Peer Revenue Returns subject to peer review

Papers

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

All of arXiv

Monday 31 August

Differential Geometry

Dimensional reduction associated with unchanged direction trajectories in flat fiber Robertson Walker spacetimes

In this article, we establish a new dimensional reduction result in (Galilean or Lorentzian) Robertson Walker spacetimes with Euclidean fiber. In this context, we prove that every unchanged direction observer can be embedded into a three dimensional totally geodesic timelike submanifold.
Read the full abstract

Computation and Language

Bridging Lexical Divergence: LLM-Assisted, Cost-Efficient, Zero-shot Scientific Entity Linking

Scientific domain entity linking (EL) differs from general domain EL because mentions and entity names often lack lexical overlap. Another challenge is that specialized terminology is used in the scientific domain, which is rarely encountered in models pretrained on general domains. Therefore, models trained on general domains transfer poorly to scientific domains. To address this, in-domain fine-tuning is the natural remedy. However, many scientific domains lack expert-annotated data, motivating the need for a zero-human-annotation approach. Existing zero-shot methods heavily rely on LLMs to generate aliases across entire mention corpora, which incurs substantial computational cost, and those methods provide no mechanism to filter out noise from LLMs. To address these challenges, we propose Sci-ZSEL, a framework that selectively generates entity aliases with an LLM to control computational cost, and applies an ontology-aware filter to remove aliases that semantically drift toward ontology neighbors. Then, filtered aliases are used to construct pseudo-labeled mention-entity pairs for fine-tuning. To enable evaluation of EL under low lexical overlap, we also release a new animal science EL benchmark linked to three livestock trait ontologies, where mentions and entities exhibit substantially lower lexical overlap than in existing benchmarks. Across five benchmarks, Sci-ZSEL outperforms the non-fine-tuned baseline, is most useful on nonoverlapping mentions, and combining it with curated synonyms gives the best performance in most settings.
Read the full abstract

All authors Md Rasel Khondokar, Qiao Qiao, Farjana Sultana Samia, Nhat Le, Yuepei Li, Qi Li

Software Engineering

Don't Let the Model Write the YAML: Deterministic, Minimal-Diff GitOps Remediation from LLM-Proposed Field Changes

LLM agents increasingly diagnose incidents and propose remediations. In a GitOps workflow, applying a fix means editing a version-controlled config file, and the obvious implementation, having the model author the edited file or a diff, is what practitioners reach for first. Evaluating that choice on real Kubernetes manifests, we find no text-generation strategy is safe for unattended automation. Unified diffs are unsafe: under strict patching almost none apply, but that is an artifact, since a tolerant tool (GNU patch) applies 96%, yet silently misapplies about 1 in 7 (14-20%) with no error signal. Full-file rewrite is capability-dependent: a small model corrupts the file, while a frontier model is usually correct but non-deterministic (it silently drops a field or edits a neighbor on some runs) and must regenerate the whole file, costing O(file size) per edit. We present an alternative that separates the semantic decision (which resource, field, and value) from the syntactic act of editing the file. The agent emits only a structured field-change intent; a deterministic pipeline indexes manifests by (kind, name), locates the target scalar's exact character span via the YAML parser's node position marks, and replaces only that span in the raw text. Because the file is never re-serialized, the diff is minimal by construction, formatting and comments are preserved, and the edit is correct and deterministic independent of the model, at O(1) generation cost. The contribution is the pairing of an LLM-proposed intent with a deterministic, fail-closed application contract for GitOps. We implement it in KubeAstra (Apache-2.0) and release the benchmark. Our claim is scoped to faithful application of a known change; whether the change is right is left to human PR review.
Read the full abstract

Artificial Intelligence

ConvDeck: Conversational Paper-to-Slide Generation via Stage-Specific User Feedback

Automatic academic paper-to-slide generation is inherently iterative, because creating an effective presentation requires repeated cycles of generation, critique, and revision. Recent multi-agent systems partially acknowledge this through internal critique-and-revise loops, while conversational approaches allow users to refine generated slide decks through dialog. However, these refinement processes either remain largely closed to the user or introduce feedback only after a complete deck has been produced, limiting the user's ability to participate in the iterative refinement of narrative flow, content allocation, and presentation emphasis. To address this gap, we introduce ConvDeck, a multi-agent pipeline for conversational paper-to-slide generation that distributes interaction across the pipeline through stage-specific loops, allowing users to iteratively refine both the presentation outline and the final slide deck at the stages where each kind of decision is made. These loops are driven by a refinement mechanism in which agents can think, speak, and act, enabling them to either directly apply edits or respond conversationally to clarify user feedback and discuss revision options. Our evaluation shows that stage-specific conversational feedback improves user-goal satisfaction while preserving narrative coherence, content quality, and visual presentation.
Read the full abstract

All authors Tarik Can Ozden, Sachidanand VS, Furkan Horoz, Ozgur Kara, Dilek Hakkani-Tür, Junho Kim, James Matthew Rehg

Solar and Stellar Astrophysics

A Probabilistic Framework for Incorporating Helioseismic Far-Side Active Regions into SFT Models

Surface Flux Transport (SFT) models are routinely used to model the Sun's photospheric magnetic field and provide inner boundary conditions for coronal and heliospheric models, yet they remain fundamentally limited by the lack of direct information about flux emergence on the far side of the Sun. To address this, we develop a framework for incorporating helioseismic images of far-side active regions (HIFAR), into the Advective Flux Transport (AFT) model. Using near-simultaneous magnetic flux maps (HIFARM) inferred from HIFAR, and STEREO/EUVI 304 A observations from 2010 May 13 to 2014 August 18, we develop a logistic regression model to estimate the probability that a helioseismic detection corresponds to an AR based on descriptors of magnetic flux, field strength, and location. The model achieves a precision of 0.93 for active region (AR) detections at a probability threshold of 0.73, selected to reject 90% of "ghost" ARs. We also derive an empirical scaling relationship between HIFARM and AFT flux to place HIFARM on the AFT flux scale and demonstrate that the framework reproduces the flux evolution of ARs consistent with near-side observations over multiple solar rotations. At the global scale, we find that conventional near-side-driven AFT simulations underestimate the total unsigned magnetic flux by ~10-20% of the Sun's total magnetic flux budget. The probabilistic AR detection model and the complete framework developed in this work provide a practical pathway for incorporating helioseismic far-side ARs into SFT models, a step toward realistic modeling of full-Sun magnetic field.
Read the full abstract

All authors Bibhuti Kumar Jha, Lisa A. Upton, K. D. Leka, Ruizhu Chen, Junwei Zhao, Shea Hess Webber, Ignacio Ugarte-Urra, Kiran Jain

Machine Learning

QTEA: Ternary LLMs with Sparse Residual Salient Weight and By-Column Optimization

Weight-only post-training quantization (PTQ) can alleviate the computational burden of serving large language models (LLMs) at scale. However, existing PTQ methods often fail to generalize across models and suffer severe accuracy loss below 2 bits. Many leverage unstructured sparsity to mitigate this loss, but at the cost of regularity and GPU-friendly execution. We present QTEA, a sub-2-bit PTQ framework that quantizes weights into ternary values and uses salient weights as residual error compensators. To maintain hardware efficiency, residuals are assigned to selected columns with semi-structured $1:4$ sparsity within the salient columns. We further add column-wise rescale refinement to GPTQ-style column-by-column quantization, alternately updating per-column scales and ternary assignments to reduce reconstruction error. We also identify order-dependent error propagation in GPTQ and introduce error decay to attenuate late-stage error accumulation. On Qwen3-14B, QTEA compresses all weights to an effective 1.7 bits per weight while improving average accuracy over the strongest ternary PTQ baseline by 16.7%. It also achieves 1.40$\times$ and 2.61$\times$ lower perplexity on WikiText and C4 respectively. This trend holds on Llama3-8B, where QTEA obtains a 6.6% accuracy gain and 1.34$\times$ / 1.95$\times$ lower perplexity on the same datasets. Finally, we develop a lookup-table based kernel that achieves 7.2$\times$ faster per-token generation over an FP16 baseline. Code is available at https://github.com/Intelligent-Microsystems-Lab/QTEA.
Read the full abstract

All authors Yipin Guo, Arun M George, Jie Fu, Tareq Mahmoud, Sixue Xing, Siddharth Joshi

Optics

Modeling and Optimization of Cascaded Four-Wave Mixing toward the VUV in Xenon-Filled Negative-Curvature Fibers

In this work, we thoroughly characterize third- and fourth-harmonic generation through cascaded four-wave mixing in a xenon-filled negative-curvature hollow-core fiber operated under a pressure gradient. The experimental results are supported by both a semi-analytical model and numerical simulations, which accurately reproduce the pressure dependence of the harmonic conversion. We identify the optimum pressure conditions for third and fourth-harmonic generation and experimentally observe the cascaded four-wave-mixing process extending into the vacuum ultraviolet (VUV), up to the sixth harmonic at 172 nm, with the accessible spectral range limited by the available VUV filter.
Read the full abstract

All authors Jose Otavio Rosa, Frédéric Gérôme, Fetah Benabid, Jonas H. Osório, Jonathas de Paula Siqueira

Computation and Language

LLM-as-a-Demographic: Whom Sociodemographic Prompting Helps, and Whom It Hurts

Large language models (LLMs) are increasingly used as judges for subjective tasks, where annotators disagree and the relevant question is not only how accurate a judge is, but whose judgments it reproduces. Sociodemographic prompting conditions the judge on an annotator's demographic profile to align its judgments with the corresponding group's. We test whether this alignment emerges distributionally, comparing the predicted label distributions of 23 open-weight LLMs on three subjective tasks against those of real annotator groups, under three conditions: no demographic information, single-attribute profiles, and intersectional profiles over gender, age, race, and education. Three findings emerge. First, a judge prompted with no demographics is not perspective-neutral: models best reproduce the judgments of White, college-educated annotators. Second, demographic conditioning is asymmetric: it moves the judge toward majority groups and away from minority groups, most strongly on offensiveness, where intersectional profiles amplify the harm. Third, by comparing base and instruct models we identify instruction-tuning as a possible source of the asymmetry. Demographic conditioning should therefore be used with caution to estimate group judgments: conditioning moves predictions away from the reference distributions of the minority groups the method is often invoked to serve.
Read the full abstract

High Energy Physics (Phenomenology)

Radiative Corrections and Reheating in Supergravity Models of Inflation

We consider the effects of radiative corrections in Starobinsky-like models of inflation, concentrating on models of inflation formulated in N=1 no-scale supergravity. Inflaton couplings to matter fields are necessary for reheating and these have an impact on loop corrections to the inflaton potential. Whilst corrections due to the supergravity couplings of the inflaton to Standard Model (MSSM) fields are negligible, we use Planck data to obtain interesting constraints on GUT bilinear couplings, vevs, and gauge boson masses that could be sharpened by future CMB measurements.
Read the full abstract

All authors John Ellis, Tony Gherghetta, Kunio Kaneta, Wenqi Ke, Keith A. Olive

Quantum Physics

Quantum Secret Sharing with a Helper and Programmable Access Structures

Quantum secret sharing (QSS) is a process in which the state of a quantum system is partitioned into multiple shares, allowing only specific subsets of shareholders to reconstruct the state, while others gain no information about its identity. In this work, we propose a variant of QSS, called “helper QSS,” which designates a special shareholder whose participation with any other group of parties is sufficient for recovering the state. We present different families of helper codes that function for an arbitrary number of parties and system sizes. As an application, we introduce the idea of a programmable access structure in QSS, which allows for a third-party programmer to independently choose the access structure of the code even after the shares are delivered to all the shareholders. This choice is made blindly, meaning that the programmer has no information about the secret being encoded, and it is implemented using the nonlocal process of quantum steering.
Read the full abstract

All authors Eric Chitambar, Sarah Hagen, David W. Kribs, Mike I. Nelson, Andrew Nemec

Computational Complexity

A Dichotomy for Complex Boolean Holant with Binary Disequality

We classify Boolean Holant problems with a supplied binary disequality signature and an arbitrary fixed finite set of algebraic complex-valued signatures. Each problem is polynomial-time computable or $\#\mathsf{P}$-hard, and the tractability criterion is decidable. The proof combines arity reduction with the growth of the available binary transfer group and uses stable pairs to recover the structure of higher-arity signatures. We also prove a dichotomy for the disequality-edge model when the signature set contains an endpoint-nondegenerate irreducible eight-vertex signature, resolving the corresponding arity-four cases in the main proof.
Read the full abstract

Systems and Control

Uncertainty-Aware Parameter Estimation for Condition Monitoring of Power Converters

Changes in circuit parameters can indicate component degradation in power converters, but measurement noise and limited parameter observability can make these changes difficult to detect reliably. This paper proposes an uncertainty-aware parameter estimation framework to distinguish genuine parameter changes from variations caused by estimation uncertainty. Using a differentiable switched converter model, the method requires only one or two samples per switching period. Model-derived residual whitening and a Bayesian posterior approximation quantify estimation uncertainty and reveal poorly identifiable parameters. Simulations demonstrate improved estimation accuracy over least-squares and physics-informed neural-network benchmarks, with uncertainty estimates consistent with repeated noise realizations. Experiments on a laboratory buck converter validate the estimated parameters against independently derived references and demonstrate the detection of a controlled output-capacitance reduction. The proposed framework provides a basis for reliable parameter-change detection in power converter condition monitoring.
Read the full abstract

Dynamical Systems

Dual mixing and quasi-mixing of inner function restrictions

Dual quasi-mixing and dual mixing of a measure preserving transformation are uniform individual ratio limit properties of the transfer operator which entail the quasi-mixing and mixing properties of Krickeberg (respectively). The real restrictions of normalised, parabolic, non-Möbius inner functions of the upper half plane preserve Lebesgue measure and are dual quasi mixing; such a restriction being dual mixing iff it is exact and singular dual quasi-mixing otherwise.
Read the full abstract

Numerical Analysis

Solving the Incompressible Navier-Stokes Equations on Oriented Curved Surfaces Discretized by Point Clouds

We present a meshfree numerical solver for the incompressible Navier-Stokes equations on oriented curved surfaces that are represented by surface point clouds. On curved surfaces, numerical challenges pertaining to stiffness and pressure-velocity coupling are exacerbated. Moreover, vector calculus on curved surfaces differs from its Euclidean counterpart. The presented method operates on surface point clouds in an Eulerian frame of reference without requiring a computational grid or mesh. It achieves consistent approximation in space and time with high order of accuracy; we demonstrate up to order six. The incompressibility constraint is locally imposed as a weak artificial compressibility approximation, avoiding global matrix inversion. We show that the method provides consistent and convergent approximations of surface vector fields and differential operators. We study the relationship between error, spatial resolution, and artificial Mach number and characterize the frequency spectrum of the artificial oscillations. We provide numerical solutions of the incompressible Navier-Stokes equations on symmetric surfaces, such as the sphere and torus, and on parametric and non-parametric asymmetric surfaces. Since the proposed method works directly on unstructured surface point clouds, it provides a promising approach for simulations on image-derived geometries, such as in biological morphogenesis from microscopy videos.
Read the full abstract

Functional Analysis

Hom–Lie Algebras and Explicit MSS Partition Bounds for $(6,3)$ Biangular Frames

We study the algebraic structure of $(6,3)$ biangular Parseval frames. The two-distance property yields adjacency matrices $A_1,A_2$ whose span forms a three-dimensional commutative algebra, and adjoining the commutator $[A_1,A_2]$ produces a three-dimensional Lie algebra $\g$. The Gram matrix $G = I + c_1 A_1 + c_2 A_2$ induces a derivation $\alpha(X) = [G,X]$ on $\g$, equipping $(\g, [\cdot,\cdot], \alpha)$ with a Hom–Lie algebra structure. We compute all structure constants explicitly in terms of the strongly regular graph parameters $(k_1,\lambda_1,\mu_1,k_2,\lambda_2,\mu_2)$ and the frame angles $(c_1,c_2)$, and use this framework to derive explicit bounds for the partial frame operators arising in Marcus–Spielman–Srivastava (MSS) partitions. These bounds depend directly on the structure constants and refine the universal MSS estimate in the regime of highly unbalanced partitions.
Read the full abstract

High Energy Physics (Theory)

Anomaly Matching between Topological Superconductors and $\mathcal{N}=8$ Supergravity

The boundary of a $3{+}1$-dimensional topological superconductor carries a $\mathbb{Z}_{16}$-valued global (Dai-Freed) anomaly, famously matched by the $16$ fermions of a Standard Model generation once a certain $\mathbb{Z}_4$ symmetry is gauged. In a supergravity completion, though, a gravitino contributes $-7/16$ rather than $\pm1/16$, obstructing this $\mathbb{Z}_4$ in the Minimal Supersymmetric Standard Model. We observe that four-dimensional $\mathcal{N}=8$ supergravity evades this fate. Its fermions, eight gravitini in the $\mathbf 8$ and fifty-six spin-$\tfrac12$ states in the $\mathbf{56}=\Lambda^3\mathbf 8$ of the $SU(8)$ R-symmetry, cancel the anomaly. The cancellation holds for every admissible $\mathbb{Z}_4$ structure, is consistent with the vanishing of the continuous $SU(8)$ anomaly, and follows from a Smith homomorphism and the Rarita-Schwinger ghost-subtraction weight identifying the supergravity anomaly with that of the topological superconductor. In maximal supergravity this $\mathbb{Z}_4$ stays anomaly-free and gaugeable despite its eight gravitini. All numbers are reproduced by an open-source library described in the Supplemental Material.
Read the full abstract

Computation and Language

Uncovering and Mitigating Aggregation-Induced Reward Hacking in Multi-Reward Reinforcement Learning

Reinforcement learning fine-tuning of large language models increasingly adopts multiple reward dimensions, including verifiable rules, task-specific evaluators, and learned reward models, to provide richer supervision across diverse capabilities. These dimensions are commonly scalarized with fixed aggregation weights. We identify a failure mode in which aggregation itself induces reward hacking: static projection aliases qualitatively different reward profiles into a single scalar, steering optimization toward whichever dimensions are easiest, densest, or systematically favored by the reward signal. Over training, this traps the policy in suboptimal profiles and prevents convergence to better-balanced ones that would yield higher task performance. To address this, we propose Adaptive Multi-Reward Projection (AMRP), a lightweight online method that reallocates aggregation weights using three signals, relative shortfall, reward volatility, and recent progress, increasing pressure on lagging, unstable, or stagnant dimensions while relieving saturated ones. Across structured reasoning, citation-grounded generation, and open-ended alignment under GRPO, AMRP consistently improves reward-profile balance and downstream performance over fixed and dynamic weighting baselines; it also remains effective with GDPO and PPO, supporting compatibility across RL algorithms. Our code is available at https://github.com/yyhappier/AMRP.git.
Read the full abstract

All authors Yu Yuan, Yaoyou Fan, Lili Zhao, Guangting Zheng, Kai Zhang, Lu Pan, Ke Zeng, Qi Liu

Materials Science

Cryogenic Enhancement of Electron Spin Polarization from a Strained GaAs/GaAsP Superlattice Photocathode

We report electron spin polarization of 95.0 +/- 0.8 (stat) +/- 2.4 (sys)% from a strained GaAs/GaAsP superlattice photocathode grown by metalorganic chemical vapor deposition (MOCVD) and cooled to dry-ice temperature (195 K). We achieved this polarization with 97.8% circularly polarized excitation light and a quantum efficiency of 0.7% at the peak polarization wavelength. This measurement exceeds the values of previously reported GaAs-based photocathode polarizations, which have clustered near 92% for two decades. We vary the temperature of the cathode and measure the polarization and quantum-efficiency spectra at 295 K, 273 K, 195 K, and 77 K. The polarization rises from 91.2(1)% at 295 K to its maximum at 195 K, while the spectral peak shifts from 775 nm to 739 nm (78 meV) over the full temperature range, tracking the widening band gap. The spectra probe two depolarization mechanisms: a thermalized transport channel that is suppressed on cooling, and energy-dependent hot-electron relaxation that persists and sets the low-temperature saturation. The polarization recovers after a full cooling and warmup cycle and is stable while the quantum efficiency decays, which disfavors surface energy filtering as the origin of the gain. These results indicate that modest cooling to 195 K, for which dry ice suffices, is a practical route to higher-polarization GaAs-based electron sources.
Read the full abstract

All authors Matt Grau, Colin Kirk, Greg Blume, John Hill, Sushil Poudel, Alimohammed Kachwala, Marcy Stutzman, Joseph Michael Grames, Sylvain Marsillac, Matt Poelker

Artificial Intelligence

AI Should Not Only Be Helpful. It Should Be Contingent. Artificial Intimacy, Sycophancy, and the Future of Social Learning

Conversational artificial intelligence is increasingly embedded in everyday social environments, where it functions as both an informational tool and a source of interpersonal feedback. This perspective introduces contingency, i.e., the degree to which system responses vary with user behavior and its interpersonal consequences, as a central construct for evaluating AI systems. We argue that current alignment approaches, including reinforcement learning from human feedback, tend to prioritize user approval and conversational fluency over behaviorally informative feedback, leading to sycophantic patterns of noncontingent affirmation. Drawing on behavioral science and social learning theory, we propose that contingent feedback is a key mechanism through which individuals develop interpersonal skills. When AI systems provide feedback weakly coupled to social consequences, they may reduce opportunities for adaptive calibration in real-world interactions, particularly during adolescence, a critical period for social development. We outline a framework for contingent AI, including trajectory-based evaluation and models of social consequence prediction, and propose a research agenda spanning developmental psychology, human-AI interaction, and machine learning. More broadly, we argue that AI systems should be evaluated not only by user satisfaction, but by their impact on human social learning.
Read the full abstract

Combinatorics

Localization of the Caro-Wei bound and its applications to bipartiteness

We confirm a conjecture of Brause, Randerath, Rautenbach and Schiermeyer (2016) by proving a localized lower bound on the independence number of a graph that strengthens the classical bounds of Fajtlowicz (1978) and of Caro (1979) and Wei (1981), which in turn settles a conjecture by Bertram and Horák (1996). Our proof is based on a new Motzkin–Straus-type inequality involving local clique numbers and the independence number. We then apply the developed methods to study spectral and algebraic measures of graph bipartiteness. In particular, we extend a theorem of Brandt (1998) on spectral bipartiteness from regular $K_{r+1}$-free graphs to all $K_{r+1}$-free graphs, we improve a general upper bound for the least signless Laplacian eigenvalue of $K_{r+1}$-free graphs, and we disprove a conjecture of de Lima, Nikiforov and Oliveira (2016) in the case of $K_4$-free graphs.
Read the full abstract

Analysis of PDEs

Exit Times for Brownian Motion and Location Detection

In 2023, Wyman and Xi asked “Can you hear your location on a manifold?” We pose a related question, asking if probabilistic data can be used to recover location within a manifold. More precisely, we ask if location can be recovered from the distribution of exit times of Brownian sample paths. We show that exit time moments determine location up to symmetry for a variety of two-dimensional domains. Beginning with (convex) domains whose boundaries are conic curves, we show that in elliptic, parabolic, and hyperbolic domains, the first two exit time moments are sufficient to detect location up to symmetry. We establish similar results in unbounded wedge domains and equilateral triangular domains. We show that the full sequence of exit time moments determines location in rectangular domains. We end with two location detection results on general planar domains. The key tool driving our work is a “Poisson hierarchy,” which establishes a connection between exit time moments of Brownian motion and solutions to a family of PDE problems.
Read the full abstract

Fluid Dynamics

Chordwise micro-jet placement reveals a trade-off between mean hydrodynamic performance and unsteady loading under cloud cavitation

Prescribed tangential micro-jet injection can modify cloud-cavitation dynamics, but the chordwise location that improves mean hydrodynamic performance may differ from the location that minimizes unsteady loading. This study compares five injection locations (x/c = 0.15, 0.30, 0.45, 0.60, and 0.70) on a Clark-Y hydrofoil at Re = 7 x 10^5 and cavitation number 0.8. Transient large-eddy simulation with a Volume-of-Fluid formulation and the Schnerr-Sauer cavitation model is used in a two-dimensional parametric framework, with one representative three-dimensional case included to illustrate spanwise cavity deformation. The analysis considers vapor topology, turbulent kinetic energy, velocity and pressure fields, cycle-averaged surface pressure, hydrodynamic forces, force fluctuations, spectra, and cavity-thickness histories. Injection at x/c = 0.15 gives the highest cycle-averaged lift-to-drag ratio among the tested cases: drag decreases from 0.137 to 0.107 and the lift-to-drag ratio increases from 5.693 to 6.261, while lift decreases from 0.78 to 0.67. In contrast, x/c = 0.60 gives the lowest recorded force-fluctuation RMS, with lift- and drag-coefficient RMS values of about 0.112 and 0.0165. The preferred injection location is therefore objective-dependent: the location favored by mean hydrodynamic performance does not coincide with the location favored by unsteady-load reduction under the present conditions.
Read the full abstract

All authors Amirmehran Mahdavi, Soheil Motezakkeri, Yasin Ebrahimi, Ali Alavi

Physics and Society

Rock, Paper, Scissors, ... Dynamite - A Model of Disruption from New Technologies

We seek to understand the effect of adding disruptive highly-capable new technologies to competitions by assessing the addition of Dynamite to Rock-Paper-Scissors. We find that providing a versatile Dynamite move to only one player provides limited value (win probability increases from 50% to 55.5%) and is played rarely. That value decreases further if the game is expanded beyond just the original three moves. We also observe several mechanisms by which prior moves can become strategically unplayable, or obsolete. We hope that this model illustrates some non-intuitive aspects of developing new versatile technologies. We also hope that it illustrates some pitfalls for developers and integrators to avoid in order to create value rather than merely capability.
Read the full abstract

Computer Vision

Distributed Implicit Harm: A Compositional Safety Blind Spot in MLLM-Based Video Moderation

Despite their growing use in video moderation, multimodal large language models (MLLMs) exhibit a compositional safety blind spot: videos composed of seemingly benign components can convey harmful meaning when interpreted as a whole. We refer to this phenomenon as Distributed Implicit Harm (DIH), where harm arises from relations among components distributed along a decomposition axis of the video, rather than from any single explicit cue. Among many possible axes, we study two representative cases: temporally distributed harm across visual segments (DIH-T) and cross-modal harm between audio and visual streams (DIH-M). Studying and mitigating DIH at scale requires data that is difficult to collect: such videos lack compositional harm annotations, evade retrieval based on local visual cues, keywords, or single-modality signals, and are consequently absent from existing safety datasets. To bridge this gap, we develop a multi-agent synthesis framework that composes individually benign components into harmful scenarios and generates diverse DIH videos with explicit reasoning annotations, yielding a dataset of over 9,000 videos spanning visual-only and audio-visual settings. Benchmarking over 30 MLLMs spanning frontier proprietary models and leading open-source systems reveals substantial and consistent deficits in detecting both DIH-T and DIH-M. Notably, this failure persists even among the strongest frontier models: they often correctly assess individual components in isolation but fail to recognize the harmful meaning that emerges from their composition. We further evaluate these models on a manually collected set of real-world DIH videos from social media and observe the same failure mode, highlighting DIH as a practical and underexplored challenge for video moderation.
Read the full abstract

All authors Ruotong Wang, Zihao Zhu, Siwei Lyu, Xin Tao, Baoyuan Wu

Distributed Computing

BlockMGARD: Accelerating Adaptive Scientific Data Reduction with Region-of-Interest Error Control on GPUs

The growing scale of scientific data makes lossy compression essential for reducing data volume under controllable error. Transformation-based compressors using multilevel decomposition, such as MGARD, achieve strong compression ratios but map poorly to GPU architectures. We propose BlockMGARD, an adaptive, Region-of-Interest (ROI)-supported GPU lossy compressor, with four contributions: (1) an In-cache Block decomposition leveraging GPU on-chip memory and constant lookup tables to accelerate decomposition; (2) a hybrid hierarchy combining In-cache Block and global decomposition to balance speed and compression ratio; (3) an end-to-end pipeline with fine-grained ROI error control for feature preservation; and (4) an evaluation against state-of-the-art methods on five real-world datasets. Compared to MGARD-X, BlockMGARD achieves up to 4.2x and 9.1x higher compression and decompression throughput, and up to 8.63x higher compression ratio than uniform-tolerance baselines under ROI-aware error control. Across four GPUs, BlockMGARD achieves near-ideal linear scaling and up to 1.58x I/O cost reduction over MGARD-X.
Read the full abstract

All authors Yanliang Li, Qian Gong, Qing Liu, Jaemoon Lee, Norbert Podhorszki, Scott Klasky, Xin Liang, Jieyang Chen

Systems and Control

Exploiting the Benefits of V2B Application on Peak Shaving of Data Center Loads

The accelerated growth in data center projects has introduced a demand-driven bottleneck throughout power grids and contributed to a substantial increase in carbon emissions. These concerns are fueling discussions on methods to use existing energy assets to drive operational efficiency. To this end, this paper explores the benefits of Vehicle-to-Building (V2B) applications to support peak shaving of data center cooling loads. Initially, a literature review was conducted considering V2B constraints and optimization methods including SoC limitations, EV participation, tariffs, and building loads. This analysis was then used to develop a conceptual case study of a 10 MW data center in Loudoun County, VA by simulating a temperature-dependent load profile and adjusting the V2B participation of 40 commercial and passenger EVs. Simulation results indicate that, depending on seasonal variations in cooling load demands, strategic deployment of V2B assets between 12-5pm can offset gross cooling loads by 13-36%.
Read the full abstract

Chemical Physics

Smoothed truncated Coulomb potential for periodic Gaussian-basis Hartree–Fock exchange

Truncated Coulomb (TC) potentials reduce finite-size errors and accelerate thermodynamic-limit convergence in periodic Hartree–Fock (HF) calculations, but their use with Gaussian basis sets is complicated by the evaluation of electron repulsion integrals (ERIs), particularly for nonspherical truncation domains and all-electron calculations. We introduce the smoothed truncated Coulomb (sTC) potential as a systematically improvable approximation to a parent TC potential. A real-space Gaussian convolution smooths the sharp truncation boundary, and a single dimensionless parameter, $\eta$, controls the width of the smoothing window, which can be tightened to systematically approach the TC reference. The smoothing by sTC enables a dual-space algorithm for evaluating periodic Gaussian-basis ERIs that requires neither a large plane-wave basis nor new molecular integral kernels and is applicable to both pseudopotential and all-electron calculations, including the important Wigner–Seitz-cell truncation boundaries. Benchmarks spanning insulating, semiconducting, layered, metallic, and molecular-crystal systems show that sTC-based HF closely reproduces TC results in pseudopotential calculations and extends TC-quality calculations to all-electron settings. Across these systems, sTC substantially improves thermodynamic-limit convergence relative to the probe-charge Ewald method while remaining practical for all-electron calculations in which direct TC-based calculations are computationally challenging.
Read the full abstract

Combinatorics

Varieties of chain complexes and mixed dimer covers

A quiver representation consists of a collection of vector spaces along with a set of arrows, which are linear maps between these spaces. In this work, we study quiver representations in equioriented type $A$ which are also chain complexes; that is, in which consecutive arrows compose to zero. We show that orbits of these representations under a change of basis action are in bijection with mixed dimer covers of a $2 \times n$ grid graph. The latter object can be endowed with a partial order which is a distributive lattice, and we show that the degeneration order on the orbits of chain complexes is a coarsening of this partial order. In addition, we use recent matrix formulae of Claussen and Ovenhouse to enumerate these orbits. This also computes the Kostant partition function applied to height-restricted, type $A$ roots. When the dimension vector is uniform, we discuss a correspondence with paths of a beam of light bouncing between glass plates and give an explicit generating function.
Read the full abstract

All authors Portia X. Anderson, Esther Banaian, Melanie J. Ferreri, Nicholas Mayers, Shiyun Wang, Alexander N. Wilson

General Relativity and Quantum Cosmology

Unimodular Gravity with Arbitrary Diffusion Function: A Dynamical System Reconstruction Approach

We investigate cosmological diffusion models in unimodular gravity within a dynamical systems reconstruction framework. By treating the logarithmic slope of the diffusion sector as an invertible dynamical variable, the diffusion function can be systematically reconstructed from the phase-space structure of the cosmological evolution. Under these conditions, we determine the physically admissible fixed points of the system, identifying novel matter–diffusion scaling solutions associated with power-law diffusion sectors, as well as purely diffusion-dominated configurations capable of driving late-time accelerated expansion without requiring an explicit cosmological constant term. The local behavior around the fixed points is then extended to the full cosmological evolution, providing a framework to explore the global implications of diffusion cosmologies. Beyond the asymptotic fixed-point structure, we further develop a reconstruction formalism based on the dynamical evolution of the diffusion slope, allowing for trajectories interpolating between different diffusion regimes during the cosmic history. Our results establish a systematic framework for constructing and classifying viable diffusion cosmologies in unimodular gravity directly from the phase-space dynamics.
Read the full abstract

General Relativity and Quantum Cosmology

Gravitational waves from core collapse of rotating very-massive stars: 3D numerical relativity computation

We numerically study the collapse of rotating, very-massive stellar cores with masses of $\approx 200$, $300$, $500$, and $1100M_\odot$ into black holes using both axisymmetric and three-dimensional (3D) numerical relativity. Our results indicate that when the dimensionless spin of the resulting black hole exceeds 0.8, a massive disk consistently forms around it. These massive, compact disks, carrying more than about 10% of the black hole's mass, are prone to non-axisymmetric deformations that trigger gravitational-wave bursts with frequencies around 10-50 Hz. Such waves could be detected by the Einstein Telescope and Cosmic Explorer, even from sources a few Gpc away. We also summarize the gravitational-wave signals from axisymmetric collapse, which tend to have lower amplitudes and higher frequencies than those caused by non-axisymmetric instabilities.
Read the full abstract