Peer Revenue Publish or profit

Papers

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

All of arXiv

Saturday 5 September

Materials Science

Elucidating Guest-Host Mechanisms in ZIF-L for Tuneable Highly Luminescent 2D Materials

3D metal-organic frameworks (MOFs) are well known effective hosts for luminescent guests that improve tuneability, photostability and material fabricability. 2D guest@MOF systems, while less explored, offer competitive advantages over 3D guest@MOF systems due to their optical transparency, inter-layer spacing, and vertical thinness. This work examines luminescent organic dyes@ZIF-L to establish the underlying mechanisms of guest incorporation in ZIF-L and demonstrate the advantages of the 2D ZIF-L architecture for the material's functional luminescence. Analysing a case study system, fluorescein@ZIF-L (F@ZIF-L), using nanoscale FTIR, diffraction, and topology mapping, confirmed that fluorescein (F) resided in the ZIF-L framework cavities. Supported by surface energy simulations, the extent of guest incorporation was found to be indicated by a morphological continuum, from the characteristic leaf-shaped ZIF-L to rectangular F@ZIF-L. By modifying synthesis temperature and solvent ratios, the luminescent properties of F@ZIF-L could be rationally tuned in terms of guest loading (% mol) and arrangement (i.e. guest monomer to aggregate ratio). When optimised, F@ZIF-L exhibited tuneable emission chromaticity, 99.7% photoluminescent quantum yield, minimal guest leaching in solution over 12 months, and high photostability. Perylene@ZIF-L exhibited unique white light emitting properties, with CIE coordinates (0.33, 0.34) arising from a combination of yellow alpha-phase excimer and blue monomeric perylene emission. Finally, oriented luminescent thin films of guest@ZIF-L materials were grown on malleable Zn foils, demonstrating an in situ fabrication technique. Together, the work highlights the potential for luminescent dye@ZIF-L systems in developing tuneable and resilient luminescent components of next-generation optoelectronics, sensors, and lighting systems.
Read the full abstract

All authors Dylan A. Sherman, Lorenzo Donà, Cyril Besnard, Lars Mester, Ben Slater, Judit Farrando-Pérez, Christopher Allen, Joaquín Silvestre-Albero, Jin-Chong Tan

Cryptography and Security

Evaluating Deep-Search Agents under Hierarchical Web Evidence Poisoning

Search-augmented LLM agents are increasingly used for consumer decisions, making them vulnerable to Generative Engine Optimization (GEO) poisoning. Existing benchmarks largely measure whether manipulated content is retrieved or endorsed, but do not track whether an agent verifies suspicious evidence, revises adopted claims, or recovers before producing its final recommendation. We introduce HAE-GEO, a benchmark that tracks the full trajectory from exposure to recovery under progressively more persuasive Web poisoning. Agents interact via a multi-turn Search-Scrape interface across three attack levels (L1 direct assertion, L2 contextual camouflage, and L3 apparent corroboration), supported by a controlled corpus of 72,039 clean pages and 770 poisoned pages per level spanning 8 product categories and 154 brands. Evaluation combines deterministic behavioral measures with six semantic rubric dimensions. Evaluating 10 agents, we find three recurring patterns: evidence recognition degrades under the corroboration trap; agentic search improves final resistance without improving evidence recognition or utility; and defense prompting increases verification, yet rarely converts verification into recovery.
Read the full abstract

All authors Zhongan Bi, Qiwen Wang, Jianrong Jiang, Jigang Ding, Wenwen Xiong, Changhua Meng, Xuanang Gao, Kepeng Lin, Changjiang Jiang, Yiang Chen, Huan Yao, Wei Wang, Zhenyu Ma, Wenhui Dong

Analysis of PDEs

A globally defined compressible polyconvex isotropic energy satisfying the true-stress-true-strain monotonicity condition (TSTS-M++)

We construct a compressible isotropic energy on $\mathrm{GL}^+(3)$ combining polyconvexity and global true-stress-true-strain monotonicity (TSTS-M++), answering a question of Wollner, Holzapfel and Neff. For every $\mu>0$ and $k>0$, the energy $W_k(F)=\frac{\mu}{2k}[\exp(k(\lVert F\rVert^2+3J^{-1}+J-7))-1]$, $J=\det F$, is polyconvex and strictly rank-one convex. Its Cauchy-stress response satisfies TSTS-M++ globally if and only if $k\ge1/(8\sqrt3)$. The reference state is stress free with positive infinitesimal shear and bulk moduli. Stress bijectivity has the smaller sharp threshold $k_{\rm B}\approx0.0083$: at equality the stress map is a global homeomorphism with a nondifferentiable inverse, and above it a global smooth diffeomorphism. Thus, for $k_{\rm B}\le k<1/(8\sqrt3)$, the Cauchy-stress map remains globally bijective while TSTS-M++ fails at finite strain. In the TSTS-M++ regime, each prescribed inner radius of a finite plane-strain annulus with a traction-free outer wall has a unique radial equilibrium, and the inner pressure increases smoothly and strictly from zero to infinity. Comparisons with other energies distinguish the roles of polyconvexity, TSTS-M++, reduced convexity and endpoint growth. Under incompressible spherical inflation, global pressure inversion holds for the Demiray family with $b\ge b_*$, including $W_k|_{J=1}$ for $k\ge k_*$ and the stress-free $I_1$-exponential energy, and for the exponentiated Hencky energy when $k_{\mathrm{eH}}>3/8$. Incompressible neo-Hooke instead has a limit point. A further compressible $I_1$ energy retains the global constitutive properties and reference moduli of $W_k$ but exhibits two pressure turning points under incompressible spherical inflation, showing that these properties allow a descending pressure branch.
Read the full abstract

All authors Dongxin Bai, Yunhao Wu, Yong Li, Kai Zhang, Patrizio Neff

Computation and Language

Factors Influencing the Emergence of Dependency Length Minimization in Neural Agent Simulations

Given various grammatical options, language users prefer the word order choice that reduces the overall length of syntactic dependencies, a principle known as dependency length minimization (DLM). The origins of this preference remain an open question, particularly whether it originates from constraints on efficient information processing. Computational simulations provide a powerful approach to identifying the factors influencing the emergence of linguistic phenomena. However, previous simulations of DLM have not examined realistic interaction contexts and have produced mixed results. The present study investigates the emergence of DLM in artificial languages using a recently proposed language learning and communication framework based on recurrent neural networks (RNNs). In this framework, agents are trained to speak and interpret artificial languages and then use these languages to communicate. Using this framework, we study the impact of several factors related to processing limitations in a communicative setting, such as noise during listening, limited speaker capacity, and incremental sentence processing. Our results reveal a complex interplay among these factors in shaping word order preferences in neural agents. Specifically, in the full meaning space, agents regularize toward a single dominant word order, while in the half meaning space they show a short-before-long preference that only aligns with DLM in verb-initial languages. A consistent DLM preference emerges only when agents are subject to incremental processing pressure. These findings suggest that limitations in human cognitive processing may indeed play a role in shaping DLM. Our findings provide insights into the conditions under which neural models replicate human-like preferences and highlight the challenges of designing emergent communication models that capture human cognitive biases in language processing.
Read the full abstract

All authors Yuqing Zhang, Tessa Verhoef, Gertjan van Noord, Arianna Bisazza

Optics

Motion-Tracking Terahertz Time-Domain Spectroscopy Enabled by Time-Programmable Frequency Combs for Moving-Object Sensing

Terahertz time-domain spectroscopy (THz-TDS) provides direct access to both spectral information and time-of-flight features, making it attractive for dynamic sensing. However, conventional high-speed THz-TDS methods typically rely on a fixed delay trajectory, limiting their ability to selectively acquire only the relevant temporal window while tracking target motion. Here, we demonstrate motion-tracking THz-TDS based on time-programmable frequency combs (TPFCs), in which the THz measurement window is actively locked to a moving time-domain waveform. By continuously modulating the phase-lock set point of one TPFC, we perform real-time apodized acquisition of a specific temporal window centered on the main THz peak at rates up to 308 Hz. Simultaneously, the residual peak position within the window is detected in real time and fed back to the other comb, enabling the measurement window to follow target displacements exceeding the original acquisition window. The feedback command together with the residual peak position further enables reconstruction of the relative target displacement. As a proof-of-concept demonstration, we use a moving gold mirror to simulate respiration- and heartbeat-related motion superimposed on a much larger displacement representing body motion, and successfully resolve the two vital-sign-related components from the measured displacement. This work extends arbitrary-delay control with TPFCs to THz-TDS and establishes a motion-compensated spectroscopy platform for simultaneous displacement tracking and THz waveform acquisition.
Read the full abstract

All authors Riku Shibata, Shun Fujii, Tomofumi Ikari, Shinichi Watanabe

Representation Theory

A Note on Additive Diameter

We record two results on additive diameters of group representations. First we give a corrected version of Proposition 5.3 from the first arXiv version of arXiv:2504.07573. Let $0<\varepsilon<1/3$, let $n>9/\varepsilon^2$, and let $U\leq\mathfrak{sl}_n(\mathbb{C})$ with $\dim U>\varepsilon n^2$. Then $\operatorname{diam}^{\mathrm{SL}_n(\mathbb{C})}_{+}(\mathfrak{sl}_n(\mathbb{C}),U)\leq 32/\varepsilon+8$. The proof uses an averaging argument for the action of the symmetric group on the off-diagonal matrix positions. This argument was subsequently developed further in arXiv:2609.03882. We also prove the right-hand inequality in Question 6.7 of arXiv:2504.07573. If $G$ is a complex algebraic group, $V$ is a finite-dimensional $G$-module, $U\leq V$, and $\mathfrak g=\operatorname{Lie}(G)$, then $\operatorname{diam}^{G}_{+}(V,U)\leq\operatorname{diam}^{\mathfrak g}_{+}(V,U)$. No irreducibility or connectedness assumption is needed.
Read the full abstract

Data Structures and Algorithms

Fast Evaluation of Polynomials with Rational Preprocessing

Horner's rule evaluates a monic degree-$n$ polynomial using $n-1$ multiplications. We show that with rational preprocessing of the coefficients, any such polynomial can be evaluated using only $\lfloor n/2 \rfloor + 1$ multiplications over fields of characteristic zero or of characteristic $p>n$. This resolves the multiplication side of a conjecture of Rabin and Winograd (Comm. Pure Appl. Math 1972), who achieved $n/2 + 2\lceil\log_2 n\rceil$ multiplications and conjectured the logarithmic overhead was necessary. We show that this multiplication count can't be beaten in general, proving that three multiplications do not suffice for degree~$6$. This strengthens the lower bound of Pan (STOC 1978), who proved a tight bound for general, complex preprocessing. In characteristic~2, for every $n>1$ and every finite field of size at least $2n$, we prove that an $n$-multiplication chain cannot parametrize all value vectors at $2n$ distinct evaluation points, even with arbitrary preprocessing. We give $\lfloor n/2 \rfloor + 1$ multiplication schedules over characteristic~2, each with an explicit inverse, for every odd degree $n\le 25$ and conjecture that this is possible for all $n$. We also give an injective polynomial construction for universal hashing that uses $N$ multiplications to hash $2N$ values with a single random key. This improves the best previous construction by Daniel J. Bernstein (cryp.to).
Read the full abstract

Networking

Delay and Throughput Analysis of Computation Offloading in Mobile Edge Computing: A Queueing Network Approach

Mobile edge computing (MEC) enables mobile devices to offload computation to nearby edge servers and to the cloud in order to reduce end-to-end delay for different applications. In this paper, we study static computation offloading when each task consists of multiple dependent subtasks represented by a rooted directed tree. We develop a product-form queueing-network (PFQN) model with an approximation to capture the computation and communication dynamics of tree-structured task execution in a multi-tier MEC system. Based on this model, we derive closed-form expressions for server utilizations and waiting times, and then construct a recursive algorithm for evaluating the average delay of general tree-structured tasks. We formulate the static offloading design problem as the minimization of the rate-weighted average task delay over the routing probabilities, and solve it through a differentiable optimization framework based on softmax parameterization, log-sum-exp smoothing, and a stability barrier on server utilizations. Numerical results show that the proposed PFQN approximation provides accurate delay estimates and that the delay-optimized static policy consistently outperforms the considered baseline algorithms in terms of both average task delay and maximum stable throughput.
Read the full abstract

Computer Vision

What Does Animal Re-Identification Learn? Linear Biological Concepts and Their Origins in Visual Representations

Conservation increasingly relies on camera traps that collect more wildlife imagery than experts can manually analyze, making animal re-identification (Re-ID) essential for monitoring individuals and populations. Yet understanding which cues drive model decisions is challenging for ViT-based Re-ID models, whose metric-learning objectives provide no explicit supervision for biological concepts. We ask whether such models nonetheless organize their representations along biologically meaningful axes. Using a DINOv3 backbone fine-tuned for Western lowland gorilla Re-ID with triplet-margin loss, we find that sex and age emerge as linear directions that generalize to held-out individuals, reaching up to 0.91 AUROC and being recoverable from a single image per individual. Activation steering further shows that the sex direction is causally used by the model, flipping a significant fraction of predictions to the opposite sex. Comparing off-the-shelf and fine-tuned backbones shows that Re-ID training does not create these concepts, but relocates them across the network. Finally, data attribution reveals that the representation we find reflects a graded biological axis, is redundantly encoded across the population and shaped by visually ambiguous individuals. Together, these findings show how interpretability can uncover both the biological structure and failure modes of Re-ID representations, providing a step toward auditable computer vision for wildlife monitoring.
Read the full abstract

All authors Robert Nolting, Alexandra Schild, Moritz Weckbecker, Maximilian Schall, Gerard de Melo

Materials Science

Oxidation of Tantalum Nano-Film by Microwave Exposure

Oxidation and ablation of 200 nm tantalum films were carried out by three routes: (i) femtosecond (fs-)laser direct write, (ii) high-temperature annealing (HTA) in a tube furnace, and (iii) annealing in a 2.45 GHz microwave cavity. Complete conversion of the 200 nm Ta layer into 409 nm of Ta2O5 required one hour at 600 C in the furnace, but only minutes at ~ 50 W of microwave power. Fs-laser (515 nm/200 fs) oxidation of the Ta nano-film set in at an average single-pulse fluence of ~ 0.1 J/cm2 under strong pulse-to-pulse overlap (900 pulses per focal spot), i.e. within a narrow window bounded from above by the onset of ablation. Under microwave annealing, both the cavity resonance frequency and the quality factor Q changed markedly at the metal-to-oxide transition, reflecting the collapse of the real and imaginary parts of the permittivity at 2.45 GHz. This dielectric contrast turns the cavity into a sensor: the oxidation can be followed in real time from the shift of the cavity resonance, providing a non-invasive, in-situ diagnostic tool.
Read the full abstract

All authors Massimiliano Zamengo, Lina Grineviciute, Hsin-Hui Huang, Haoran Mu, Julianija Nikitina, Saulius Juodkazis, Junko Morikawa

Machine Learning

IXPLORE: Bounded Ideal Point Estimation with Grid-Based Uncertainty Quantification

Ideal point estimation is widely used to analyze and visualize political data. However, selecting the corresponding spatial model involves various trade-offs: while model-based approaches such as Item Response Theory (IRT) are based on utility functions rather than optimized for predictive accuracy, most Machine Learning (ML) alternatives struggle to generalize beyond training data when embedding sparse test responses. We introduce IXPLORE, a bounded ideal point estimation algorithm that combines a predictive fit objective with a sparsity-aware likelihood function. On five benchmark datasets spanning surveys, roll calls, and deliberation, this approach surpasses model-based and ML-based algorithms on reconstruction and imputation error - especially for users with sparse responses. Furthermore, we show that non-linear feature transforms can further reduce the reconstruction error while remaining visually interpretable. To quantify uncertainty, IXPLORE applies grid-based posterior inference on a bounded 2D latent space. Available as a Python package on PyPI, IXPLORE offers a flexible framework for constructing bounded, interpretable political maps with fast inference and strong imputation performance.
Read the full abstract

Computer Vision

FujinSplat: Seeing Through Smoke with RAW-Domain Gaussian Splatting

The appearance of a smoky scene is shaped by two processes that a camera records together: the participating medium alters scene radiance in a view-dependent way, and the image signal processor (ISP) then remaps the result through a nonlinear tone and color transformation. Recovering a clean 3D scene requires separating both. Per-view sRGB dehazing acts only after the ISP has entangled them; standard 3D reconstruction ignores the medium and absorbs it into scene geometry and radiance. FujinSplat addresses the problem in the RAW domain, where the two processes remain separable. A per-scene Base ISP is fitted from the scene's hazy RAW captures to its own camera renderings and then frozen, providing a fixed photometric anchor that performs no dehazing. Analyzing expert corrections reveals a compact, low-dimensional correction space identifiable from RAW alone. FujinSplat therefore fits per-view action answers at the training poses and trains a single scene-agnostic controller to regress them from RAW; the corrected views supervise one static 3D Gaussian representation, jointly with a bounded per-view residual that reconciles cross-view photometric inconsistencies. On the RealX3D real-world smoke benchmark FujinSplat clearly outperforms the strongest comparable baseline, ahead of both physics-based reconstruction and restoration-then-3DGS pipelines. Code is available at https://github.com/I2WM/FujinSplat.
Read the full abstract

Machine Learning

Granular-Ball Quantum Clustering for Resource-Efficient and Robust Learning

Quantum clustering aims to exploit quantum feature representations to uncover complex data structures beyond conventional Euclidean geometry. Yet this sample-level kernel construction requires O(n^2) quantum circuit executions for n data points, creating a major bottleneck under near-term quantum resource constraints. Prior solutions fail to resolve this efficiency-accuracy dilemma: classical granular-ball clustering reduces sample complexity but relies on Euclidean metrics that cannot capture quantum correlations, while existing quantum compression schemes prioritize efficiency over structural preservation, degrading performance on non-convex or noisy data. Here we propose Granular-Ball Quantum Clustering (GBQC), a framework that tightly couples granular-ball structural abstraction with quantum feature learning. GBQC first compresses raw data into compact, representative granular balls via a PCA-guided splitting strategy, reducing kernel evaluations by 80% compared to full-sample methods. A quantum cohesion mechanism then filters noisy granules in Hilbert space to improve clustering robustness. Extensive experiments on synthetic, noisy, overlapping, and real-world datasets demonstrate that GBQC consistently achieves superior clustering accuracy and robustness compared with representative classical and quantum clustering methods. Meanwhile, the proposed granular-ball compression significantly reduces quantum kernel evaluations and computational overhead, enabling quantum clustering experiments on larger datasets within parameterized quantum learning frameworks. These results suggest that granular-ball representations serve not only as a compression mechanism to reduce quantum computational costs but also as an effective structural abstraction mechanism that improves clustering quality by eliminating redundant and structurally ambiguous learning units.
Read the full abstract

All authors Suzhen Yuan, Qilin Xie, Lifeng Shen, Shuyin Xia, Jermiah D.Deng, Guoying Wang

General Physics

Stratification Criteria for Machine Learning Pattern Discovery in Particle Physics: Preparing for the AlphaFold Moment

Machine learning capabilities are expanding into scientific domains at an accelerating pace. When applied to high-energy physics pattern discovery, they will generate candidates faster than traditional evaluation can absorb. ML finds patterns in past data it is inherently post hoc. Whether those patterns reflect structure or coincidence is unknowable at discovery time; this limitation applies equally to human and computational pattern-finding. What differs is scale: ML candidate generation is effectively unbounded, while human evaluation capacity remains fixed. When generation rate exceeds evaluation bandwidth, binary accept/reject degenerates to random sampling. Information-theoretically, the only response that preserves ranking under finite evaluation budget is stratification. By focusing on stratification rather than binary filtering, rule adjustments can be made retroactively, thresholds tuned as results accumulate, and evaluation bandwidth focused on top-ranked candidates. This paper attempts to codify those criteria, proposing seven computationally evaluable standards for stratifying ML-generated patterns. The goal is not to deliver verdicts but to prioritize which candidates merit pre-registration and longitudinal tracking. The framework preserves the essential paradigm: pattern plus theory equals potentially real physics. Patterns alone, however striking, remain candidates until theoretical understanding arrives. Making these criteria explicit enables prefiltering at scale while creating a collaborative resource rather than a competitive one. ML capabilities extend what physicists can search while preserving how physicists evaluate. We offer this provisional framework for community calibration, with the goal of developing validation infrastructure before the capability fully arrives.
Read the full abstract

Superconductivity

Microscopic calculation of coherence lengths and magnetic penetration depth in multiband superconductors

We present an extended Ginzburg-Landau (GL) method for calculating the superconducting coherence length and magnetic penetration depth at temperatures well below the transition temperature $T_{\mathrm c}$. In contrast to conventional GL theory, which expands the free energy in both the order parameters and their gradients, our method applies a perturbative expansion only to the covariant-gradient terms, while retaining the full dependence on the superconducting order parameters. The coefficients of these terms are determined from finite differences of microscopic free energies evaluated at small imposed pair momenta. The method applies to both single-band and multiband superconductors and therefore provides a framework for incorporating more realistic electronic structures. For the models examined here, the extended GL method agrees well with real-space Bogoliubov-de Gennes (BdG) calculations over a wide temperature range, while requiring substantially less computational effort.
Read the full abstract

All authors Tristan Ryoma Fuchs, Takuya Nomoto, Hikaru Watanabe, Ryotaro Arita

Analysis of PDEs

Remarks on a Liouville-type theorem by Chae and Wolf for stationary Navier-Stokes equations

In this note, we revisit a Liouville-type theorem of Chae and Wolf for stationary Navier-Stokes equations in $\mathbb{R}^3$ [J. Differential Equations 261 (2016) 5541-5560]. We show that their logarithmic improvement of the classical $L^{9/2}$ condition is part of a substantially broader weighted framework. More precisely, we prove that a solution $u\in\dot H^1(\mathbb{R}^3)$ is necessarily trivial whenever $$ \int_{\mathbb{R}^3}|u(x)|^{9/2}\,\omega(|u(x)|)\,dx< +\infty, $$ for every positive, nondecreasing and bounded weight $\omega$ satisfying a mild growth condition near the origin. This structural condition encompasses the logarithmic weight due to Chae and Wolf, as well as a hierarchy of iterated-logarithmic weights and (genuinely) non-logarithmic examples, including a dyadic weight. Our result identifies a broader class of weighted integrability conditions under which the triviality of stationary Navier-Stokes solutions follows, and shows that the mechanism underlying the Chae-Wolf improvement is not intrinsically tied to a specific logarithmic weight or to a single logarithmic scale.
Read the full abstract

Cryptography and Security

Robustness Evaluation and Detection of Transferable Adversarial Attacks in ML-Based NIDS

Machine learning-based network intrusion detection systems (ML-based NIDS) are vulnerable to adversarial evasion, where malicious samples are perturbed to evade detection and be misclassified as benign. Despite growing research on adversarial attacks and defenses for ML-based NIDS, comparative evaluations of multiple attack types, detection models, and defense strategies under a common setting remain limited. In this paper, we evaluate eight adversarial evasion attacks, fifteen detection models, and three representative defense strategies using the NF-UQ-NIDS dataset, which includes recent traditional and IoT network traffic with twenty distinct attack categories. The evaluation compares model performance on clean test data and on robustness evaluation sets that include adversarial samples, analyzes attack success consistency across models, and examines the effect of defense strategies on adversarial robustness. To support model comparison, we introduce the Robustness Index (RI), a compact comparative metric that rewards high balanced accuracy and macro-F1 score computed on the robustness evaluation set while penalizing high attack success rate (ASR). We further present AR-NIDS, a two-stage framework that uses an adversarially trained ensemble to distinguish normal, attack, and adversarial samples, followed by an adversarial attack classifier to identify the attack type. Under the evaluated transfer-based setting, the proposed adversarially trained ensemble achieves the strongest overall trade-off between classification performance on the robustness evaluation set and evasion resistance, reducing the average ASR from 0.41 to 0.03 and achieving an RI of 0.98.
Read the full abstract

Computation and Language

Tri-PvP: Exposing Modality Bias in Omni-Modal Large Language Models through Perceptual-Propositional Evidence Conflicts

Omni-modal large language models (OLLMs) jointly process vision, audio, and text, yet their modality bias under cross-modal conflict remains underexplored. Existing benchmarks conflate two distinct forms of evidence within a single modality: perceptual signals (e.g., a photograph or recording of a dog) and propositional signals (e.g., the declarative claim "this is a dog"), such that any measured modality bias is inherently confounded with evidence-form bias, precluding clean attribution to either source. To address this, we introduce Tri-PvP, an 8,000-sample tri-modal conflict benchmark crossing vision, audio, and text, where vision and audio each take perceptual or propositional form. Evaluating five OLLMs, we find robust visual bias across most models and evidence-type conditions. Crucially, we reveal a systematic asymmetry in evidence-form bias: models exhibit a stronger bias toward perceptual signal in vision but propositional in audio. Further analyses via layer-wise linear probing and contrastive decoding reveal that modality bias is already linearly decodable from early representation layers and can only be partially mitigated, calling for mitigation strategies beyond surface-level interventions.
Read the full abstract

All authors Yen-Ting Piao, Shu-Yun Chen, Chin-Hui Chu, Chun-Wei Chen, Shih-Yun Shan Kuan, Hung-yi Lee, Yun-Nung Chen

Robotics

How to Learn from What a Human Would Avoid? Intervention-Aware World Models with Real-World RL for Dexterous Manipulation

Multi-fingered dexterous manipulation remains a frontier for real-world reinforcement learning (RL) due to the high-dimensional action space and the prohibitive cost of hardware failures. While human-in-the-loop (HIL) RL allows operators to intervene before failures occur, current pipelines often treat these interventions as reactive corrections, discarding the rich safety signal inherent in the operator's decision to take control. In this paper, we ask: How can we learn from what a human would avoid? We present WHIRL, a safety-aware RL framework that transforms binary human interventions into forward-predictive signals for proactive risk avoidance. Our approach centers on an intervention-aware latent world model with four prediction heads: dynamics, reward, termination, and a novel per-state intervention-probability head that learns to predict the likelihood of a human takeover at future states. This head provides an actor-side risk-shaping term that discourages the policy from entering "intervention-prone" regions, modeling the operator's internal safety threshold. We evaluate our framework on a 16-DoF LEAP Hand across tasks spanning convex and irregular object grasping, prismatic manipulation, and long-horizon multi-stage tasks. Our results show that predictive risk-shaping enables the system to achieve a 96.7 percent success rate on complex grasping tasks while reducing the operator intervention burden by up to 84 percent in step-weighted terms. By closing the loop between human intuition and predictive world modeling, this work provides a practical safety-aware recipe for training complex dexterous agents in the real world while reducing operator fatigue and hardware-risk exposure.
Read the full abstract

All authors Jiaju Yin, Zhenhui Zhang, Lixin Xu, Heng Zhang, Jun Shao, Yating Feng, Arash Ajoudani, Renjing Xu

Information Theory

Cadence: Error-Bounded Lossy Compression of Demand Time Series with a Time-Series Foundation Model

We present Cadence, an error-bounded lossy compressor for numeric time series pairing a 330M-parameter time-series foundation model (Google TimesFM-3) with an adaptive arithmetic coder, guaranteeing $|\hat{x}_t-x_t|\le\tau$ on every sample. One negative result constrains the design space: for lossless coding a foundation model is worth nothing, because bits saved are logarithmic in predictor accuracy, $\Delta b=\log_2(\mathrm{MAE_{old}}/\mathrm{MAE_{new}})$. So the $1.51\times$ advantage TimesFM-3 holds over a 32-tap linear predictor buys 0.60 bits of 20.28, a median gain of +0.03%. Error-bounded coding escapes this at one point: once a forecast lands inside the band the residual index is zero and the sample nearly free. Cadence contributes: (1) an adaptive range coder with context-modelled binarization, beating xz/zstd on real indices by 9.7% (15/15) and reversing a finding from a general-purpose back end; (2) a determinism result – predictions are not bit-identical across batch sizes, and no PyTorch configuration repairs this, forcing group size and execution device into the container format; and (3) domain localization on corpora postdating any plausible training cutoff. On 49 EIA-930 balancing-authority demand series (2026) Cadence gains 13.3% over the best of six classical predictors, and 28.3% on 50 MTA ridership series (2026): 21.4% median over 297 series-tolerance pairs, winning all 297. Against downsampling, what time-series databases deploy for retention, its guaranteed worst-case error is $28$–$56\times$ tighter at equal size. End-to-end, once the context bootstrap is paid for, gains run from 6.8% at six months of hourly data to 15.1% asymptotically. Attempting to falsify the domain claim on SDRBench, theory predicts failure and delivers: -0.8% median, 0 of 27 pairs gaining. Three further negative results and eight retracted claims are reported in full.
Read the full abstract

Computer Vision

Depth-to-Image Synthesis-Driven Generative Unguided Depth Completion

Guided depth completion methods heavily depend on RGB quality and alignment, while unguided ones often suffer from limited precision due to the absence of explicit visual cues. In this paper, we present Depth-to-Image Synthesis-Driven Generative Unguided Depth Completion (GUDC), a new completion paradigm that innovatively bridges advanced 2D generative models with unguided depth completion, enabling semantics-aware depth inference without real RGB inputs. Our key idea is to exploit ControlNet's powerful depth-conditioned generation capability to synthesize pseudo-images directly from sparse depth, effectively converting the original unguided setting into a semantics-guided one. To address the potential image-depth misalignment caused by depth sparsity, we propose a multi-level dense-to-sparse representation distillation strategy for ControlNet fine-tuning, where dense-depth features act as teacher signals to distill consistent structural representations for sparse-depth inputs. Furthermore, during pseudo-image-guided completion, we propose a pseudo-image semantic attention fusion module to adaptively extract informative semantic cues from pseudo-images while suppressing artifacts (e.g., texture hallucinations). Extensive experiments on KITTI and NYUv2 validate that our GUDC achieves superior accuracy and robustness over existing methods.
Read the full abstract

Machine Learning

Memory in Deep Time-Series Models

Deep learning for time series has progressed through successive architectural paradigms, from recurrent networks and transformers to structured state-space models, retrieval-augmented predictors, foundation models, and tool-using agents. These developments are typically studied in isolation, organized by architecture or modeling era. We argue that they can instead be viewed through a common question of how does a time-series model retain and access information beyond its immediate input? This question is motivated by a fundamental limitation of conventional time-series modeling: information relevant to a prediction may lie far beyond a feasible input window, while compressing history into a fixed-size state can discard information that may become useful later. We formulate this challenge as a memory problem and organize existing time-series methods along a spectrum from internal memory, encoded in parameters and fixed-size states, to external memory that is addressable, retrievable, and increasingly maintained by agents. We then develop a unified taxonomy of memory mechanisms and review three classes of external memory, including explicit modules, retrieval augmentation, and agentic stores, under a common framework for what is retained, how it is written and accessed, and how it persists. A cross-cutting analysis maps these mechanisms to time-series tasks and identifies gaps in both methods and evaluation. We conclude by outlining open problems in building memory systems that can selectively retain, retrieve, revise, and forget information as temporal environments evolve. The result is a framework for studying memory as a first-class dimension of time series modeling, independent of the underlying backbone.
Read the full abstract

All authors Minh Hoang Nguyen, Huu Hiep Nguyen, Manh Nguyen, Van Dai Do, Dung Nguyen, Hung Le

Physics and Society

Price Dislocations, News Citations, and Epistemic Leverage on Polymarket

Prediction-market probabilities increasingly appear in news coverage, yet little is known about which market movements become news or how much trading money sits behind the numbers journalists quote. Unlike a poll, a market price can be moved by anyone willing to trade, so the cost of manufacturing a number that circulates as news bears directly on the information environment. We link 173.7 million signed Polymarket trades to news coverage from 2024-2025. From 6,990 articles mentioning prediction-market venues, an LLM-based, human-validated matcher extracts 1,582 sentences quoting market odds and attributes 918 to the specific market whose price they cite. We then detect 44,976 price dislocations, movements of at least five percentage points backed by concentrated one-sided trading, and ask whether a market is cited more often afterward. In the days after a dislocation, a market's citation rate is about 33% higher than its matched baseline (log citation-rate ratio $\tau_{\mathrm{cite}}=0.283$, permutation $p=0.001$), robust to binary and Poisson count outcomes. Yet move size is not the strongest predictor of citation: prominence dominates (standardized $\beta=0.610$ vs. $\beta=0.159$ for move size). Finally, we combine the dollar flow behind a given price change with observed citation rates into a metric we call epistemic leverage, the dollars needed to move a market five points and have the move cited. It stays near \$0.7-1.0 million across prominence quintiles, because cheaper-to-move markets are proportionally less likely to be cited. The implied threat model centers not on the long tail of cheaply moved markets but on the few prominent markets newsrooms treat as informational infrastructure, where a seven-figure price of influence sits within the budgets of actors with a large stake in the quoted number. We release aggregate event-study data and validation materials.
Read the full abstract

Computer Vision

Geometry-Aware Test-Time Learning for Quantitative Spatial Reasoning

Quantitative spatial reasoning in visual-language models (VLMs) aims to infer spatial distances and directional relationships among objects in 3D space from a 2D image and a natural language query. Despite recent progress, VLM spatial reasoning remains brittle under distribution shifts, largely due to the high cost of 3D supervision. As a result, models often produce inconsistent or contradictory predictions when faced with novel object configurations or rephrased spatial queries, revealing a misalignment between learned representations and underlying geometry. To address this, we propose TTL-SR, a geometry-aware Test-Time Learning framework for quantitative Spatial Reasoning that leverages geometric consistency constraints and unlabeled test data to adapt models to target domains. Specifically, TTL-SR augments the input query with geometrically coupled auxiliary queries, filters unreliable predictions via adaptive geometric triggering to construct structured token-level pseudo-labels, and updates model parameters under a geometry-aware multi-objective loss using only test data. Experimental results demonstrate that TTL-SR significantly boosts spatial reasoning performance, yielding 6.47% and 9.41% accuracy gains for Qwen3-VL-4B-Instruct and SpatialRGPT-VILA-1.5-8B on Q-Spatial-ScanNet dataset, respectively.
Read the full abstract

All authors Gege Zhang, Shuaicheng Niu, Gang Dai, Lei Sun, Shuangping Huang

Soft Condensed Matter

Controlling pair dynamics of rotating magnetic microparticles through radial and transverse interactions

Rotating magnetic microparticles are building blocks for field-driven assembly and microrobotic control. As the elementary interaction rule for larger assemblies, pair motion in these systems is governed not only by magnetic forcing, but also by hydrodynamic coupling and other long-range interactions. Here we develop a reduced framework for two synchronized rotating magnetic particles by resolving the interactions into radial components that change the interparticle distance and transverse components that rotate the line of centers. The competition between magnetic dipolar interaction, additional radial repulsion, and rotation-induced transverse coupling selects three pair-motion modes: rigid-body rotation, contact-separation rotation, and irreversible separation. We derive transition criteria for the rigid-body state, reversed orbital motion, and the separation boundary, and obtain an asymptotic solution for the separation dynamics. Lattice Boltzmann simulations of particles rotating near a wall provide a hydrodynamic realization of the model, in which inertial secondary flow generates radial repulsion and rotational flow produces transverse coupling. The resulting phase diagram in the physical $(\mathrm{Re},C_m)$ plane is consistent with the reduced-model predictions. These results provide design rules for programming elementary pair interactions in rotating magnetic-particle systems and may help guide the control of microrobotic assemblies.
Read the full abstract

All authors Dongfang Fu, Leilei Wang, Kailai Wang, Xu Zheng, Zaiyi Shen

Algebraic Geometry

Boundedness of Stein degrees in positive characteristics

In this paper, we prove Shokurov's conjecture in relative dimension $1$-case in sufficiently large characteristics. Depending on this result, we prove the boundedness of Stein degrees for horizontal boundaries in a log Calabi-Yau fibration in sufficiently large characteristics.

Differential Geometry

Rigidity of Ricci-Pinched Complete Self-Shrinkers in Arbitrary Codimension

Let $M$ be an $n$-dimensional complete connected self-shrinker in $\mathbb{R}^{n+p}$. Denote by $\operatorname{Ric}$ and $\mathbf{H}$ the Ricci curvature tensor and the mean curvature vector of $M$, respectively. We prove that if the self-shrinker $M$ satisfies $\operatorname{Ric}\ge(\frac{n-2}{n^2}+\varepsilon_n)|\mathbf{H}|^2g$, then $M$ is either a linear subspace or a round shrinking sphere, where $\varepsilon_n$ is an explicit positive constant equal to $\frac{1}{3n^2}+O\left(\frac{1}{n^5}\right)$. Moreover, we obtain a rigidity theorem that is sharp in codimension two for the self-shrinker satisfying $\operatorname{Ric}\ge\frac{n-2}{n^2}|\mathbf{H}|^2g$.
Read the full abstract

Computation and Language

ModularPhaseNet: Finite-Cyclic Phase Geometry for Computable Semantic Hierarchy, Direction, and Context Consistency in Standard Transformers

We propose ModularPhaseNet, a classical and integer-computable discretization of the continuous complex phase geometry introduced in QuantumPhaseNet. The real-valued hidden states of a standard Transformer are retained, while only an auxiliary phase channel is quantized into a cyclic subgroup G = <g> of order q | (p-1) in the multiplicative group of F_p. A continuous phase e^{i phi} is represented by z = g^a mod p; phase composition becomes group multiplication, relative phase becomes group division, conceptual hierarchy is induced by a filtration of cyclic quotients, semantic direction is represented by oriented relative group elements, and contextual consistency is measured by gauge-invariant cycle holonomy. The method introduces three components into an otherwise standard Transformer: a finite-phase encoder, a quotient-filtration hierarchy module, and a group-valued connection module. Their outputs enter self-attention as real-valued bias terms. Training uses distributions in the real group algebra or straight-through Gumbel-Softmax, whereas inference uses exact modular exponentiation and precomputed tables. No quantum hardware, complex-valued matrix multiplication, or discrete-logarithm computation is required. We prove quantization-distortion bounds, nesting of quotient-induced partitions, gauge invariance, a discrete integrability result for flat connections, and boundedness of the resulting attention output. The central empirical hypothesis is that these exact discrete invariants improve hierarchy recovery, discourse alignment, contradiction detection, and calibrated hallucination-risk prediction under a controlled compute budget. This paper reports the theory together with a pre-registered evaluation plan; the experiments described in Section 14 have not yet been carried out, and no empirical result is claimed here.
Read the full abstract

Mesoscale and Nanoscale Physics

Static and Ensemble-Dependent Thermodynamics of the Strain-Induced Parity Anomaly in Gapped Graphene

A static deformation of graphene can act on its Dirac electrons as a valley-odd magnetic field. In sublattice-gapped graphene this field makes the two valleys add in the parity-odd response rather than cancel. We derive the equilibrium thermodynamics of this effect and separate it from the finite-frequency transport response. At fixed electrochemical potential, reversing the pseudomagnetic field removes every nonzero pseudo-Landau level in the continuum theory. The remaining grand-potential difference is fixed by the spectrally asymmetric zeroth level. The static charge response is a thermally broadened plateau confined to the gap and has no metallic $m/|\mu|$ tail. Near a band edge, pseudofield reversal transfers $\mathrm{B}\ln2$ of entropy per unsplit zero-mode state in the low-temperature window. The fixed-$\mu$ heat capacity has two side lobes per edge and a universal peak $0.439229\,D_B k_{\mathrm B}$. We then formulate a definite constant-gate-voltage circuit and show that the measured sheet heat capacity depends on the electrical boundary condition. The full massive-Dirac density of states and an exact finite-field pseudo-Landau-level calculation give the same gate crossover in their common limit. At fixed carrier number, the low-temperature edge value is $-2(\ln2)^2 D_B k_{\mathrm B}$, rather than a node. A finite geometric capacitance gives a continuous and experimentally tunable interpolation. Finally, we give a trace-free triaxial strain geometry, a disorder–interaction window, and realistic calorimetric and quantum-capacitance scales. The field-reversal protocol isolates an equilibrium electromechanical anomaly without a real magnetic field.
Read the full abstract