Computation and Language
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 11 September
We present ORQA, a method for testing occupation-level knowledge in large language models. Prior methods either map abstract LLM skills to occupations via task definitions or utilize expert knowledge which is difficult to obtain at scale and expensive. ORQA complements both of these methods by connecting O*NET occupations to trusted occupation-specific websites (such as regulatory agencies, licensing bodies, professional organizations, and government publications) and converting these into source-traceable question-answer pairs. A combination of an automated pipeline and human review produces a set of high quality questions about occupations. The question set created via our method covers 116 occupations from all 21 major groups in the SOC, with 480 questions sourced from 187 different websites. Each question is designed to probe a real-world skill question that is relevant to the occupation in question. We test 15 state-of-the-art frontier and open-weight models via this method. Claude Opus 4.6, GPT-5.4 and Claude Sonnet 4.6 all perform the best at approximately 58-62% while smaller open-weight models achieve approximately 33-41% performance. Performance varies significantly across occupations. Healthcare-related occupations achieve the highest performance (78%) while Office and Administrative Support achieve approximately 40%. Performance on individual occupations (e.g. Sheet Metal Workers and Fish and Game Wardens) is essentially zero. We also find that open-ended questions and weighting by wage bill do not significantly affect the ranking of models on this benchmark. We believe that leveraging existing trusted occupation-specific information to test LLM knowledge in professional domains may be a scalable and useful method for evaluating occupation-level AI performance in the future. Results and data are available at orqabench.org.
Machine Learning
Certified AI Triage of ICU Alarms
In the VTaC benchmark 71% of ventricular-tachycardia alarms are false, but silencing a real one can delay recognition of a dangerous arrhythmia. We reframe alarm reduction as three-way triage (retain, suppress, or defer) and bound the decision this analysis treats as harmful: among suppressed alarms, the fraction that were genuine stays below a user-set budget with 95% confidence, under i.i.d. event sampling. Alarms sharing a waveform record are dependent, so the clustered analysis is a sensitivity check. On the official split a 5% budget certifies in all three seeds, suppressing 74.8% of false alarms while silencing 1.5% of genuine ones, at AUROC 0.953 and Challenge Score 83.33, numerically comparable to the strongest of the eleven published systems. Our central finding measures what multiplicity costs: the correction charges for every candidate, so a finer grid can certify strictly less. Under held-out calibration the 885-cell grid we declared certifies 1 of 15 fold-runs, while choosing the grid on a separate selection partition certifies 8. We project the calibration volume each budget needs, making an uncertifiable budget a design parameter. Finally, adding a learned reliability dimension to the policy grid did not sharpen the certified frontier.
Read the full abstract
Machine Learning
LatentVerse: A Framework for Understanding Shared and Modality-Specific Information in Multimodal Latent Representations
Latent embeddings have become a central data abstraction in modern machine learning, especially in biomedicine, where foundation models are increasingly used to encode multimodal data like clinical text, medical images, omics, and physiological signals. However, the utility and value of these representations depends on understanding their quality, structure, and the information they encode. Existing analysis workflows for evaluating representations remain fragmented across custom scripts, isolated metrics, and most importantly lack multimodal analysis, limiting accessibility and reproducibility. We present LatentVerse, a representation analysis resource that combines a web-based visual analytics platform for accessible, report-driven exploration with a command-line interface for scalable technical workflows. LatentVerse unifies diagnostics for various representation quality metrics and extends to multimodal settings by decomposing embeddings into shared and modality-specific components. We evaluate LatentVerse through controlled unimodal and multimodal simulations, discovery-oriented analyses on real biomedical embeddings, and a user study across diverse use cases. By supporting thorough and interpretable evaluation of latent spaces, LatentVerse makes foundation model representations more understandable in biomedical and data science applications.
Read the full abstract
High Energy Physics (Phenomenology)
DIS Dijet Production: An operator basis bridging eikonal and TMD regimes
We propose an operator basis for the unpolarized deep-inelastic scattering (DIS) quark-antiquark dijet production process smoothly connecting the all-twist eikonal regime to the back-to-back leading-twist TMD regime at arbitrary Bjorken x. Starting from the background-field quark propagator we construct an operator basis that organizes the eikonal and twist expansions within a unified formulation. Utilizing this operator basis we derive the dijet production amplitudes retaining all contributions required by either the leading-eikonal or the leading-twist description. In the eikonal limit the resulting amplitudes reproduce the all-twist Color Glass Condensate result, while in the back-to-back limit they reduce to the leading-twist gluon TMD-based result at arbitrary x. The operator basis can be systematically extended to include sub-eikonal and higher-twist corrections. Further, by examining the relationship of this operator basis to the improved TMD (iTMD) factorization we recover the iTMD structure in the eikonal limit. However, we find at non-zero x the transverse resummation generates longitudinal phases that prevent factorization in terms of a conventional TMD operator.
Read the full abstract
Materials Science
Observation of topological surface phonons in diamond with nonlinear optics
Topological quantum states in electronic systems have profoundly transformed the understanding of phases of matter. Recent theories predict novel vibrational topological quantum states, i.e., topological phonons in various solids. However, this paradigm is yet to be established due to the lack of convincing experimental verification of topological surface phonons—a hallmark signature of topological phonon states. Here, we report the discovery of topological surface phonons in diamond using sum-frequency spectroscopy—a nonlinear optical spectroscopy capable of probing surface phonons with high sensitivity. Diamond, known for exceptional hardness and thermal transport, is unveiled as a phonon topological semimetal hosting topological nodal-lines and nexus triple points in the bulk. With consistent theory and experiments, we uncover the resultant topological surface phonons on diamond (111) and (100) surfaces. Moreover, by chemically modifying these surfaces, we reveal the disorder effect on topological surface phonons. These findings pave the way for bridging two fundamental domains: quantum topology and lattice dynamics, besides having important implications on diamond-based devices and functional interfaces.
Read the full abstract
Classical Analysis and ODEs
Weighted Endpoint Variational and Jump Inequalites for Rough Singular Integrals
In this present paper, we study the weighted weak type $(1,\,1)$ bounds for rough maximal singular integral operators as well as the the corresponding variation and jump operators. Our first result is the weighted weak type $(1,\,1)$ bound for the maximal singular integral $$T_{\Omega}^{*}f(x)=\sup\limits_{\varepsilon>0}|T_{\varepsilon,\Omega}f(x)|=\sup\limits_{\varepsilon>0}\bigg|\int_{|x-y|>\varepsilon} \frac{\Omega(x-y)}{|x-y|^{d}}f(y)dy\bigg|,$$ where $\Omega\in L^\infty(\mathbb{S}^{d-1})$, homogeneous of degree zero, and satisfies the cancellation condition. We show that $$\|T_{\Omega}^{*}\|_{L^1(w)\rightarrow L^{1,\infty}(w)}\lesssim[w]_{A_1}[w]_{A_\infty}\log([w]_{A_\infty}+1),$$ where $w$ belongs to the Muckenhoupt class $A_1(\mathbb{R}^d)$. This represents an essential improvement of a result (Honzík, Inter.Math. Res. Not. 2020) and a result (Bhojak and Mohanty, J. Funct. Anal.2023). Our second one is the weighted weak type $(1,\,1)$ bounds for variation and jump operators corresponding to $\{T_{\varepsilon,\Omega}\}_{\varepsilon\in2^{\mathbb{Z}}}$ and $\{T_{\varepsilon, \Omega}^{\phi}\}_{\varepsilon\in\mathbb{R}^{+}}$ under the condition $\Omega\in L^\infty(\mathbb{S}^{d-1})$, where $T_{\epsilon,\Omega}^{\phi}$ represents a smooth truncation of rough singular integral operator.These results of this part are the first weighted weak type $(1,\,1)$ variation inequalities and jump inequalities for rough singular integrals.
Read the full abstract
Cryptography and Security
An Evidence-First Multi-LLM Framework for Auditable Critical-Infrastructure Dependency Modeling
Critical-infrastructure knowledge is distributed across heterogeneous, incomplete, and weakly structured evidence, making dependency models difficult to construct automatically and difficult to trust. Large language models (LLMs) can extract structured knowledge from such evidence, but direct LLM-to-graph generation risks unsupported relationships, inconsistent terminology, incorrect entity identities, and erroneous dependency endpoints. We present an evidence-first multi-LLM framework for constructing Infrastructure Knowledge Bases (IKBs) and Infrastructure Dependency Graphs (IDGs) from heterogeneous infrastructure documentation. Multiple open-weight LLMs independently extract candidate entities and dependencies from normalized evidence, after which the framework separates evidence verification, ontology grounding, entity resolution, dependency alignment, validation, fusion, and human review. Evidence support, ontology reconciliation, endpoint resolution, model agreement, and human validation remain distinct states, while provenance and unresolved cases are preserved throughout. The validated IKB is then projected deterministically into the IDG without introducing new LLM-generated knowledge. Evaluation in nine infrastructure projects shows that entity recovery achieves substantially higher recall than complete directed dependency recovery and that canonical endpoint resolution is a major constraint in dependency construction. Cross-model overlap is also much lower for dependencies than for entities, indicating that the models often produce non-overlapping candidate assertions rather than a stable majority consensus. These findings support an auditable evidence-to-IKB-to-IDG process in which uncertainty is preserved and resolved progressively rather than collapsed into a single confidence or voting decision.
Read the full abstract
Materials Science
Revealing Hidden Inversion Symmetry Breaking in ZrTe$_5$ via Phonon-Assisted Heterodyne Amplification
ZrTe$_5$ is a sensitive topological material where small perturbations can alter its electronic structure. Its equilibrium crystal structure has been widely regarded as centrosymmetric, while recent experiments have raised the possibility of inversion-symmetry breaking. Here we probe this hidden symmetry lowering using nonlinear optical spectroscopy. Although conventional second-harmonic generation does not resolve an equilibrium symmetry-breaking signal, terahertz-field-induced second-harmonic generation (TFISH) reveals it through phonon-assisted heterodyne amplification. A coherently driven infrared-active phonon acts as a local oscillator for the vanishingly weak second-order susceptibility $\chi^{(2)}$, converting an otherwise undetectable symmetry-breaking response into a phonon-frequency modulation of the TFISH signal. The field-linear scaling of this modulation demonstrates $\chi^{(2)}$ is an equilibrium susceptibility rather than a response induced by the THz field. Polarization- and temperature-dependent measurements identify a bulk polar distortion along the crystallographic $a$ axis that persists to room temperature, while the $c$ axis remains nonpolar. These results provide direct optical evidence for equilibrium inversion-symmetry breaking in bulk ZrTe$_5$ and establish a structural constraint for understanding its electronic and topological properties.
Read the full abstract
Optics
Enhanced chiral response of hybrid photonic structures empowered by Mie resonances
Chiroptical response provides a powerful route to control light-matter interactions through light handedness, with crucial importance in polarization control, chiral sensing, and nonlinear photonics. Hybrid structures may offer a novel route to engineer such responses by coupling plasmonic field localization with dielectric Mie resonances within a single nanostructure. However, in chiral metasurfaces, hybridization has largely been treated as a means of resonance enhancement, while its role in actively reshaping and reversing intrinsic plasmonic chirality remains largely unexplored. Here, we demonstrate a hybrid metasurface in which a dielectric Mie resonance mediates enhanced chiral response of plasmonic building blocks. By coupling a Si nanocylinder to trapezoidal Au nanobars, a weak and mode-selective plasmonic chiral seed is reorganised into two opposite-handed hybrid resonances, producing near-unity circular dichroism with spectral sign flipping. Extending this mechanism to nonlinear regime, the same hybrid-mode selectivity produces strongly nonlinear chiral emission and near-unity third-harmonic generation circular dichroism of either sign. We believe these results establish plasmonic-Mie hybridization as a novel mechanism for engineering high-purity, sign-switchable linear and nonlinear chirality in compact metadevices.
Read the full abstract
Functional Analysis
Nonexistence of Approximately Dual Frames via Weighted Composition Operators
Reconstruction from the coefficients of a frame often requires a dual frame, and the canonical dual is obtained by inverting the frame operator, which is rarely available in a closed form. It is therefore natural to ask whether the duality can instead be implemented by an operator of prescribed form. We study this problem in the context of weighted composition operators on Hardy and Bargmann-Fock spaces. We first characterize those weighted composition operators which carry a frame together with a dual of it to another such pair. We then prove that if the image of a frame under a weighted composition operator is itself a dual frame, then such an operator is a positive scalar multiple of the identity and the frame is tight on both Hardy and Bargmann-Fock spaces. Relaxing duality to approximate duality, we show that on the Hardy space, composition is still forced to be the identity under mild regularity for the weight function. Approximate tightness, however, fails completely. On the Bargmann-Fock space, the situation is rigid: such an approximate dual exists precisely when the ratio of the optimal frame bounds does not exceed an explicit sharp threshold determined by the prescribed error.
Read the full abstract
Machine Learning
When Connected Does Not Mean Similar: Charting the Homophily Boundary of SNAP-KG for Streaming Entity Integration
SNAP-KG is a framework for assigning newly arriving entities to semantic communities in a growing knowledge graph (KG) using only their raw features, with no graph access and no retraining at inference time. It was evaluated on five multi-view benchmarks and a 2.4M-node OGB-WikiKG2 KG. In each of these datasets, at least one graph view is homophilous, meaning that connected nodes usually belong to the same class, and SNAP-KG performs well on all of them. This paper asks what happens outside that setting. We extend the evaluation to three heterophilous graphs (Texas, Wisconsin, Chameleon) and measure the edge homophily of every view. When no homophilous view is available, clustering quality drops sharply for both SNAP-KG and the transductive baselines used in its original evaluation. What decides this is the homophily of the relation, not the number of relations. Multi-view fusion still helps, but only when at least one homophilous relation provides a reliable foundation. The homophily assumption is therefore shared by the whole method family, not specific to SNAP-KG. We argue that heterophilous multi-view clustering is a separate research problem, outside the scope of this work. As future work, we outline how a heterophily-aware teacher could be distilled into SNAP-KG's projector to serve both homophilous and heterophilous KGs.
Read the full abstract
Metric Geometry
Gromov-compactifiability of Wasserstein spaces
We characterize Gromov-compactifiability of Wasserstein spaces over Polish metric spaces for $1\leq p\leq\infty$. The space $\mathcal{P}_1(X)$ is Gromov-compactifiable if and only if $X$ is bounded and Gromov-compactifiable, whereas $\mathcal{P}_p(X)$ is Gromov-compactifiable for every $1<p<\infty$. At the other endpoint, $\mathcal{P}_\infty(X)$ is Gromov-compactifiable if and only if $X$ is. For $p=1$, the sufficiency argument combines an averaged triangle defect with tightness. For $1<p<\infty$, the main tool is the uniform convexity of $L^p$. The $p=\infty$ proof uses the finite-set criterion and a mass redistribution construction.
Read the full abstract
Computation and Language
CueMem: Cue-Guided Context Reconstruction for Long-Term Conversational Memory
Long-term conversational agents must answer user queries by recalling information from extended dialogue histories, yet directly using the full history is costly and often unreliable, while compressed memory units may lose fine-grained evidence needed for question answering. Motivated by the reconstructive view of autobiographical memory, we propose CueMem, a cue-guided framework that treats extracted memory records as retrieval cues rather than self-contained evidence and reconstructs query-relevant dialogue context from their source turns. During memory construction, CueMem extracts fine-grained memory cues from dialogue turns and links each cue to its source turn. At query time, it retrieves query-relevant cues, maps them to source-turn anchors, and expands from these anchors over a turn graph that captures temporal proximity and semantic relatedness, reconstructing a compact evidence context from the original dialogue for LLM answer generation. Experiments on LoCoMo and LongMemEval show that CueMem consistently outperforms representative long-term memory baselines. Further analyses show that graph-based context reconstruction helps recover supporting dialogue evidence while reducing query-time input tokens and latency compared with the full-history LLM setting. These results highlight retrieval cues as an effective alternative to self-contained memory evidence for long-term conversational question answering.
Read the full abstract
Computation and Language
SynthSentry: Detecting Synthetic Data Contamination in Language Model Training Data
Large language models trained recursively on their own or other models' outputs undergo model collapse, in which distributional tails and factual accuracy deteriorate while fluency survives. Prior work diagnoses collapse after training; the actionable problem is screening a corpus of unknown provenance before training. We introduce SynthSentry, a corpus-level, model-agnostic contamination signal requiring no access to the generating model, no generation history, and no synthetic labels. The score is a distributional divergence over three statistics: lexical diversity collapse, n-gram tail truncation, and perplexity variance across reference models. We evaluate on corpora contaminated by small open-weight generators and an instruction-tuned open-weight model under a leave-one-generator-out protocol. A domain-stratified study measures false positives on naturally repetitive human text (legal, clinical, source code). The score ranks corpora by severity with little loss when whole generator families are held out. Per-domain calibration holds near its nominal false-positive budget once covariance shrinkage and a bootstrap threshold replace a naive quantile, which runs four times over budget. A downstream fine-tuning check showed no contamination-driven accuracy deficit at our scale, so whether pruning recovers one remains open; the same run shows over-pruning risk once pruning exceeds the true contamination fraction. We frame screening as a data-curation defense rather than a post-hoc diagnosis and release the scoring toolkit. All results are small-scale; scope is English-language, batch-mode corpus screening. Contamination sources are single-generation or hand-authored rather than recursively generated, so results speak to synthetic contamination generally and not to recursion depth.
Read the full abstract
Data Structures and Algorithms
Spatial Mixing and Deterministic Approximate Counting of Multi-spin Systems beyond Bounded Degree Graphs
We develop a framework for deterministic approximate counting of multi-spin systems beyond bounded-degree graphs. The algorithm recursively constructs rational polytopes containing the true marginal vectors and uses linear-fractional programming to obtain certified bounds on marginal ratios. For positive interactions on graphs of polynomial connective constant $D$, we establish strong spatial mixing whenever $Dc<1$, where $c$ bounds the Birkhoff contraction coefficients of the interactions. Under the same condition, we also give a deterministic fully polynomial-time approximation scheme (FPTAS) for the partition function. We further extend the framework to proper colorings of sparse Erdős-Rényi random graphs using recursion on permissive blocks. For every fixed $\eta\in(0,1)$, sufficiently large fixed $d$, and fixed integer $q\ge(2+\eta)d$, we obtain an FPTAS for counting proper $q$-colorings of $G\sim\mathcal G(n,d/n)$ with high probability over $G$. This improves the leading constant $3$ in the earlier counting guarantee of Yin and Zhang (APPROX/RANDOM, 2016) to $2$, and asymptotically matches the spatial mixing regime established by Yin (ICALP, 2014).
Read the full abstract
Superconductivity
Emergent magnetism, heavy electrons and pressure-induced reentrant superconductivity in the iron substituted 4d transition-metal sulfides
Superconductivity emerging from or in the vicinity of magnetic states is generally considered to be mediated by spin fluctuations and thus lies beyond the scope of conventional electron-phonon coupled BCS framework. Here we report the emergence of novel ferromagnetism in the d-electron rhodium sulfide Rh17S15 superconductor, characterized by an enhanced Sommerfeld coefficient {\gamma} arising from the flat topological band and many-body correlations. We further demonstrate that the ferromagnetism can be tuned via Fe substitution at the Rh sites, leading to a spin glass ground state induced by the competing ferromagnetic and antiferromagnetic exchange interactions. Fe doping results in a further enhancement of both the {\gamma} and electron effective masses. At a doping level of x = 0.67 in Rh17-xFexS15, {\gamma} reaches 312 mJ mol-1K-2, second only to the well-documented d-electron heavy-fermion material LiV2O4. Furthermore, upon applying pressure, superconductivity is first suppressed; under high pressures, however, we observe the reentrant superconductivity in both pristine and Fe-doped samples. Our results not only demonstrate the unusual magnetic states and possible heavy-fermion features in these frustration-free, d-based superconductors, but also suggest that the superconductivity in this system is likely mediated by the intrinsic spin fluctuations and may thus be unconventional.
Read the full abstract
Computer Vision
EgoMaize: A First-Person Maize Instance Segmentation Benchmark under Severe Field Occlusion
Close-range first-person field images are important for mobile maize phenotyping because many plant-level traits depend on in-canopy structures that are difficult to ob serve from overhead views. However, post-seedling maize fields create a difficult in stance segmentation setting: stems, leaves, tassels, and neighboring plants are elon gated, repetitive, and strongly occluded. We introduce EgoMaize, a compact benchmark for first-person maize instance segmentation, where the task is to predict ownership consistent plant masks and plant-owned stem/tassel cues from close-range field images with severe same-class overlap. Existing visible-only labels can fragment one physi cal plant into disconnected supervision, while full-amodal labels may require unverifi able completion behind neighboring plants or field objects. EgoMaize therefore uses an evidence-closed annotation workflow for occluded maize regions and assigns unreli able maize regions to ignore rather than background. Baseline results show that pre trained query-based grouping, boundary refinement, and high-resolution crop refine ment help different aspects of the task, but no architecture solves the coupled chal lenges of fine structure recovery, same-class instance ownership, and occlusion reason ing; occlusion-level analysis further shows that performance decreases as plant visi bility becomes more limited. The dataset and code are publicly available at https: //github.com/JaaaaaaaD/EgoMaize.
Read the full abstract
Robotics
A Deployable Architecture for Robot-Mediated Tasks (DART): Evaluation in Socially Assistive Robot-Guided Cognitive Behavioral Therapy Exercises
Socially assistive robots (SARs) can support structured health and well-being interventions, but hardware and cost constraints limit interaction complexity and longitudinal real-world deployments. We present DART: Deployable Architecture for Robot-Mediated Tasks, an architecture that extends SARs through a web application and cloud infrastructure, enabling visual content, user input, remote computation, and persistent data storage synergistically with the robot's physical embodiment, speech, and movement. We evaluated DART by instantiating it in an interatively-developed full-stack HRI system for helping university students with elevated generalized anxiety to complete cognitive behavioral therapy (CBT) homework exercises. The resulting system, which used the low-cost open-source Blossom robot platform, was refined and evaluated through a participatory design process and multiple user studies, and finally evaluated in an in-lab study with 103 participants, and then a six-week in-home deployment with four participants. In the in-lab evaluation, participants showed significant within-session reductions in stress, state anxiety, and negative affect, and gave the platform a mean System Usability Scale score of 78.89. In the home deployment, the mean System Usability Scale score was 87.5, with positive qualitative feedback on usability. Participants across both groups identified speech input, visual presentation, and web-robot synchronization as priorities for improvement. These findings validate DART as an effective architecture for extending the capabilities of a low-cost SAR in both in-lab single-session and in real-world longitudinal deployments.
Read the full abstract
Systems and Control
e-Traceroute: Physically traceable electricity routing for carbon-free energy utilization
Maximizing the self-consumption of residential photovoltaic generation is a promising pathway to meet urgent decarbonization targets for 2030 (and even for 2035). A viable solution is to create a sharing economy for idle battery capacity within a community. To achieve this, utilizing shared physical assets necessitates complete physical traceability of power flows—the capability to strictly trace the ownership of stored energy among multiple participants. Furthermore, physical traceability is an essential function for demonstrating the use of carbon-free energy resources. Such tracing is impossible in conventional systems owing to two fundamental limitations: the mixing of power flows in a common bus and the decoupling of power and information delivery. This study presents a novel physical-layer technology, called e-Traceroute, that overcomes these limitations and realizes physically traceable electricity exchanges. Specifically, physically distinguishable power routing and data transmission are unified over the same power lines. Prototyping experiments demonstrate successful integration of power transfers and information transactions, validating that the proposed system serves as a physical foundation for a reliable and scalable sharing economy that drives bottom-up decarbonization.
Read the full abstract
Robotics
DWMP: Leveraging Dual World Models for Humanoid Obstacle Traversal
Humanoid robots must traverse cluttered obstacle fields using onboard proprioceptive and visual observations, yet existing methods usually process multimodal observations without explicitly considering their different characteristics: proprioceptive observations are low-dimensional but governed by highly nonlinear robot dynamics, while egocentric visual observations are high-dimensional, noisy, and redundant. We propose DWMP (Dual World Model Policy), a framework that provides the actor with separate but complementary world-model representations for humanoid obstacle traversal. A Koopman-based dynamics world model lifts proprioceptive observations into a latent space where their temporal evolution is approximately linear, making the dynamics features easier for the actor to learn from. An RSSM-based visual world model compresses egocentric depth observations into compact stochastic states while preserving obstacle-related geometry. The student policy receives the fused latent representation for action generation, combining linearized proprioceptive dynamics with compressed visual perception. Experiments in simulation and on a Unitree G1 humanoid robot show that DWMP improves obstacle traversal performance over baselines and supports real-world deployment under randomized obstacle layouts.
Read the full abstract
Quantum Physics
The Low-Individual-Degree Test Without the Diagonal-Lines Test Is Not Quantum-Sound
To prove the quantum soundness of the classical low-individual-degree test, the authors of \cite{JNVWY20LID} defined three subtests, namely the axis-parallel lines test, the self-consistency test, and the diagonal-lines test. An interesting question is whether the diagonal-lines test can be removed. In this paper, we show that the diagonal-lines test cannot simply be removed without another compatibility mechanism. Consequently, replacing the "conditional linear functions" by "coordinate deletion functions" in the proof of MIP*=RE, as mentioned in \cite{JNVWY20LID}, does not by itself preserve the required soundness. The authors of \cite{JNVWY20LID} found an example that requires the diagonal-lines test when \((m, d, q) = (2, 2, 4)\); we give an example when \((m, d) = (2, 2)\) and \(q\) is any odd prime.
Read the full abstract
Machine Learning
ParaRecover: A Process-Level Benchmark for Error Localization and Recovery in Parallel Tool-Use Agents
Existing agent benchmarks mainly evaluate final task success or tool-call correctness, providing limited insight into whether agents can reliably diagnose and recover from intermediate execution failures. This limitation becomes particularly critical in multi-turn parallel tool-use scenarios, where errors may propagate across dependent branches and trigger cascading failures. We introduce ParaRecover, a process-level benchmark for evaluating error localization and recovery in multi-turn parallel tool-use agents. Built upon a fine-grained taxonomy of 14 error types covering planning dependencies, tool selection, and argument matching, the benchmark comprises 10,626 instances spanning two difficulty levels. To enable finegrained, process-oriented evaluation, we further propose the SDE rubric, which measures structural integrity, diagnostic reasoning, and evolutionary strategy during agent execution.Experiments across more than ten mainstream LLMs reveal that even state-of-the-art models still struggle with multi-turn error propagation,implicit tool-use failures, and precise replanning. Moreover, we demonstrate that the SDE rubric provides effective supervision signals for improving agents' reflective recovery capabilities. Our data and code are available at https://github.com/gbw206/ParaRecover.
Read the full abstract
Applied Physics
Convective Heat Transfer Optimization for Liquid Cooling Plates Driven by Field Synergy and Fractal Geometry
The rapid development of liquid-cooled data centers has imposed imperative demands on the performance of liquid cooling plate. The density-based topology optimization (TO) is an effective approach to resolving the growing thermal-hydraulic performance requirements of liquid cooling plate. However, existing TO methods can hardly optimize convective heat transfer directly which is the intrinsic heat transfer mechanism, due to the highly complex and evolving structural topologies, varying flow and temperature fields, making it extremely challenging to explicitly describe the heat transfer coefficient and heat transfer area during TO process. A convective heat transfer topology optimization (CTO) method is proposed in this study, where the iteratively evolving heat transfer coefficient is explicitly depicted by the field synergy theory in the thermal objective, and directly described by the velocity and temperature fields without relying on specific geometry. Combined with the explicit depiction of heat transfer area by the fractal geometry theory, a CTO framework is built for a direct optimization of convective heat transfer under both the laminar and turbulent flow conditions. The CTO tends to generate more hierarchical and directional structural topologies in optimization results, which is conducive to reducing low-velocity stagnation zones and improving flow direction in branched channels, achieving enhanced synergy and thermal-hydraulic performance in the optimized liquid cooling plates. Compared with the TO results without incorporation of field synergy theory, the CTO can reduce average temperature rise by 20% while improving the Nusselt number by 15% under laminar flow conditions, and reduce maximum temperature rise by 10.2% and pressure drop by 25% under turbulent flow conditions.
Read the full abstract
Computer Vision
VS-Splat: Voxel-Selective feed-forward Gaussian Splatting for end-to-end 3D object reconstruction from sparse-views
Feed-forward Gaussian splatting models have demonstrated remarkable effectiveness in reconstructing three-dimensional (3D) objects from a few two-dimensional (2D) images, even if they are unseen. As existing methods typically predict Gaussian primitives uniformly across the 3D space, most primitives are placed in non-object regions. This may hinder the representation of fine object details. This paper proposes a Voxel-Selective Gaussian Splatting model (VS-Splat), a new end-to-endfeed-forward Gaussian splatting framework that predicts many primitives only within selected voxels that are likely to belong to an object, without 3D structural supervision. To achieve this, we propose a new learnable voxel selection approach that identifies object-centric voxels only with 2D rendering supervision. Our sparse-view rendering experiments with three benchmark datasets show that proposed VS-Splat outperforms several state-of-the-art methods. We further demonstrate its effectiveness as a backbone for an existing densification method and show that anoptional extension improves its robustness to inaccurate camera pose estimates.
Read the full abstract
Computer Vision
UniMo: Unifying Human and Animal Motion Generation
The conditional generation of 3D motion has emerged as a key research topic due to its wide applicability across robotics, AR/VR, gaming, and content creation. However, extending recent advances in text-driven human motion generation to the animal domain remains challenging due to two core limitations. First, animals exhibit highly diverse skeletal topologies, unlike the standard human structure, making unified modeling across species difficult and leading to inefficient per-species models. Second, existing animal motion datasets suffer from limited scale and annotation quality, constraining model performance. To address these challenges, we propose UniMo, a unified point cloud-based motion generation framework that bypasses topological discrepancies by converting parametric skeletons into unparametric representations, further enhanced by dynamic sampling that allocates more points to active joints. Additionally, we present UniML3D, a large-scale motion-language dataset spanning both human and animal categories, containing 145,907 motion sequences and 433,388 captions-over 102x larger than existing animal datasets. Our method achieves state-of-the-art results on UniML3D and three public benchmarks including HumanML3D, KIT-ML, and AnimalML3D, demonstrating the feasibility and effectiveness of unified human-animal motion generation. Website: https://steve-zeyu-zhang.github.io/UniMo.
Read the full abstract
Computation and Language
I Am No One: Style-Aware Paraphrasing for Text Anonymization
Authorship attribution models can re-identify users from seemingly anonymized text by exploiting stable stylistic fingerprints, even after explicit identifiers are removed, posing a growing privacy risk for text publishing and analytics. This risk extends to speech-derived text such as ASR transcripts of meetings and call-center conversations, where stylometric leakage can persist even after acoustic anonymization. Differential privacy-based anonymization often severely degrades text quality and utility. We propose a style-aware, prompt-driven anonymization approach that uses pretrained large language models to construct compact stylistic profiles from minimal samples and rewrite text to suppress identifiable style markers while preserving meaning. Across blog and review datasets, our approach reduces authorship attribution F1 by 60-70% while maintaining content quality and readability, substantially outperforming DP-based and non-DP baselines.
Read the full abstract
Combinatorics
Moment obstructions and continuum-to-discrete bounds for checkerboard no-three-in-line sets
Fix one colour class in the checkerboard colouring of an $n\times n$ integer grid, and let $M_4(n,\varepsilon)$ be the largest subset having at most two points in every row, column, and diagonal of slopes $\pm1$. We prove the near-saturation bound $M_4(n,\varepsilon)\leq2n-4$ for $n\geq6$. The proof uses first and second moments of the four line families: a hypothetical set of size $2n-3$ produces row, column, and diagonal deficits whose exact moment identities contradict Cauchy–Schwarz. A finite argument handles $n=6$. The same identity extends to arbitrary deficit multisets. It gives $M_4(n,\varepsilon)\leq2n-d$ whenever $d\geq4$ is an integer and $n\geq3d-4$, and an entirely discrete asymptotic estimate \[ M_4(n,\varepsilon)\leq(\sqrt{21}-3)n+8. \] We also prove a general continuum-to-discrete theorem for the associated four-direction fractional packing problem. Applying it to the exact continuum dual certificate constructed in earlier work yields, for both colours, \[ L_{\mathrm{mono}}(n,\varepsilon)\leq\alpha n+O(1), \qquad \alpha\approx1.5768233968738, \] and hence the same upper bound for $M_4$ and for checkerboard no-three-in-line sets.
Read the full abstract
Astrophysics of Galaxies
Stellar bar occurrence in TNG50 and TNG100: disentangling environmental effects and assembly history at z = 0
In this work, we investigate the dependence of bar presence on environment using TNG50 and TNG100 simulations at z=0. We select disc galaxies within $2 R_{200}$, with stellar mass M$_{\star} \geq 10^{10}$ M$_{\odot}$ (TNG100) and M$_{\star} \geq 10^{9}$ M$_{\odot}$ (TNG50). With these criteria, our final samples consist of 1719 and 859 satellite galaxies in TNG100 and TNG50, respectively. The environment is characterised by host halo mass: massive (M$_{\rm h} \geq 10^{14}$ M$_{\odot}$), intermediate ($10^{13} < \rm M_{\rm h} < 10^{14}$ M$_{\odot}$), and low-mass haloes (M$_{\rm h} \leq 10^{13}$ M$_{\odot}$). We find that our samples recover the morphology-density relation, with thin disc fractions increasing with cluster-centric distance and decreasing halo mass. The bar fraction and the bar radius normalized by the effective radius decrease mildly with cluster-centric distance, while absolute bar radii remain constant. At fixed distance, the bar fraction increases with halo mass, partly because more massive galaxies reside in more massive haloes. In addition, we find that galaxies with earlier assembly times show a higher bar fraction, especially within $0.5 R_{200}$, suggesting a connection between earlier assembly and a higher probability of hosting a bar. Overall, we find a mild environmental dependence, with a stronger connection to stellar mass and assembly history.
Read the full abstract
Number Theory
Critical convergence and Hausdorff measures for generalized Flint Hills series
We study the generalized Flint Hills series $\mathcal{F}_{s,t}(x)=\sum_{n\ge1} n^{-s}|\sin(\pi nx)|^{-t}$ for $s>0$ and $t>1$. An explicit comparison with a series over continued-fraction denominators yields the Hausdorff dimension $\min\{1,2t/(s+t)\}$ of its divergence set. At each critical exponent $\tau=1+s/t>2$ we construct numbers of irrationality exponent $\tau$ realizing both convergence and divergence; the convergent examples establish Meiburg's conjecture in the range $t>1$. Both parts of the critical fibre have Hausdorff dimension $2/\tau$. For $s>t$ and $h_\kappa(r)=r^{2/\tau}(\log(1/r))^\kappa$ we prove that the divergence set has zero $h_\kappa$-measure for $\kappa<-1$ and infinite measure for $\kappa\ge-1$. The divergent part of the critical fibre satisfies the same law, whereas its convergent part has infinite measure for every $\kappa$. The key estimate selects a rapidly growing subsequence of convergent denominators and gives a double-logarithmic bound on the approximation error. The convergence of the classical Flint Hills series $\sum_{n\ge1}(n^3\sin^2 n)^{-1}$ remains undecided.
Read the full abstract
Machine Learning
Theoretical Guarantees for One-Shot Magnitude Pruning and Compute-Adaptive Early Exit
We study compute reduction in neural networks through a unified partial versus full computation view, captured by one-shot magnitude pruning in the static regime and early exit in the adaptive regime. In an asymptotic single-neuron model, we prove a concentration theorem for one-shot magnitude pruning with explicit rates. We also introduce the conditional perceptron for early exit and show that its excess generalization error decays as a power of the compute gap, with an exponent that grows to infinity as the alignment between partial and full computations tends to one. We then extend the analysis to deep networks, characterizing how pruning-induced distortions accumulate with depth and deriving a corresponding compute-accuracy tradeoff for frozen-backbone early exit under a neural network Gaussian process model. Numerical simulations corroborate the predicted scaling laws.