Peer Revenue Play credits. Real papers.

Papers

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

All of arXiv

Thursday 10 September

Analysis of PDEs

Long-time dynamics of partially dissipative hyperbolic systems with non-autonomous coefficients

We study quasilinear symmetrizable partially dissipative hyperbolic systems with non-autonomous relaxation coefficients in $\mathbb{R}^d$ ($d\geq1$). The existence of global strong solutions is established in a critical regularity setting for systems satisfying the so-called Shizuta-Kawashima (SK) and entropy conditions. When the initial data are additionally bounded in a lower-regularity norm, we prove that the corresponding solutions converge to equilibrium at optimal algebraic decay rates. Furthermore, we show that the conservative part of the solution behaves asymptotically as the solution of a non-autonomous parabolic equation. Our results apply to the compressible Euler system with the time-dependent damping coefficient $\frac{K}{(1+t)^{\alpha}}$ ($\alpha<1$, $K>0$ or $\alpha=1$, $K\gg 1$) in the velocity equation. The natural low/high-frequency splitting of the autonomous theory persists in the non-autonomous setting, but with a frequency-threshold that evolves in time. To handle this moving frequency structure, we introduce a new class of hybrid Besov spaces adapted to time-dependent thresholds and derive hypocoercive estimates in each frequency regime. Our results reveal the qualitative and quantitative effects of general time-dependent relaxation coefficients on dissipation and large-time dynamics.
Read the full abstract

Astrophysics of Galaxies

Constraints on the fuzzy dark matter mass using globular clusters in dwarf galaxies from Euclid ERO data

We constrain the particle mass of the fuzzy dark matter model using globular cluster candidates in dwarf galaxies, from the Euclid Early Release Observations, within the Fornax and Perseus galaxy clusters. We model the orbital evolution of globular clusters by combining fuzzy dark matter density profiles with the corresponding dynamical friction formalism, and integrate their orbits using galpy. In the fuzzy dark matter framework, both the halo density profile and the dynamical friction force are modified by the wave-like nature of the ultra-light dark matter particle, whose mass is parameterized as $m_{22}=m_\chi/10^{-22},\mathrm{eV}$ and explored over the range $0.1 \leq m_{22} \leq 100$. This framework naturally alleviates the long-standing globular cluster timing problem originally identified in the Fornax dSph. We find that globular clusters no longer experience orbital decay toward the centers of their host galaxies for $m_{22}\leq1.46^{+0.51}_{-0.32}$ in the Fornax cluster and $m_{22}\leq1.09^{+0.14}_{-0.21}$ in the Perseus cluster. Future Euclid data releases, providing substantially larger samples of globular cluster candidates, will enable significantly tighter constraints on the dark matter particle mass within the fuzzy dark matter scenario.
Read the full abstract

All authors Lisa Besnard, Pierre Boldrini, Paola Di Matteo, Chervin Laporte, Teymoor Saifollahi

Quantum Physics

Virtual quantum neural networks

Quantum neural networks are a prominent model of quantum machine learning. Their training consists in the minimization of a given loss function over a parametrized family of quantum circuits, mathematically described by unitary operators, or, more generally, completely positive linear maps. In this work, we extend the notion of quantum neural network, using random sampling and classical data processing to enlarge the optimization space in a way that includes linear combinations of completely positive maps. Our extended model, called virtual quantum neural networks, leverages its enlarged optimization space to achieve increased expressivity and improved noise robustness. These benefits are illustrated in three representative tasks: quantum error mitigation, binary classification, and estimation of ground-state energies. Overall, virtual quantum neural networks offer a flexible learning paradigm that expands the space of achievable computations and strengthens the applications of near-term quantum hardware.
Read the full abstract

All authors Benchi Zhao, Xuanqiang Zhao, Yinan Li, Yingzhou Li, Giulio Chiribella

Logic in Computer Science

Rice's Theorem under Self-Modification: Elevation Operators and a Normal Form

We ask whether it can be certified algorithmically that a self-modifying program keeps a behavioural property, a safety property in the motivating case, after its next rewrite (preservation) and along its whole evolution (persistence). When the rewrite depends only on behaviour, preservation is a behavioural property and Rice's theorem applies. When the rewrite reads the code, preservation is no longer behavioural; yet, under a uniform disruption condition, the s-m-n reduction that proves Rice's theorem works inside a single class of behaviourally identical programs, and preservation inherits the degree of the halting problem. One step never exceeds the degree of the property, while persistence can climb one level of the arithmetical hierarchy. We then isolate the mechanism shared by rewriting, supervision and system comparison, the elevation operator, and prove a normal form: the preserving set is determined by a single finite trigger and a polarity, and the Rice-Shapiro theorem restricts the polarity to the arithmetical class of the property. Runtime monitors, consistency supervision, conformance to a reference and observational equivalence are instances, and no sound theory covers the preserving systems.
Read the full abstract

2609.11326 · 10 Sep · v3: journal version. Shortened; neutral terminology; new Proposition 7.12 showing that the class of elevation operators is complete for anchored normal forms; comparison with enforcement by program rewriting (Hamlen, Morrisett and Schneider) added; illustrations moved to an appendix. 35 pages. Companion paper: arXiv:2606.28639 (applied consequences)

Neurons and Cognition

Degeneracy along the sensorimotor hierarchy: motor control within a framework larger than redundancy

Motor control has described the surplus of solutions available to the nervous system as redundancy, a term that names duplication: interchangeable elements, robust to loss but incapable of differential adaptation. Biology has had a second term for twenty-five years. Degeneracy names elements that are not interchangeable and are nonetheless isofunctional with respect to a given output, and it supports adaptability, since non-identical elements necessarily diverge in some context. Circuit neuroscience has relabeled its own results accordingly, while motor control has kept the older vocabulary. Neuromechanical models, by placing a spinal circuit in the loop with a musculoskeletal apparatus, bring the two traditions onto the same class of objects. We restate the Edelman and Tononi distinction for sensorimotor systems and derive an operational requirement: not the existence of multiple solutions, but their divergence in contexts they were not selected for. Three influential studies each meet part of that requirement and none meets all. We then argue that degeneracy and redundancy coexist along the sensorimotor hierarchy in a proportion that varies continuously, and that this proportion is measurable: computing degeneracy twice for the same configuration, once with muscle activation as the output and once with the movement produced, isolates what the musculoskeletal apparatus contributes. Five predictions follow, with the single outcome that would refute the proposal. We set out the adaptations the measurement requires in a nonlinear, non-stationary, closed-loop system, and what changes for motor control once solutions are no longer assumed equivalent: the question shifts from which rule selects a command to what the repertoire of the system still allows.
Read the full abstract

Analysis of PDEs

Regularity of a free boundary determined by the value of the gradient. Part 1: Third order asymptotics

This is the first part of a project concerning the regularity of the free boundary of a function, where the free boundary depends on the gradient of the function. We study the minimizer of the expression \begin{equation*} J(u) : = \int_{B_1} F(\nabla u) \ dx, \end{equation*} where $F(x)$ is a uniformly convex function whose second derivatives might jump at $|x|=1$. This results in an Euler-Lagrange equation that varies over the free boundary, which we define as $$ \Gamma = \partial \{x \in B_1: |\nabla u| >1 \} \cap \partial \{x \in B_1: |\nabla u| <1 \}. $$ We consider two-phase flat points for which the second derivatives of $F$ jump over $\Gamma$. In this paper, we show that under some regularity and non-degeneracy assumptions, a minimizer $u$ can be expressed as $$ u(x)= a + \nu \cdot x+ \delta p(x) + \delta \epsilon q(x), $$ where $a \in \mathbb{R}$, $\nu \in \mathbb{R}^n$, $0 < \delta, \epsilon << 1$. Here $p$ is a $C^1$ function, which consists of one polynomial in the upper half ball and another polynomial in the lower half ball. The function $q$ is a rest term with bounded $L^2$ norm. Furthermore, assuming that we have a sequence, $u^j$, of minimizers, we show that $q^j$ converges weakly to a function $q^0$ that satisfies a certain PDE. In addition, we show that $q^0$ is $C^3$ in $B_r^+$ and $B_r^-$, respectively, for $0<r<1$. This is the main result of this paper which is intended to be used to show regularity of the free boundary.
Read the full abstract

Combinatorics

Polynomial positivity cones for Coxeter roots and walks in trees

For a finite simple graph $G$ and an integer $k\ge0$, let $w_k(G)$ denote the number of walks of length $k$. We prove the conjecture of Täubig, Weihmann, Kosub, Hemmecke, and Mayr for every finite tree and determine all equality cases. If $T$ has $n\ge1$ vertices, then $n w_{k+1}(T)-2(n-1)w_k(T)\ge0$ for every $k\ge1$; for $n\ge3$, equality holds if and only if $T$ is a star and $k$ is even, whereas for $n=1$ or $n=2$, equality holds for every $k\ge1$. For non-Dynkin trees and even indices, the proof is based on a polynomial positivity cone associated with the adjacency operator of a finite graph and a positive real root of its simply-laced Coxeter system. For finite connected bipartite non-Dynkin graphs, we establish sufficient positivity conditions in terms of Coxeter orbits and inversion sets, and verify these conditions for indicator roots supported on connected induced subtrees. For non-Dynkin trees, this yields the rooted even-index inequality and, after summation, the corresponding global inequality. We also prove that if $G$ is a finite connected bipartite non-Dynkin simple graph, $\varnothing\ne U\subseteq V(G)$, and the subgraph of $G$ induced by $U$ is a tree, then $|U|w_{k+1}(G,U)-2(|U|-1)w_k(G,U)\ge0$ for every $k\ge0$, where $w_k(G,U)$ counts the length-$k$ walks in $G$ whose initial and terminal vertices lie in $U$; the intermediate vertices are unrestricted. The remaining even-index cases for finite Dynkin trees are handled by generating-function recurrences, while the odd-index cases follow from a spectral covariance identity.
Read the full abstract

All authors Dongxiu Cai, Zhenbo Chen, Jiasheng Zeng, Xiao-Dong Zhang

Computation and Language

MultiHuSE: A Multimodal Dataset for Humour Styles and Emotions

Computational recognition of verbal humour remains a challenging task, requiring an understanding of language, delivery style, emotions, and cultural context. Most existing approaches focus on binary classification and lack datasets that capture psychological dimensions of humour alongside variations in expression. We introduce MultiHuSE, a multimodal dataset comprising 2,407 high-definition videos of 50 demographically diverse actors performing 1,463 text samples across four psychological humour styles (affiliative, aggressive, self-enhancing, and self-deprecating), as well as neutral content. A subset is additionally annotated for underlying emotions. The dataset uniquely captures multiple actor interpretations of the same texts, enabling systematic analysis of expressive diversity. Baseline experiments show that multimodal fusion outperforms unimodal approaches (80.1% vs. 77.4% accuracy) in humour style classification, with particularly strong gains for affiliative humour (66% to 74%). While text provides the strongest individual signal, fusion models deliver meaningful improvements. We hope that MultiHuSE provides empirical support for psychological theories linking humour and emotion, while also opening new avenues for research in human communication, well-being, and AI-driven interaction. The dataset is available for academic use under an End-User Licence Agreement.
Read the full abstract

Artificial Intelligence

AI Exposure and AI Resilience: A Two-Dimensional Assessment Framework for Software and Software-Based Business Model

Artificial intelligence is changing both software production and the economics of software-based business models. Classical technology due diligence mainly examines technical properties such as architecture, scalability, and technical debt. These criteria do not fully capture how AI can affect a company's value proposition, competitive position, margins, or access to customers. This paper develops Artificial Intelligence Exposure and Resilience (AI-ER) as a two-dimensional assessment framework. AI exposure describes the pressure for change that AI creates for a business model. AI resilience describes the company's ability to absorb that pressure, adapt to changed conditions, and use AI in an economically viable way. Metrics for both dimensions are derived from current AI capabilities, their deployment conditions, and relevant research on business models and organizational adaptability. The model keeps exposure and resilience separate and adds an explicit assessment of evidence quality and confidence. It can be applied first with public information and later refined with internal evidence. The result is a traceable company profile that supports comparison without concealing uncertainty in the underlying evidence. The paper also specifies an initial score logic and a procedure for empirical validation.
Read the full abstract

Computers and Society

A global mobile network coverage raster product at 1km resolution, 1999–2030

Where a mobile signal is available shapes who can work, learn, bank, seek health care and respond to crises in the digital age, yet no globally consistent, sub-national record of mobile network coverage exists. We present such a record: annual 1km maps of the probability of 2G, 3G and 4G coverage for 214 countries and territories for the years 1999 to 2030. The maps are produced by three independent models: a calibrated machine-learning model, a techno-economic simulator of network build-out, and a spatial deep-learning model. The three estimates are then combined, per country and technology and in proportion to their measured accuracy, into a single best estimate with per-pixel 90% uncertainty bands; all four layers are released as part of the dataset. Because mobile roll-out closely follows a country's socio-economic conditions (population distribution, electrification, physical infrastructure), the models are grounded in existing geospatial data and tuned on 2,409 quality-screened operator-reported coverage maps, which are available up to 2020. For 2021–2024 the maps are predicted from recent geospatial data alone; for 2025–2030 they are extrapolated from demographic and infrastructure projections. On countries held out during training, the machine-learning model attains AUC 0.89–0.92. Baseline comparisons and the combined product's external validation are reported in Technical Validation. The dataset supports mapping the global digital divide, linking connectivity to household-survey outcomes, and humanitarian and infrastructure planning.
Read the full abstract

All authors Till Koebe, Theophilus Aidoo, Ali El Chami, Ali Kanso, Akansh Maurya, Purushottam Sharma, Ingmar Weber, Ridhi Kashyap

Artificial Intelligence

Magenta: Closing the Loop Between Mathematical Reasoning and Lean Verification

Most of mathematical knowledge has been communicated through so-called informal use of mathematics and natural language. With large language models (LLMs) being highly adept in using natural language, they achieve strong performance, yet not perfect, in informal mathematical reasoning. Restraining LLMs to informal reasoning misses out on the opportunity to use the discrete verification abilities that machines offer through machine-checkable proofs. In this paper, we bridge the gap between informal and formal reasoning by integrating Lean signals into the informal reasoning process. We introduce Magenta, a training-free agentic pipeline that, given only a natural-language problem, produces an answer, expresses it as a Lean 4 statement, and constructs a machine-checked proof. A statement judge verifies whether the formalisation preserves the original problem, while an error-attribution judge routes failed attempts either to mathematical re-derivation or local Lean repair. Magenta achieves 100% accuracy across all evaluated olympiad benchmarks, including AIME 2025, AIME 2026, and HMMT February 2026. When paired with the open-weight K2-Horizon-7B reasoner, it solves all six IMO 2026 problems. Our analysis shows that statement adjudication is essential for preventing false certificates and that feedback-guided correction outperforms independent resampling on difficult problems.
Read the full abstract

All authors Joshua Ong Jun Leang, Haonan Li, Zheng Zhao, Xinyi Shang, Wenda Li, Zhengzhong Liu, Eric Xing, Shay Cohen, Eleonora Giunchiglia

Artificial Intelligence

Mr.LHDR: A Benchmark for Multimodal Real-World Long-Horizon Deep Research Agents

Deep research agents are increasingly capable of web search, tool use, multimodal evidence analysis, and information synthesis. However, existing benchmarks mainly evaluate medium-horizon exploration and rarely test whether agents can sustain long, dependency-heavy research processes. We introduce Mr. LHDR (Multimodal real-world Long-Horizon Deep Research), a benchmark for evaluating real-world deep research over long, irreducible chains of interdependent evidence across eight categories. Each question is constructed from a hidden Node-Relation graph and requires an average of 12.1 necessary intermediate conclusions with a mean dependency depth of 10.4 before reaching a short, unique, and verifiable answer. Questions incorporate multimodal evidence, including images, maps, PDFs, logos, charts, tables, and video frames, with at least one non-text element that changes the reasoning state. Mr. LHDR evaluates both final answers and the correctness of intermediate conclusions under annotated dependencies. We evaluate general models, deep research systems, and agent frameworks using Overall Accuracy (OA), Strict Accuracy (SA), Checklist Score (CS), and Dependency-Aware Checklist Score (DACS). Results show that even the strongest system achieves only 43.1% OA and 34.3% SA, indicating that final-answer accuracy substantially overestimates complete research success. Removing images reduces DACS by 12.6 points, demonstrating the importance of multimodal evidence, while SA consistently declines as reasoning chains become longer. These findings reveal sustained, dependency-consistent evidence integration, rather than isolated fact retrieval, as a key bottleneck for current deep research agents.
Read the full abstract

All authors Minghao Guo, Meng Cao, Sui Zhao, Siyu Ning, Xin Wang, Haoze Zhao, Jiaxuan Yang, Haihong Hao, Mingfei Han, Shunlin Rong, Haijun Wu, Xiaodan Liang, Xiaojun Chang

Computer Vision

Mi-Ripple: Restoring Images Degraded by Iterative AI Editing

Iterative reference-conditioned image editing can introduce grid-like and granular textures, commonly described as digital ripple. We present Mi-Ripple, a diagnosis-guided restoration workflow that suppresses this digital ripple while protecting image structure. Mi-Ripple separates periodic lattice artifacts from content-entangled granular texture, then combines selective spectral notching, structure-aware smoothing, and cleaned-reference regeneration. This separation enables low-distortion filtering when artifacts are spectrally isolated and visual reconstruction when filtering would erase legitimate detail. Across fourteen notch-only executions, whole-image residual standard deviation is 0.08–0.44 in CIELAB lightness units. In a paired regeneration example, reference cleaning reduces output debris density by 45%. Mi-Ripple links measurable artifact reduction to visibly cleaner generated images, rather than optimizing a spectral score alone.
Read the full abstract

Operator Algebras

Continuous family of compact quantum metric space structures from cocycle twisted crossed product $\textrm{C}^{\ast}$-algebras

We establish the existence of a three-parameter family of compact quantum metric space structures on cocycle twisted crossed products by discrete groups. We are mainly interested in the case where the acting group has exponential/subexponential growth. We prove that the family is jointly continuous with respect to the parameters when the acting group is exact. We obtain quantitative upper and lower bounds for the associated metric dimensions. In particular, the bounds are helpful to prove the failure of lower semicontinuity of the metric dimension with respect to the quantum Gromov-Hausdorff distance. We also prove invariance of metric dimension under zero quantum Gromov-Hausdorff distance.
Read the full abstract

Artificial Intelligence

Routing by Reasoning Need: Trajectory-Aware Decoding Control for Diffusion Vision-Language Models

Diffusion vision-language models generate answers through iterative refinement, exposing intermediate answer trajectories that can be inspected and controlled at inference time. However, this controllability creates a reasoning-need mismatch, where a universal generation length is applied to questions with different reasoning demands. Visually closed questions may be harmed by continued refinement after a stable answer has formed, whereas reasoning-sensitive questions may be harmed by premature commitment. We formulate this problem as reasoning-budget mismatch and study it in LLaDA-V. Rather than choosing a universal generation length, our training-free controller routes each example to early commitment, baseline preservation, or reasoning-supportive decoding using trajectory signals from answer closure, commitment evidence, and representation revision pressure, without using ground-truth answers. Across answer-focused, mixed-reasoning, and CoT-sensitive benchmarks, routed control improves robustness over fixed long decoding, pure short decoding, and single-rule interventions. The gains are not explained by shorter outputs alone. Answer-closed examples often benefit from commitment, whereas CoT-sensitive examples require preserving or supporting intermediate reasoning. Taken together, these results suggest diffusion VLM decoding should route inference-time control by the state suggested by the observed trajectory instead of relying on a universal decoding length.
Read the full abstract

All authors Yixiang Liu, Zhongxing Xu, Zhonghua Wang, Xiaoying Tang

Machine Learning

A Dynamic Fusion Large Language Model for Traffic Flow Prediction

Traffic flow prediction is a core supporting technology for intelligent transportation systems. It uses historical data to infer future traffic dynamics in specific areas, thereby helping to alleviate congestion and improve resource allocation efficiency. Traditional neural networks struggle to break through accuracy limits due to their reliance on singular feature modeling, while large language models (LLMs) suffer from insufficient capture of spatial topological information and mining spatiotemporal correlation. This study proposes a Dynamic Fusion Large Language Model (DF-LLM) for traffic flow prediction. The model incorporates three core components: spatiotemporal embedding module, spatiotemporal fusion module, and LLM backbone. The spatiotemporal embedding module enables synergistic representation of multi-scale spatiotemporal features. The spatiotemporal fusion module integrates spatial topology and dynamic dependencies via graph convolution. The LLM backbone adopts a differentiated parameter adaptation strategy to balance training efficiency and traffic data adaptability. Additionally, it introduces a context aggregation attention module to strengthens global dependencies. More importantly, the LLM backbone takes the residual connections to mitigate the gradient vanishing in deep networks. Experiments show that DF-LLM has achieved better performance by comparing the metrics on all the four datasets.
Read the full abstract

Astrophysics of Galaxies

No Detectable One-halo Galactic Conformity Signal with Halo-mass Estimates Consistent with Weak-lensing Constraints

One-halo galactic conformity is the tendency for satellites in halos with quenched centrals to have lower star-formation activity than those in halos with star-forming centrals at fixed halo mass. It is an important probe of the galaxy–halo connection and halo-wide quenching processes that may couple central and satellite evolution. However, its existence remains controversial, because conformity must be measured at fixed halo mass, while halo masses are difficult to estimate accurately. In this Letter, we measure one-halo conformity in SDSS using five stellar-mass-complete samples and three halo-mass estimates: an ML estimate whose star-forming and quenched stellar mass–halo mass relations (SHMRs) agree with independent weak-lensing constraints, and two conventional abundance-matching (AM) estimates. We quantify conformity as the difference in median $\log({\rm sSFR})$ between satellites of star-forming and quenched centrals, using both satellite-level and halo-level statistics. The two AM estimates produce strong positive conformity signals, consistent with previous AM-based measurements, but these signals are not reproduced with the ML halo masses. For the halo-level statistic, the representative AM-based signals are $+0.38\pm0.04$ dex and $+0.23\pm0.04$ dex for the luminosity-ranking and mass-ranking AM halo masses, detected relative to no conformity at about $10\sigma$ and $6\sigma$, respectively. In contrast, the ML result is consistent with no conformity, $+0.00\pm0.03$ dex; the satellite-level statistic gives a similar result. Thus, with halo-mass estimates consistent with weak-lensing constraints, we find no detectable one-halo conformity signal in the present SDSS sample, suggesting that the strong AM-based signal is largely driven by halo-mass estimation biases.
Read the full abstract

All authors Dingyi Zhao, Yingjie Peng, Luis C. Ho, Houjun Mo, Cheqiu Lyu, Kai Wang, Yijun Wang, Tao Wang, Jing Dou, Zeyu Gao, Qiusheng Gu, Yukun Liu, Roberto Maiolino, Alvio Renzini, Bitao Wang, Yu-Chen Wang, Bingxiao Xu, Feng Yuan, Kunyao Zhao, Xingye Zhu

Computer Vision

GRIPNet: Gaussian Radial Intensity Prior Guided Architecture for Pulmonary Nodule Detection in CT

Lung cancer causes more deaths than any other malignancy, and low-dose CT screening is the main pathway to early diagnosis. That pathway hinges on the smallest lesions, yet nodules below six millimeters remain hard to detect, because most methods treat a nodule as a generic object and ignore the imaging physics behind its appearance. We show that this appearance is highly regular. Intensity peaks at the geometric center of a nodule and decays radially in a Gaussian pattern, and a fit to 18,218 annotated lesions from three public benchmarks yields a mean radial coefficient of determination above 0.86 in every dataset and size stratum. A square convolution samples both axes uniformly and is mismatched to this radial signal, most severely for small nodules. Guided by this evidence, we propose GRIPNet (Gaussian Radial Intensity Prior Network), a detector in which every module maps to a measurable property of the intensity distribution. Pinwheel convolutions decompose radial gradients, a dual-frequency module separates boundary detail from structural context, dilated masked attention matches the decay extent, and an adaptive loss reweights samples by conspicuity. GRIPNet raises mAP@0.5 to 95.3, 91.6 and 97.9 percent on KanserSet, LUNA16 and Lung-PET-CT-Dx while sharpening high-IoU localization at real-time speed.
Read the full abstract

Analysis of PDEs

On an evolutionary equation involving the nonlocal infinity Laplacian

We study the evolutionary equation $$\frac{\partial u}{\partial t} = (\mathcal{L}_{\infty})^{\alpha}u \quad \text{in} \quad \Omega,$$ where $(\mathcal{L}_{\infty})^{\alpha}u$ denotes a nonlocal infinity Laplacian acting on the function $u$, $0<\alpha \leq 1$ and $\Omega$ is a bounded open set in $\mathbb{R}^{n+1}$. We prove existence and uniqueness using Perron's method for the Dirichlet problem when $\Omega$ is a cylinder and $0<\alpha<1$.
Read the full abstract

Computer Vision

Your Model Already Knows Don't Teach It, Learn to Ask It: Soft Prompting for Few-Shot Adaptation of Vision-Language Models

We address few-shot object detection with vision-language models (VLMs) in out-of-domain settings such as aerial, industrial, and medical imagery, using only ten annotated images for supervision. Existing adaptation methods are discrete prompt optimization and LoRA fine-tuning. We revisit a third option: soft prompting, where a small number of continuous prompt tokens are optimized while the pretrained backbone remains frozen. We identify two key design choices. First, placing prompt tokens at the cross-modal boundary between visual and text tokens outperforms other placements (10.0 vs. 8.4 mAP). Second, initializing prompts from the empty space token outperforms semantic and random initialization. With these choices, one to three learned tokens (7,168 parameters on average) match the best LoRA configuration on Roboflow20-VL (14.2 mAP, 10-shot) while training over 20,000x fewer parameters. Soft prompting remains harder to optimize, exhibiting higher variance across random seeds. Unlike LoRA, however, it causes no forgetting: the LoRA rank matching our accuracy reduces NaturalBench VQA accuracy by 35% relative, rising to 56% at the largest rank, whereas soft prompting leaves pretrained performance unchanged. The learned tokens behave like prompts rather than weights. They transfer to a newer model without retraining (+0.8 mAP on Qwen3.5-9B) and can be verbalized into readable prompts competitive with prompt-search methods (matching DetPO and outperforming GEPA). The approach also extends beyond detection. On RoboCasa manipulation tasks, the frozen $\pi_{0.5}$ vision-language-action policy benefits from soft prompting, matching the LoRA baseline on two of three tasks when tokens are placed at the gradient bottleneck. These results suggest modern VLMs already encode much of what is needed for specialized domains; the challenge is learning how to ask.
Read the full abstract

All authors Gautam Rajendrakumar Gare, Siyi Li, Hewei Wang, Cesar Daniel Hernandez, Wei Zhao, Wolfgang M. Pauli, John Galeotti, Deva Ramanan

Statistical Methodology

Byzantine-tolerant distributed learning of finite mixture models under partial corruptions

Finite mixture models characterize heterogeneous populations and are increasingly fitted to distributed data using split-and-conquer procedures that aggregate local mixture estimates at a central server. The aggregation step can be seriously compromised when transmitted local mixture estimates are partially or completely corrupted. To guard against Byzantine failures, existing robust aggregation methods have been developed for settings in which a local mixture estimate is either entirely authentic or entirely corrupted. Such methods can discard useful information when only some component estimates are corrupted. We consider component-wise Byzantine failure, in which some component estimates may be corrupted, but for each mixture component, a majority of the corresponding local estimates remain authentic. We propose component-wise filtered mixture reduction (CFMR), which selects a data-driven anchor, aligns transmitted components, filters each aligned cluster by a majority radius, and aggregates the retained estimates through mixture reduction. By filtering out only unreliable components, CFMR preserves authentic information from partially corrupted machines without requiring knowledge of the failure rates. We establish an adaptive convergence bound for CFMR and show that, under suitable conditions, it attains the oracle rate that would be achieved if the authentic component estimates were known in advance. Simulations and a real-data application show that CFMR remains close to the component-level oracle, whereas whole-machine filtering and unprotected aggregation can deteriorate substantially.
Read the full abstract

All authors Yimei Zhang, Jiahua Chen, Xiaozhou Wang, Yan Shuo Tan, Qiong Zhang

Robotics

2AM: Grounding Agent-Side Memory as Guidance for Steerable Action Models in Long-Horizon Manipulation

Long-horizon robot manipulation requires memory, but not necessarily inside the action policy. To address such tasks, current agentic systems often combine VLAs with planners and geometric tools, sometimes using additional depth or calibrated geometry. These systems confound attribution: gains may come from richer observations or alternative motor tools, while failures may stem from either the policy or an under-specified language interface. We isolate this question through a deliberately constrained design: less tool breadth, but greater interface bandwidth. 2AM makes a multimodal Agent the sole holder of task memory and a single RGB-based, episodically stateless Action Model the sole executor of task-relevant motion. The Agent compiles interaction history into subtask language and optional 2D grasp, place, and move hints that bind its physical intention at different time scales. To teach this steerability to the VLA, we augment demonstrations with structured hint labels and train under condition dropout, spatial noise, and temporal jitter to tolerate imperfect Agent outputs. On LIBERO-Mem, without depth, online geometry, or planner-based object motion, 2AM reaches 76.3% average completion, a 61.5-point improvement over the strongest reported baseline of 14.8%, together with 63.0% relaxed and 11.8% strict success. These results show that task memory can remain Agent-side. They further show that Action Model capability depends not only on what the policy has learned, but on how precisely the Agent can steer it. Project website: agent-mem-action-model.github.io
Read the full abstract

All authors Yutong Hu, Fengjiao Chen, Xuezhi Cao, Renaud Detry

Materials Science

Interface-Controlled Phase Stability in Polymorphic HfO$_2$ Revealed by Machine-Learning Atomistic Simulations

HfO$_2$ exhibits rich polymorphism, and competition among different phases underpins many of its functional properties. Yet bulk free-energy relations alone cannot explain phase selection at mixed-phase boundaries, where interface orientation and structural continuity constrain collective rearrangements. Here, using machine-learning atomistic simulations and a Hf-centered local phase classification scheme, we show that crystallographic interface matching redirects phase competition and accessible transformation pathways. The M(100)/T(100) interface remains pinned as an M/T mixture throughout 3 ns simulations from 300 to 1800 K. M/PO, M/AO, and PO/AO interfaces retain two-phase coexistence up to 900 K, whereas all T/PO interfaces become PO-dominant. At 1800 K, all non-M interfaces become T-dominant, while M-containing interfaces retain a monoclinic majority. Nudged elastic band calculations reveal lower-barrier routes through interface states. For M(100)$\rightarrow$T(100), the interface-mediated route in a long cell ($\sim$12 nm) yields a barrier of 136.03 meV/f.u., 29.2 % lower than the direct route in a short cell ($\sim$3 nm). This difference is associated with sequential phase-front motion absent from the short cell. These results identify phase boundaries as active participants in phase stability and transformation and establish interface orientation and crystallographic matching as variables for stabilizing metastable polymorphs and directing phase conversion in HfO$_2$.
Read the full abstract

Cosmology and Nongalactic Astrophysics

Initial clustering of primordial black holes: A general formulation for arbitrary local non-Gaussianity

Initial spatial clustering of primordial black holes (PBHs) induced by local-type non-Gaussianity (LNG) can substantially modify cosmological constraints on PBH abundance. Several inflationary scenarios that enhance curvature perturbations at small scales relevant to PBH formation predict LNG that is not necessarily perturbative. Therefore, it is crucial to establish a theoretical framework capable of investigating the initial clustering induced by arbitrary LNG. Here, we present a general analytical formulation for the PBH two-point correlation function applicable to arbitrary LNGs in the alternative approach. Under the assumptions that PBHs form only at the large peaks of perturbations, and that large-scale modes weakly modulate the local variance of small-scale perturbations, we derive an analytic expression for the PBH bias parameter, directly connecting initial clustering to the primordial trispectrum in the collapsed limit. We demonstrate the versatility of our formula by computing the bias parameters in the ultra-slow-roll inflation, curvaton, and modulated reheating scenarios. We also formally generalize the framework to broad power spectra to account for correlations across different PBH mass scales. Because our formulation does not rely on weak or perturbative non-Gaussianity assumptions, our result provides a universal theoretical basis for evaluating initial clustering impact on PBH observables.
Read the full abstract

High Energy Physics (Phenomenology)

Broad p-Wave Sommerfeld Resonances From Dimensional Deconstruction

We investigate whether dimensional deconstruction can qualitatively modify long-range forces. We show that the deconstruction of a five-dimensional AdS spacetime effectively generates an inverse-square potential, which naturally gives rise to broad p-wave Sommerfeld resonances without fine-tuning. We then apply this mechanism to dark matter phenomenology.
Read the full abstract

Analysis of PDEs

Sharp sign criterion for global existence and blow-up of strong solutions with arbitrarily large initial data in the 3D spherically symmetric compressible Navier-Stokes equations

We establish a sharp criterion for global-in-time existence versus finite-time blow-up of strong solutions to the 3D spherically symmetric compressible Navier-Stokes equations on a solid ball, based solely on the initial sign of effective velocity. The viscosity coefficients are assumed to satisfy the Bresch-Desjardins structure $\mu=\rho^{\alpha}$, $\lambda=(\alpha-1)\rho^{\alpha}$, with the effective velocity given by $v=u+\alpha \rho^{\alpha-2}\rho_r$. Previously, global existence results for strong solutions in higher dimensions were restricted to the case $\alpha\le 1$ with the endpoint $\alpha=1$ corresponding to the viscous Saint-Venant (shallow water) system. In this paper, we extend the global existence theory beyond this threshold to the supercritical regime $\alpha>1$. Specifically, whenever the initial effective velocity is nonnegative on the boundary of the ball, we prove the global-in-time existence of strong solutions for arbitrarily large initial data. The main difficulty lies in deriving a uniform lower bound for the density, which requires handling singular estimates at the center of the ball. To overcome this, we exploit several novel quantities near the center and employ a refined maximum principle to establish, for the first time, the Lipschitz continuity of the velocity at the center. Subsequently, we derive Dini-Gronwall-type inequalities for suitably constructed functions, which close the density lower bound estimate. Conversely, we can construct a family of initial data for which the initial effective velocity is negative on the boundary, such that the corresponding strong solutions blow up in finite time, with vacuum appearing exactly on the boundary at the blow-up time. Our results establish a clean dichotomy between global regularity and singularity formation, governed purely by the initial sign of the effective velocity.
Read the full abstract

Software Engineering

Exploring the Role of Security Experience and ChatGPT Usage Strategies on Secure Software Engineering Education

The rapid adoption of Large Language Models (LLMs) is reshaping software engineering education, but their role in secure software engineering education remains underexplored. We report an exploratory empirical study of how 26 graduate students in a part-time MSc Cybersecurity programme used ChatGPT during a vulnerability-fixing assignment. To characterise ChatGPT use, we analysed students' ChatGPT interaction logs using a structured double-coding procedure and examined whether usage patterns and prior cybersecurity expertise were associated with assignment performance. The results show that students with varying levels of cybersecurity expertise used broadly similar ChatGPT strategies. Individual usage patterns showed descriptive differences by grade, but none remained statistically significant after correcting for multiple comparisons. In contrast, diversity of ChatGPT usage, i.e., the number of distinct usage patterns adopted, was positively associated with performance, even after controlling for cybersecurity expertise. These exploratory findings suggest that the way students engage with ChatGPT may be more informative than whether they use it, and motivate future controlled studies to guide students toward effective LLM use in secure software engineering education.
Read the full abstract

All authors Alessio Ferrari, Minh An Nguyen, Kushal Ramkumar, Liliana Pasquale

Computation and Language

Automatic Lyric Transcription for Greek Songs: Scaling and Task Composition Effects in Whisper Adaptation

Automatic Lyric Transcription (ALT) remains substantially more challenging than speech recognition due to melodic variability, rhythmic irregularity, and accompaniment interference. This is heightened in low-resource languages like Greek, where no prior benchmark for ALT exists. We present the first controlled study of Whisper adaptation for Greek ALT, investigating model scaling effects, task composition via multitask training in transcribe-translate ratios, and two-stage speech-to-singing adaptation. We also curate a segment-level aligned singing dataset based on the Greek Audio Dataset (GAD) using source separation and CTC forced alignment. Results show that scaling consistently improves performance, while multitask learning acts as a beneficial regularizer primarily for smaller-capacity models. The 2-stage adaptation in Whisper Large-v3 achieves a Word Error Rate (WER) of 27.2%, a significant improvement over zero-shot baselines, establishing the first Greek ALT benchmark.
Read the full abstract

All authors Maria Frangiadaki, Dimitrios Damianos, Kosmas Kritsis, Vassilis Katsouros

Probability

A threshold for full packing dimension of Hölder images of sets and measures

We study when the image of a measure under a random Hölder map attains full packing dimension. Our main tool is a family of packing intermediate dimension profiles $\dim_{P,\theta}^{s}\mu$, indexed by $\theta\in(0,1]$ and $s>0$, which refine the packing dimension profiles of Falconer and Howroyd and reduce to them at $\theta=1$. For a large family of random $\alpha$-Hölder maps $f_\omega:\mathbb{R}^n\to\mathbb{R}^m$, which includes index-$\alpha$ fractional Brownian motion as a particular case, we prove that for every compactly supported Borel probability measure $\mu$ on $\mathbb{R}^n$, $$\dim_P \mu_{f_\omega}=m \quad\text{almost surely} \quad\Longleftrightarrow\quad \alpha m\leq \lim_{\theta\to 0}\dim_{P,\theta}^n \mu.$$ We further study the profiles $\dim_P^s\mu$ themselves, obtaining a quantitative lower bound and a Marstrand-type identity for their limiting behavior as $\theta\to 0$. Finally, we obtain the analogous characterization for analytic sets: $$\dim_P f_\omega(E)=m \quad\text{almost surely} \quad\Longleftrightarrow\quad \alpha m\leq \lim_{\theta\to 0}\dim_{P,\theta}^n E,$$ where $\dim_{P,\theta}^n E=\sup\{\dim_{P,\theta}^n \mu :\, \mu\in \mathcal{M}_c^+(E) \}$.
Read the full abstract

Analysis of PDEs

A Lagrangian superposition principle for the continuity equation with reaction

We study the continuity equation with reaction $\partial_t\mu + \operatorname{div}(v\mu) = w\mu$ on $[0,T]\times\mathbb{R}^d$, driven by Borel velocity and reaction fields $v$ and $w$. We provide the first version of a Lagrangian superposition principle for the equation in its natural generality, assuming only finite quadratic energy without imposing boundedness conditions on the reaction term neither any form of compactness for the support of the measure. More precisely, every solution $\mu \in \mathcal{C}([0,T];\mathscr{M}_+(\mathbb{R}^d))$ with finite quadratic energy is represented by a probability measure $\eta$ on absolutely continuous curves in the geometric cone over $\mathbb{R}^d$, concentrated on the solutions of the characteristic system $x' = v(x)$, $k' = w(x)\,k$, through the $2$-homogeneous marginal $h^2_t(\eta)=\mu_t$. We also establish the converse implication. The proof passes through a regularization of the triple $(v,w,\mu)$ which produces fields satisfying only local bounds; the core of the argument is therefore a representation theory under such local assumptions, whose key tool, of independent interest, is a two-time representation formula relating $\mu_s$ and $\mu_t$ along the flow of $v$ for arbitrary times $s,t$.
Read the full abstract