0.6447 Sharpe ratio, 0.8588 Sortino ratio, and 0.4592 Calmar ratio, while reducing maximum drawdown from 34.10% for Buy and Hold to 18.47%. Across five SPY seeds, PPO-HRAP remains stable with mean total return $0.2725 \pm 0.0109$ and mean Sharpe ratio $0.6219 \pm 0.0565$. Single-run cross-asset tests on QQQ and DIA further show that the proposed method ranks first on total return and Sharpe ratio for all three reported assets. These results suggest that blending learned actions with a
volatility-aware regime prior is a practical way to improve risk-adjusted trading behavior, although the current policy still incurs high turnover and cross-asset robustness beyond SPY remains limited to single-run evidence.
Read the full abstract ▾
Computation and Language
Evaluating Biomedical Reranking for LLM-Based Question Answering over Longitudinal Clinical Notes
Maryam Shahbaz Ali, Laura B. Strachan, Caitlin Sherman et al. (+3 more)
Patient-specific clinical question answering requires locating the right evidence within long, heterogeneous longitudinal clinical records in which relevant facts may be scattered across encounters, repeated in copied-forward notes, or expressed using different clinical terminology. We evaluated whether biomedical reranking can improve evidence selection and downstream answer quality in a locally deployed
retrieval-augmented generation pipeline for longitudinal clinical notes. The pipeline combines PubMedBERT dense
retrieval, BM25 lexical
retrieval, weighted reciprocal-rank fusion, and MedCPT cross-encoder reranking. Across 1,000 open- and closed-ended question-answer pairs from a cohort of 200 bariatric surgery patients, reranking increased exact source-chunk
retrieval within the top 10 items, Hit@10 from 46.6% to 60.6% and mean reciprocal rank from 0.2371 to 0.3252. With Qwen3-8B generation, local judge-assessed answer correctness increased from 44.8% to 48.6%. These results show that biomedical reranking can improve the placement of relevant clinical evidence within a limited context window, although gains in
retrieval do not translate proportionally into gains in answer correctness.
Read the full abstract ▾
All authors Maryam Shahbaz Ali, Laura B. Strachan, Caitlin Sherman, Mark Kovler, Eleanor Mackey, Syed Muhammad Anwar
Artificial Intelligence
TRACE: Trajectory Return Attribution and Contrastive Erasure for Multi-Turn Safety
Fengpeng Li, Kemou Li, Qizhou Wang et al. (+3 more)
Safety-aligned large
language models (LLMs) often refuse a harmful request but comply once the same goal is spread over several turns. Preference objectives score whole responses to single
prompts, so their training loss alone cannot control risk on unseen histories. Our analysis gives sufficient conditions under which suppression at supervised single-turn contexts yields a bound on multi-turn trajectory risk. The bound accounts for coverage, transfer slack, and leakage, and characterizes contraction relative to a base-policy risk budget evaluated on the trained policy's contexts. TRACE (Trajectory Return Attribution and Contrastive Erasure) turns this principle into a
token-level objective. On the safe response, each
token is weighted by the discounted return of a refusal-attributable advantage. The advantage compares a frozen reference model with its refusal-ablated copy, allowing earlier response
tokens to receive credit from later refusal-related evidence. At high-gap positions on rejected responses, TRACE combines the observed
token with policy-selected alternatives in the erasure target. A gradient-norm penalty replaces the retain set. Across five open-weight models and seven multi-turn attacks, TRACE gives the lowest attack success rate (ASR) in all 35 model and attack pairs, while the model utility evaluated on MMLU and HellaSwag drop by at most 1\.23 points. Source code can be found in the supplemental material.
Read the full abstract ▾
All authors Fengpeng Li, Kemou Li, Qizhou Wang, Haiwei Wu, Jiantao Zhou, Di Wang
Machine Learning
Clifford Sheaf Neural Networks
Kotaro Kamiya and Joel Nicholls
We introduce the Clifford Sheaf Neural Network (CSNN), an equivariant sheaf neural network for
geometric graphs that places a Clifford
algebra on each stalk of a cellular sheaf and transports multivector features along edges. The canonical choice of restriction map for sheaves with
algebra-valued stalks is
algebra homomorphism. Adding the constraint of equivariance, the naive choice becomes versor conjugation. However, versor conjugation is expressively weak, so we drop
algebra homomorphism and arrive at the K-term sandwich. The resulting sheaf
Laplacian is positive semidefinite by construction, needs no versor constraint, and still mixes grades. Our main contribution characterizes the resulting family of restriction maps along three axes: which grades a map couples, how much of the endomorphism space it reaches, and how well it is conditioned. The K-term sandwich spans half of the endomorphism space, and in Cl(3, 0, 0) it corresponds to the maps that commute with the central pseudoscalar. The number of terms controls expressivity. CSNN is the reversion member, a first-order model by construction and the grade-mixing corner of this family, developed as a sheaf construction for graph-level equivariant regression.
Read the full abstract ▾
Optimization and Control
Control Allocation with Adaptive Augmentation for Aerodynamic Optimization of Trailing Edge Morphing Aircraft
Mark Spiller, Lennart Kracke and Johannes Autenrieb
This paper presents a control allocation framework with adaptive augmentation for a trailing edge morphing aircraft, providing stability guarantees under uncertainty while exploiting the available morphing degrees of freedom to optimize aerodynamic efficiency. A baseline controller is designed using the nominal aircraft model to establish the desired closed-loop reference dynamics. For the aerodynamics, the wing shape is optimized dependent on the flight state to achieve a target elliptical lift distribution corresponding to minimum induced drag. The adaptive augmentation is tailored to the control allocation problem to account for the uncertain system dynamics and stabilize the aircraft around the reference dynamics. The resulting stabilization condition is formulated as hard constraint in the allocation problem while minimizing deviations from the corresponding state-dependent aerodynamically optimal wing shape. Numerical results demonstrate that the adaptive augmentation compensates for matched uncertainties while the control allocation optimizes the wing shape with respect to the reference elliptical lift distribution.
Read the full abstract ▾
Artificial Intelligence
ProtoFlow: Prototype-Guided Flow Matching for Multivariate Time Series Forecasting
Shibo Feng, Wanjin Feng, Yang Qiu et al. (+3 more)
Generative modeling has shown strong promise for multivariate time mseries (MTS)
forecasting, especially scale to high-dimensional settings.
Diffusion-based methods achieve competitive performance but typically require many sampling steps at inference. VAE-based non-iterative
forecasting frameworks have therefore emerged as an efficient alternative. Within this line of work, vector quantization (VQ) enables controllable latent space modeling by mapping multivariate series into compact discrete representations. Existing VQ-based
forecasting methods, however, typically rely on autoregressive (AR)
token generation, which suffers from exposure bias and training-inference mismatch. Flow matching provides an efficient non-autoregressive alternative for latent
forecasting, but existing formulations usually initialize transport from a generic Gaussian prior. We instead observe that the trained VQ codebook already captures representative latent prototypes and can thus serve as a more informative prior for flow matching. Based on this insight, we propose ProtoFlow, a
forecasting framework that combines vector-quantized autoencoding with Prototype-prior Flow matching. Our method first maps multivariate sequences into a discrete latent space, then constructs a structured prior from the learned codebook, and finally learns a DiT-based rectified flow to transport samples from this prior to future latent representations conditioned on historical observations. By replacing generic noise initialization with a learned prototype prior, ProtoFlow avoids the rollout mismatch of AR
token prediction and promotes faster training convergence. Extensive experiments on
benchmark datasets show that it consistently achieves superior
forecasting performance with efficient inference.
Read the full abstract ▾
All authors Shibo Feng, Wanjin Feng, Yang Qiu, Deheng Ye, Peilin Zhao, Chunyan Miao
Robotics
Cross-entropy optimization with prioritized constraints
Francisco Roldan Sanchez, Pau de las Heras Molins, David Fridovich-Keil et al. (+1 more)
When constraints conflict, an optimizer must determine which requirements to preserve and which to relax. On the one hand, a priority ordering specifies which requirements take precedence. On the other hand, penalty-based formulations encode their relative importance through numerical weights. Depending on these weights, a solution can improve its weighted score while violating intended priorities. We introduce TierCEM, a variant of the cross-
entropy method that incorporates strict constraint priorities directly into elite selection without requiring per-constraint importance weights. TierCEM works by sequentially filtering sampled candidates, from highest- to lowest-priority constraint. If and when a constraint eliminates all remaining candidates, TierCEM returns to the last nonempty set and selects elites with the smallest violations of that blocking constraint, recursively preserving satisfaction of all higher-priority constraints. We evaluate TierCEM on 2D
navigation and contact-rich pushing tasks in proprioceptive and learned world-model settings. Experiments show that reversing the constraint ordering changes which constraints are violated under conflict. Prioritizing progress toward the task objective also enables TierCEM to relax lower-priority constraints when they would otherwise prevent further progress.
Read the full abstract ▾
All authors Francisco Roldan Sanchez, Pau de las Heras Molins, David Fridovich-Keil, Georgios Bakirtzis
Machine Learning
Feature Selective Model Collapse in Diffusion Models: Total Replacement versus Fixed-Budget Training
Hanna Malet and Gabriel Turinici
Model collapse arises when generative models are trained on synthetic data produced by earlier models. The phenomenon has attracted considerable
attention because of its societal and technical implications. However, previous studies have reached seemingly contradictory conclusions: replacing real data with synthetic data causes collapse (Shumailov et al.), yet accumulating real data alongside synthetic data can prevent it. For
diffusion models, we study an intermediate regime typical of finite-budget pipelines: all past
datasets and the real data are kept, but each new model is trained on a fixed-size sample from this growing pool, so the real fraction vanishes without any data being removed. Experiments on a 2D spiral
dataset as well as the image
benchmarks (MNIST, Fashion-MNIST, and CIFAR-10) show that replacement protocol degrades
dataset rapidly as in the literature, whereas the fixed budget degrades only partially, sparing some features. A linear-response model of the multi-generational parameter dynamics, analyzed by
stochastic recursion, confirms that the two protocols differ: some features will be fragile and lost within a few generations for both protocols, while some will be robust and preserved over practically unbounded horizons under the fixed budget protocol.
Read the full abstract ▾
Machine Learning
Prediction-powered Neural Architecture Search
Pascal Janetzky, Yuxin Wang, Michael Klar et al. (+1 more)
Evaluating candidate architectures in neural architecture search (NAS) faces an inherent trade-off: on the one hand, reliable performance labels are limited because training and evaluating architectures is costly; on the other hand, zero-cost proxies (ZCPs) are cheap to compute at large scale but can be noisy. Yet, how to effectively combine these two sources of supervision remains unclear. In this paper, we propose PPNAS, a novel prediction-powered inference (PPI) approach for NAS. PPNAS fuses (1) a small set of architectures with observed performance labels and (2) a large set of architectures with ZCP information. To combine these two sources of supervision, PPNAS exploits the ordinal information provided by ZCPs to construct additional pairwise ranking supervision, while PPI debiases systematic discrepancies between ZCP-based and true performance rankings. We evaluate PPNAS in end-to-end predictor-based NAS, where it achieves state-of-the-art under limited evaluation budgets. To the best of our knowledge, PPNAS is the first prediction-powered approach for label-efficient NAS.
Read the full abstract ▾
All authors Pascal Janetzky, Yuxin Wang, Michael Klar, Stefan Feuerriegel
Computation and Language
What Wins a Vote? Formatting, Length, and Lexical Diversity in the French Compar:IA LLM Arena
Simonas Zilinskas, Maayeesha Farzana and Christophe Benavent
LLM arenas turn pairwise human preferences into model rankings. Those preferences may reflect how an answer is presented as well as what it says. We take a stylometric approach to 137,293 decisive French-
language votes from the July 2026 Compar:IA release; the primary formatting analysis includes 137,113 battles across 116 models, and the joint estimates use the 127,092 battles with all required measurements. For each battle, we reconstruct the response visible when the user voted. We then compare the raw ranking with rankings adjusted for formatting, length, readability, vocabulary variety, and sentence structure. Presentation is associated with winning, but length, bold text, and lists tend to occur together, making their individual contributions hard to separate. Across the measured features, two associations change least across specifications: bold usage (+11.0% win odds per standard deviation in the joint model) and moving-average type-
token ratio (MATTR), a measure of vocabulary variety that is less sensitive to answer length (+16.8%). The bold association is substantially smaller in observed multi-turn conversations, whereas the MATTR association changes little; because users choose whether to continue, this difference is descriptive rather than causal. The full adjustment moves 36 of 116 models by at least ten ranks. Yet comparisons with external
benchmarks do not show that adjusted rankings better measure capability. We therefore recommend publishing raw and adjusted rankings side by side as a transparent sensitivity analysis.
Read the full abstract ▾
Machine Learning
EP-Flow: Disordered Crystal Structure Prediction without Site-Level Annotations
Qiuliang Liu, Liming Wu, Qi Li et al. (+5 more)
Generative models have made rapid progress in ordered
crystal structure prediction, yet many functional materials are intrinsically disordered, with substitutional mixing, vacancies, or interstitial species controlling their properties. Existing
crystal generators either assume deterministic site occupations or require site-level disorder annotations, which are often unavailable when the chemical formula is the primary input. We formulate disordered
crystal structure prediction through an Occupancy Distribution
Matrix (ODM), a continuous site-by-species representation that unifies ordered
crystals, solid solutions, vacancy disorder, and interstitial occupancy. A valid ODM must satisfy coupled site-wise occupancy, mass-conservation, and non-negativity constraints, placing each sample on a formula-dependent transportation polytope. We propose Entropic Polytope Flow (EP-Flow), a marginal-constrained flow matching framework that canonicalizes heterogeneous polytopes into a shared double-centered space, learns a marginal-preserving flow, and recovers feasible occupancies through a Sinkhorn inverse map. By jointly generating occupancies, fractional coordinates, and
lattice parameters, EP-Flow achieves state-of-the-art performance on formula-conditioned disordered CSP
benchmarks derived from COD and MPDS, substantially outperforming adapted ordered-
crystal generators. Analyses further show that EP-Flow recovers sparse and chemically meaningful local disorder patterns rather than merely matching global composition statistics.
Read the full abstract ▾
All authors Qiuliang Liu, Liming Wu, Qi Li, Zhonglong Peng, Chang Chen, Xiaolong Chen, Wenbing Huang, Shifeng Jin
Computer Vision
ARROW: Arbitrary Reconstruction and Tracking of 4D Observations in the Wild
Ilya Fradlin, Christian Schmidt, Jens Piekenbrinck et al. (+3 more)
Dynamic scenes may be captured by a moving camera, multiple
video streams, or images taken at different times. These observations reveal complementary aspects of scene geometry and motion, yet bringing them together requires establishing correspondence across viewpoints, capture times, and visibility changes. We introduce ARROW, a feed-forward model that unifies 3D reconstruction and 3D point tracking from arbitrary image sets. At its core is a novel order-invariant querying approach, which allows the association of queries with observations across arbitrary inputs. We show that exposing the model to more diverse sets of inputs during training results in improved task performance. Moreover, the resulting model is capable of generalization to a wider range of tasks including multi-view tracking. Trained with this strategy, ARROW establishes a new state of the art in 3D tracking on WorldTrack and TAPVid-3D and outperforms dedicated multi-view trackers on an adapted RGB-only MVTracker
benchmark, while remaining competitive across 3D reconstruction tasks. Code and weights are publicly available.
Read the full abstract ▾
All authors Ilya Fradlin, Christian Schmidt, Jens Piekenbrinck, Karim Knaebel, Gonzalo Martin Garcia, Bastian Leibe
Statistical Methodology
Testing and segmentation of joint and individual components in integrative multi-source factor models
Kyoowon Kim and Sungkyu Jung
Disentangling shared (joint) structures from source-specific (individual) variations is a fundamental task in multi-source data integration. Existing joint-individual models often rely on computationally intensive optimization or loose spectral bounds, leading to suboptimal separation accuracy and poor scalability. In this paper, we propose the Multi-Source Sequential
Alignment Test (MSSAT). MSSAT leverages the
geometric observation that true joint components manifest as closely aligned score subspaces across different data sources. By deriving the asymptotic null distribution of our
alignment statistic, we develop a rigorous, resampling-free sequential testing procedure to accurately estimate the joint rank. Extensive simulations and real data applications, including a TCGA multi-omics
dataset, demonstrate that MSSAT achieves superior separation accuracy and substantially faster computation compared to competing methods.
Read the full abstract ▾
Group Theory
Finite abelian groups are Schur multipliers of finite solvable groups
Derek F. Holt
We prove that every finite abelian group arises as the Schur multiplier of a finite solvable group, from which it can be deduced that every finite abelian group is the
knot group of some extension of number fields.
Data Structures and Algorithms
Factor Three Approximation for Edit Distance
Egor Gorbachev
We give randomized
algorithms for $3$-approximate edit distance in $\widetilde{\mathcal{O}}(N^{11/6})$ time for unweighted edit distance and in $\widetilde{\mathcal{O}}(N^{40/21})$ time for arbitrary metric edit weights, where $N$ is the total input length. For non-metric costs, we prove an unconditional $\Omega(N^2)$ oracle-query lower bound for every approximation factor depending only on $N$, even for symmetric weights or weights satisfying the triangle inequality (but not both). Under the Orthogonal Vectors Hypothesis, we show a similar result for constant-size alphabets. This holds even for symmetric weights over a size-$3$ alphabet or triangle-inequality weights over a size-$2$ alphabet. In contrast, for symmetric weights over a binary alphabet we show an $\widetilde{\mathcal{O}}(N^{40/21})$-time $3$-approximation
algorithm.
Read the full abstract ▾
Robotics
Ex vivo breach detection using electrical conductivity during robotic pedicle drilling in the spine
Jorge Andrés Pérez Velásquez, François Teyssere, Thibault Chandanson et al. (+2 more)
Purpose: Pedicle screw placement is technically demanding in scoliosis treatment. High precision is required due to limited visibility, anatomical variability, and the risk of complications. Although robotic systems assist CT-based planning and execution, they still rely on ionizing intraoperative imaging and complex registration. This study proposes robotic pedicle drilling with real-time preventive breach detection using electrical bioimpedance sensing. Methods: We developed a robotic approach combined with a pedicle-drilling tool equipped with a proprioceptive electrical bioimpedance sensor developed by SpineGuard. A real-time detection
algorithm was designed to analyze the electrical bioimpedance signal during drilling and identify abrupt changes in conductivity associated with potential breaches towards the spinal canal. The method operates without external devices or sensors. Results: The ex vivo experiments showed that the proposed method prevented breaches in 100 of the 51 drilling cases. These findings demonstrate the system's ability to detect potentially hazardous events during drilling and to stop the procedure before. The ex vivo experiments demonstrated that the proposed method prevented breaches in all 51 drilling cases. Conclusions: This work demonstrates the feasibility of robotic pedicle drilling with electrical bioimpedance sensing for real-time breach prevention. Using only the tool signal, the method eliminates the need for external sensing systems and supports safer pedicle screw placement.
Read the full abstract ▾
All authors Jorge Andrés Pérez Velásquez, François Teyssere, Thibault Chandanson, Quentin Grimal, Brahim Tamadazte