Peer Revenue Now accepting positions

Papers

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

All of arXiv

Sunday 8 November

Machine Learning

When Calibration Data Matter in Sub-4-Bit Language Model Quantization

Low-bit post-training quantization reduces language-model memory but can sharply degrade downstream task accuracy. Calibration usually relies on generic text, but when task-informed data recover accuracy and which quantizers benefit remain poorly understood. We introduce FragCal-DC, which supplements generic text with training documents from task families with low retention of full-precision accuracy in a two-bit scan. We use a two- to four-bit sweep to locate functional regimes and four-seed calibration comparisons to evaluate six data-dependent methods across six dense models (1.7B-8B) and ten tasks. At their lowest tested functional bit widths, iterative SignRound V2 and OmniQuant improve at two bits, whereas single-pass GPTQ and AWQ show small mean responses at three bits. At two bits, SignRound V2 gains 1.83 percentage points in mean ten-task accuracy over generic-text calibration across six models. At 70 billion parameters, GPTQ and SignRound V2 both improve at two bits, with functional generic-calibration baselines. For two-bit SignRound V2, task-level controls show that recovery concentrates on families included in calibration, while incorrect answer designations retain about 90% of the mean accuracy gain. Changes in reconstruction error do not consistently track accuracy gains. These findings support task-informed calibration as a design principle for preserving downstream accuracy in low-bit quantization.
Read the full abstract

Astrophysics of Galaxies

Studying the Fullerene Nest: JWST Observations of the Planetary Nebula Tc 1. III. Tracing Fullerene-Linked Dust Variation

We have observed the fullerene-rich planetary nebula Tc 1 with the MIRI/MRS on JWST. The mid-infrared spectra of this spatially resolved source reveal that the dust emission is not co-spatial with the fullerene emission. Three broad features can be distinguished: a 6$\unicode{x2013}$9 $\mu$m plateau, a 9$\unicode{x2013}$13 $\mu$m plateau and the 30 $\mu$m feature. The 6$\unicode{x2013}$9 $\mu$m plateau does not change shape spectrally across the entire field of view, indicating it is produced by stochastically heated material. Spectral variations of the 9$\unicode{x2013}$13 $\mu$m plateau reveal at least three distinct components, peaking at 9.5 $\mu$m, 11 $\mu$m and 12 $\mu$m respectively. SiC produces most likely at least the 11 $\mu$m component and we find evidence that the 12 $\mu$m component might be produced by SiC grains too. The variation in relative strength of these components across the field of view hints at dust processing taking place in Tc 1. The 9.5 $\mu$m component of the 9$\unicode{x2013}$13 $\mu$m plateau has a strong correlation with the SiC component(s) and is thus likely produced by similar material.
Read the full abstract

All authors Simon Van Schuylenbergh, Jan Cami, Els Peeters, Morgan M. Giese, Charmi Bhatt, Dries Van De Putte, Cedric Baerts, Michael J. Barlow, Jeronimo Bernard-Salas, Alessandra Candian, P. Bryan Changala, Nick L. J. Cox, Harriet L. Dinerstein, Vincent J. Esposito, D. A. García-Hernández, Marco A. Gómez-Muñoz, Kay Justtanont, Kathleen E. Kraemer, Eric Lagadec, Arturo Manchado, Ana Monreal Ibero, Raghvendra Sahai, Ameek Sidhu, G. C. Sloan, N. C. Sterling, A. G. G. M. Tielens, Jeremy R. Walsh, Roger Wesson, Joshua Cole Whitman, Albert Zijlstra

Other Condensed Matter

Spin Conductivity: Insight From Perturbation Theory

The investigation on the spin conductivity has now been blooming in burgeoning altermagnetic materials. Here, an alternative and elementary derivation for the spin conductivity expression is presented. This is done within the context of perturbation theory. It indicates that our derived expression for the $\mathcal{T}$-even ($\mathcal{T}$ for time reversal) spin conductivity agrees exactly with those derived from the Kubo linear response theory and quantum kinetic theory. In addition, the expression for the $\mathcal{T}$-odd spin conductivity in the clean limit is exactly consistent with those presented previously. We show that the $\mathcal{T}$-odd ($\mathcal{T}$-even) spin conductivity arises from the contribution of zeroth-order (first-order) wave function correction. Our work provides some physical insight into the spin conductivity from the perturbation theory and shines more light on the charge-to-spin conversion mechanism in magnetic materials.
Read the full abstract

All authors Xu Chen, Mingbo Dou, Yu Sui, Xianjie Wang, L. L. Tao

Computer Vision

ADC-RANet: Improving Resolution Adaptive Neural Networks with Attention, Distillation and Calibration

Dynamic neural networks are a class of adaptive models in which the computation assigned to each input sample is determined dynamically, potentially reducing computational cost while maintaining similar performance. For image classification, the Resolution Adaptive Network (RANet) is a strong approach that progressively allocates computational resources across input scales based on sample difficulty. In this work, we introduce RANet with Attention, Distillation and Calibration (ADC-RANet), an enhanced version of RANet that improves performance across all computational budgets. Our approach combines three tightly connected components: bottom-up channel attention, top-down cascaded knowledge distillation, and overconfidence correction using MixUp, in a synergistic design that coordinates both pathways. Experiments on CIFAR-100 and ImageNet datasets show consistent gains while preserving the efficiency of RANet. Our code is available: https://github.com/jefersonseverino/ADC-RANet
Read the full abstract

All authors Jeferson Severino de Araujo, Lilian Hollard, Eric Granger, Marco Pedersoli

Astrophysics of Galaxies

Season by season causal analysis of the continuum and velocity-resolved H$\beta$ emission of NGC 5548

We investigate possible causal influences in five historical seasons (1989-1993) of NGC 5548 using a version of the Synergistic-Unique-Redundant Decomposition (SURD) method adapted for this analysis, an information-theoretic causal discovery method for time series. Two four-variable sets contain the blue wing (B), core (C) and red wing (R) of H-beta together with either total H-beta (H) or the 5100 Angstrom continuum (F). We examine 7200 decompositions with common and unequal lags, two or three seasonal flux states, and base delays of 1-30 d. We rank the individual named components within each information family, reporting their amplitudes and simultaneous leakage. In 1989, unique continuum information about the core reaches 0.799 and 0.387 of target entropy for two and three states, with leakage 0.047 and 0.276. In 1992, the red-wing history supplies unique information about the future blue wing in both variable sets and both state counts, with peak target-entropy fractions 0.106-0.166. Distinct joint influences also emerge: for the 1991 core, the two-state S(F,B) and S(F,R) components contribute 0.187 and 0.125 at 20 d with leakage 0.087, although the unique continuum term vanishes. Strong core-wing redundancy identifies candidate inputs for simpler predictive models. Large late-lag leakage, reaching approximately 0.97 in the total set, identifies targets for additional driver measurements and richer temporal histories. The resulting seasonal map of individual, shared and joint influences complements the response delays and velocity structure measured by reverberation mapping, and provides specific causal hypotheses for the evolving broad-line region.
Read the full abstract

Machine Learning

System Prompts Affect Response Bias More Than Discriminability in LLM Safety

Large language model safety evaluations often focus on a compliance-refusal tradeoff: models should refuse unsafe requests without refusing safe ones. We argue that this misses a broader form of prompt-conditioned task selection, where the same request can elicit different behaviors depending on the system prompt. We investigate this using a controlled evaluation crossing eight safety-task prompt families with three clarity levels across HarmBench, XSTest, and IHEval. Holding benchmark items fixed, we evaluate 13 models and use signal detection theory to distinguish two possibilities: whether models become better at telling safe from unsafe requests (discriminability), or simply more willing to refuse (response criterion). We find substantial differences in discriminability across models, but system prompts primarily shift the willingness to refuse, with smaller and less consistent effects on discriminability. Increasing prompt clarity makes models more willing to refuse without reliably improving their ability to distinguish safe from unsafe requests. Beyond refusal, prompt families also shift responses among six distinct behavioral policies, including clarification and safe help. Together, these findings show that system prompts primarily change how models respond to safety-relevant requests rather than consistently improving their ability to distinguish safe from unsafe ones.
Read the full abstract

Optimization and Control

A Globally Convergent Line-Search Quasi-Newton Method for Discounted Markov Decision Processes

In this paper, we develop a line-search quasi-Newton (LSQN) method for discounted Markov decision processes (MDPs). The method maintains an inverse model to approximate the inverse of the current policy Jacobian and refines it through an inner learning procedure that preserves existing secant conditions and achieves the prescribed accuracy in at most $n$ queries in exact arithmetic for an MDP with $n$ states, even when the initial inverse model is singular. A sign-based direction correction combined with an Armijo line search ensures monotonic decrease of the Bellman residual norm across policy changes without requiring a value-iteration fallback. We prove global Q-linear residual convergence and local Q-superlinear convergence of the value iterates from arbitrary finite initial value vectors and inverse models, without requiring a unique optimal policy. Numerical experiments show competitive performance against nine baselines, while public benchmarks demonstrate that the inner learning procedure typically requires far fewer than $n$ learning queries.
Read the full abstract

Human-Computer Interaction

Wording Matters: Understanding How GenAI Policy Rules Shape Teachers' and Students' Workflows and Perceptions

The growing use of generative AI (GenAI) in higher education has led institutions to develop policies governing its applications. However, we have little knowledge about how teachers and students adapt and adhere to the policies. In this paper, we analyzed GenAI policies from 101 U.S. universities, identifying nine recurring rules for GenAI use, and interviewed 20 students and 20 faculty members about the rules. We find that the wording of the rules can change how students perceive and act on the rules while introducing different challenges that teachers face in enforcing them in practice. For example, rules such as "Passing off AI-generated content as your own constitutes academic dishonesty" and "Students are accountable for any inaccurate, biased, or unethical content" both outline the acceptable use of AI, but the former can make students hide their use of AI out of fear of unwanted scrutiny, whereas the latter can encourage students to take responsibility and ownership of their use of AI and be more transparent. Students described uncertainty about acceptable use, fear of penalties and unfair treatment, and stigma that discouraged acknowledging AI use. Teachers took on approval workload, had to interpret acceptable use, and lacked reliable ways to verify compliance. We summarize these impacts in a policy-impact matrix linking each rule to its effects on teacher and student workflows and perception.
Read the full abstract

All authors Sabiha Mehreen, Xiangzhe Yuan, Rishab Nithyanand, Naimul Hoque

Computer Vision

NanoNet: Ultra-Lightweight Network for MRI Brain Tumor Classification toward Edge AI

Automated brain tumor classification from magnetic resonance imaging (MRI) can support accessible medical image analysis, but conventional deep learning models often exceed the resource budgets of highly constrained edge-AI platforms. We present NanoNet, an ultra-lightweight convolutional neural network designed for resourceconstrained MRI classification and motivated by future edge-AI applications. Inspired by MobileNetV2, NanoNet substantially reduces network depth, channel capacity, expansion ratios, and final feature dimensionality, resulting in only five inverted residual blocks and 8,132 parameters. We further investigate dependency-aware structured pruning, knowledge distillation, and INT8 quantization-aware training for additional compression. Experiments on the BRISC 2025 benchmark show that NanoNet achieves 92.80% accuracy and 92.97% macro F1 with 8,132 parameters and 10.498 MFLOPs. Across three compression seeds, it exceeds the mean accuracy of 70%-pruned MobileNetV2 by 5.20 percentage points while using 60.2% fewer parameters and 9.8% less computation. Converted INT8 NanoNet achieves 89.70% accuracy and 90.22% macro F1 across three QAT seeds. Earlier TensorFlow Lite measurements indicate 27.7 KB of baseline model storage, motivating future edge-AI deployment.
Read the full abstract

All authors Tran Nguyen-Le, Huu-Thuan Huynh, Minh-Triet Tran, Trung-Nghia Le

Materials Science

Learning Electronic Band Structures Across the s, p, d Blocks with an Empirical Pseudopotential Network

Electronic band structures govern how crystals conduct electrons and absorb light. Density-functional theory (DFT) computes them through costly self-consistent iterations, while existing machine-learning models either predict isolated scalar properties or output Hamiltonians that still require DFT solvers. Here we present EpNet, a learned empirical pseudopotential for crystals spanning 63 elements across the s, p and d blocks. An equivariant neural network builds its Hamiltonian from physics-motivated local potentials and nonlocal projectors. Trained solely on band energies, EpNet predicts complete band structures directly from atomic coordinates. On unseen chemical systems, its median errors are \winAll\ eV for band energies and \gapAll\ eV for band gaps. On 600,000 structures from an independent database, its band-gap errors are overall less than half those of existing models, at about 1% of DFT cost. EpNet thus enables fast screening of electronic properties for materials discovery and electronic-structure simulations of large systems.
Read the full abstract

Probability

Mean-level arithmetic computation via stochastic reaction networks

Reaction networks provide a natural model for analog computation in which inputs and outputs are encoded by molecular abundances. Much of the work on analog computation with chemical reaction networks (CRNs) has focused on deterministic mass-action systems. In DNA- and protein-based circuits, however, low copy numbers can make intrinsic fluctuations significant, motivating the use of stochastic models. In this paper, we develop a framework for computation with stochastic CRNs at the level of stationary means. Specifically, we construct elementary arithmetic modules for identification, addition, and multiplication, and analyze their mean-level correctness and convergence rates. We focus on these three modules because they are simple enough to analyze explicitly while already exhibiting the new issues raised by stochastic mean-level computation. Unlike in deterministic systems, the computed value cannot be read from a single output measurement. Instead, readout requires Monte Carlo estimation, with error arising from transient bias as the system approaches stationarity and intrinsic sampling noise that persists even at stationarity. We therefore study the convergence rates of output means, variances, and autocovariances to their stationary values, as well as the stationary fluctuations governing sampling error. We also show that this readout cost can depend strongly, and sometimes unexpectedly, on the architecture of the computation. In particular, comparing linear-chain and binary-tree implementations reveals that binary trees can greatly reduce stationary variance for addition, but can instead produce exponentially larger variance for multiplication.
Read the full abstract

Machine Learning

Discrete Diffusion Playground: A 2D Benchmark for Discrete Generative Models

Discrete diffusion is a promising framework for categorical generative modeling, but current evaluation protocols often fail to reveal whether models have actually learned the target distribution. We introduce the Discrete Diffusion Playground, a diagnostic benchmark that converts known 2D continuous distributions into binary token sequences through an exact discretized bijection. This creates a rare setting in which models are trained on discrete sequences, but generated samples can be decoded back into 2D and compared directly against a known ground-truth distribution. The benchmark spans smooth Gaussian mixtures and tilted checkerboards with sharp boundaries, disconnected support, and coordinate-dependent structure, allowing it to expose failures that are invisible in standard proxy metrics. It supports exact grid divergences, Wasserstein distances, density visualizations, and nonparametric two-sample tests, providing both quantitative evaluation and interpretable failure diagnosis. Using this testbed, we show that masked diffusion models can learn meaningful adaptive reveal orders but remain less stable than autoregressive baselines on smooth targets and largely fail on tilted checkerboards, where autoregressive models preserve the target geometry. Control experiments with reversed and interleaved token orderings show that autoregressive performance is not explained by favorable bit order alone. The Playground provides a simple but stringent benchmark for stress-testing discrete generative models and for developing evaluation metrics that detect real distributional mismatch rather than proxy performance. Our code is available at https://github.com/shaojus/DiscreteDiffusionPlayground .
Read the full abstract

All authors Justin Shao, Sihyun Park, David Ryan Koes, Maria Chikina

Cryptography and Security

Beyond AUROC: Deployment-Aware Evaluation of Model-Internal Jailbreak Detectors

Large language models (LLMs) are increasingly deployed in user-facing systems, where jailbreak attacks can bypass safety controls and induce harmful and policy-violating outputs. Recent model-internal detectors use hidden representations, gradients, or internal signals to identify such prompts, but their practical security value depends critically on behavior at low false-positive operating points rather than on area under the receiver operating characteristic curve (AUROC) alone. We study two data dependencies that shape this behavior: calibration data, which determine the detection threshold, and reference data, which can influence detector scores. Across four model-internal detectors, four LLMs, and twelve jailbreak families, thresholds calibrated on curated benign prompts frequently fail to preserve their false-positive rate (FPR) budgets on real-user traffic. Recalibration improves false-positive control but reveals substantial family-level coverage gaps, including cases with high AUROC but low true-positive rate (TPR). Causal reference-set interventions show that changing reference composition can substantially alter family-level detection and that improving coverage of one attack family can reduce coverage of others. These findings show that calibration and reference data are part of a detector's security behavior. We therefore recommend evaluating model-internal jailbreak defenses using realized FPR on deployment-like traffic and per-family TPR at explicit low-FPR operating points, with AUROC used as a complementary ranking metric rather than the primary security measure.
Read the full abstract

All authors Mohammad Tehabsim, Basil Al-Housani, Nour Alhussien, Phung Lai

Atmospheric and Oceanic Physics

Rescene-ESM: an Earth system model built around a frozen AI weather model

Earth system models underpin climate projections, but their cost limits how many scenarios, ensemble members and impact pathways can be explored. Machine-learning weather models forecast as skillfully as physics-based models at a fraction of the cost, yet coupled to an ocean they have produced weak El Nino variability or have learned it from climate-model output, and none carries the climate through rivers to the people exposed to it. Here we present Rescene-ESM, an Earth system model whose coupler, YAHO, yokes a frozen 20.8-million-parameter AI weather model to physically formulated ocean, sea-ice, land, river, carbon and energy components constrained by observations. Adding the couplings one at a time shows what each contributes: over a climatological ocean the AI atmosphere varies from year to year by a third of the observed amount over the oceans, and an interactive mixed layer, an equatorial wave ocean, the residual of the observed sea surface temperature (SST) equation as stochastic forcing and the land's moisture flux raise this step by step and bring El Nino and its teleconnections, leaving the mean climate almost unchanged. The full model simulates about 1,000 years per day on one graphics processing unit, holds the observed climate for a thousand years without drift, reproduces the observed 1979-2022 global warming, carries forced change with a physically based structure around the frozen network (transient climate response 1.60 K), routes runoff to 212 river gauges with a median seasonal-cycle correlation of 0.79 and, driven by emissions, separates the heatwave exposure of an SSP1 and an SSP3 world by a factor of 4.7 in 2081-2100, three-quarters of it through the warmer climate.
Read the full abstract

Signal Processing

Analysis of the Effect of Doppler on Direct-to-Cell: Real-World Performance and Proposed Improvements

Direct-to-cell (DtC) satellites serve unmodified LTE phones today from low Earth orbit, where the satellite's motion creates Doppler shifts of tens of kHz. Since large offsets such as these cannot be compensated adequately by the user equipment (UE), the satellite pre-compensates the Doppler shift of each beam, but it can do so for only one reference point in the beam. A UE elsewhere in the beam still has to compensate a residual offset; previous field studies report residuals of up to 10 kHz, beyond the $\pm$7.5 kHz that the 15 kHz subcarrier spacing can resolve. In this paper we corroborate this effect with our own detailed measurements of Starlink transmissions and further correlate ineffective handover performance to inadequate residual offset compensation. We analyze the Doppler compensation problem with detailed standard-specific simulations and propose a compensation method at the UE, based on the location of the UE with respect to the beam reference point used for the Doppler compensation and the known satellite location, that improves the efficiency of handover. With UE cell-specific pre-compensation, the neighbors are identified, read correctly and handed over to in time rather than lost to radio-link failure.
Read the full abstract

All authors Yuxi Chen, Muhammad Iqbal Rochman, Kyler McCormack, Jakob Kotz, Monisha Ghosh

Artificial Intelligence

GARP: Real-Time Generative Agents on Local Hardware in a Game Engine

Generative agents are game characters whose plans, conversations, and memories are written on the fly by a large language model (LLM). The original Generative Agents demonstration used cloud inference in a custom simulation. We present GARP, a platform for real-time generative agents inside Godot, and describe the Character Behavior System used in GARP 2. The complete loop runs on local hardware using a 3B-parameter open-weight model and no cloud service. Alongside the original planning, memory, and reflection, GARP gives each agent an emotion system driven by LLM appraisal, four graded personality axes drawn from the Big Five that set simulation parameters as well as prompt text, six needs whose urgency grows over time, and pairwise relationships updated after every conversation. The system narrows 50+ candidate actions to typically 3-8, constrains model responses with JSON schemas, and surfaces important memories in decision prompts. A layered optimization strategy (an adjustable game clock, decisions taken ahead of need, multi-tier caching, prompt compression, and fully asynchronous execution) keeps the game responsive while the model thinks. In a nine-scenario evaluation, 25 of 27 runs passed log-based checks. Both failures occurred in the birthday party scenario, where one guest omitted the event from their daily plan.
Read the full abstract

All authors Piero Molino, Patrick John Chia, Federico Bianchi, Jonathan P. Chen, Paul Szerlip

Machine Learning

Target Recoverability under Selective Labels: A Failure Mode in ECG-Based Acute Coronary Syndrome Prediction

Medical machine learning models often learn from diagnoses observed only in patients clinicians chose to test. The biases of such selective labels are well known, but less understood is what happens when a diagnosis includes cases the model's inputs cannot detect. Because testing decisions draw on information the model does not see, these cases can appear predictable among tested patients, producing relationships that fail, and can even reverse, among the untested. We formalize this mechanism and show in simulations that it worsens as testing becomes more selective or undetectable cases more common, while models trained only on detectable cases remain informative. We empirically study electrocardiogram (ECG)-based prediction of acute coronary syndrome (ACS), which is confirmed only by invasive testing and whose subtypes differ in how clearly they appear on the ECG. Three ECG foundation models fine-tuned on revascularization, our ACS proxy, performed well among catheterized patients (AUROC 0.740-0.770) but worse than chance for 30-day major adverse cardiovascular events among untested patients (AUROC 0.381-0.478). Models trained on ECG-visible subtypes retained discrimination (AUROC 0.680-0.709), whereas a model trained on unstable angina, which can occur without ECG changes, was inversely related to risk (AUROC 0.310). Under selective labels, effective learning therefore requires prediction targets that can be inferred from the model's inputs and evaluation in the intended deployment population, which often extends beyond the patients who were tested.
Read the full abstract

Computer Vision

What Does Multiple Choice Actually Measure? Disentangling Video Understanding from Option Exploitation

Video understanding evaluations are intended to measure a model's proficiency in video perception and task solving. However, current VideoQA benchmarks struggle to isolate these capabilities due to their reliance on multiple-choice (MC) answer guidance, language priors, and fragile scoring pipelines. Disentangling these confounding factors reveals how much of a model's score is supplied by the evaluation context rather than earned through faithful perception. Across six VideoQA benchmarks of varying length, we find that accuracy is significantly above chance without any visual input. This is caused by shortcutting and option-prior reliance. We express these effects in the language of faithfulness: option-exploitation and answer-conditioned reasoning traces are unfaithful even when they are correct. Together, our results show that a VideoQA score is jointly determined by language priors, option timing, and how outputs and labels are converted into correctness. We produce a novel Option-Delayed Replay (ODR) mechanism that discloses options at controlled stages of an open-ended (OE) trajectory, to obtain objective MC scores from faithful option-free inference. Our work demonstrates that agentic methods with ODR and faithful inference can match option-exploiting performance. Finally, we formalize our diagnostic collection and ODR as a reproducible evaluation protocol, and we release a human-relabelled edition of LVBench (correcting $20.5\%$ of keys) to remove annotation noise as a confounding variable.
Read the full abstract

Accelerator Physics

Symplectic Normal Coordinates Flow for Generative Modeling of Hamiltonian Systems

In conditioned generative models of physical systems, the densities at different values of the conditioning parameters are often related through symplectic transformations. For stable linear Hamiltonian systems, there exists a symplectic transformation to "normal coordinates" such that the dynamics is reduced to rotations in each phase-space plane. This picture extends to the nonlinear case where, away from resonances, the Birkhoff normal form again consists of rotations in phase space, now with amplitude dependence. In this work, we introduce an invertible symplectic flow layer consisting of amplitude-dependent rotations in learned normal coordinates. We derive the condition under which the nonlinear rotation layer is symplectic and detail an architecture that guarantees it. Experiments are carried out on three conditional generative modeling problems: a synthetic chain of coupled nonlinear oscillators and two problems from accelerator physics. We show empirically that normalizing flows with the new layers achieve competitive or superior negative log-likelihoods compared to baseline flows, with gains that grow with the strength of the nonlinearity on the synthetic system.
Read the full abstract

Analysis of PDEs

The sharp Fourier extension estimate for the sphere

The full strong Fourier extension estimate for the unit sphere is established in every ambient dimension at least three. The finite Knapp boundary is obtained by combining a uniform diagonal estimate on rescaled spherical graphs with a transverse product bound. A Whitney decomposition and a dyadic dual pairing yield strong type at the exact scaling exponents. The critical local estimate and the strict mixed range come from the recent work on Euclidean local smoothing.
Read the full abstract

Analysis of PDEs

Relative Entropy Methods for Rotating Stars in the Compressible Euler-Riesz Equations

We establish finite-time relative entropy estimates for axisymmetric finite-energy entropy solutions of the three-dimensional attractive compressible Euler-Riesz equations near a prescribed compactly supported rotating star. The entropy solutions are assumed to have the same total mass as the steady state and to satisfy the prescribed angular momentum relation. For general pressure laws, the compatibility with the angular momentum constraint, together with the nonnegative rotational remainder, yields a cancellation of the centrifugal terms. For axisymmetric perturbations, the gradient of the perturbed Riesz potential has no azimuthal component and therefore couples only to those momentum components controlled by the modulated kinetic energy. The density-weighted estimates for the Riesz potential then allow us to close the relative entropy inequality. Under an additional higher-integrability assumption on the density, the result extends to the Chandrasekhar pressure law, yielding a corresponding stability estimate for rotating white dwarf stars.
Read the full abstract

Computer Vision

Differentiable Virtual-Slot Partitioning for Adaptive Tokenization in Vision Transformers

Adaptive tokenization aims to reduce the compute spent on uninformative image regions while preserving accuracy. Existing work relies on (i) fixed or frozen scoring rules, which cannot be trained toward the task loss, or (ii) discrete masks trained with straight-through estimators, which yield noisy gradients, while fully differentiable alternatives instead leave the pre-trained regular grid. We revisit adaptive tokenization from a regular-grid perspective and ask: can a two-scale quadtree be learned end-to-end without deleting a single token during training? We show that one coarse token and its four fine children can be made the two endpoints of a single continuous interpolation. We introduce Differentiable Virtual-Slot Partitioning (\dvsp{}), which keeps four virtual slots per $2{\times}2$ region and blends their token features and attention mass with a single weight $\alpha$, exactly reproducing the discrete grid at $\alpha \in \{0,1\}$. On COCO detection with EVA-02-B, \dvsp{} retains $97.9\%$ of dense Box AP while removing $64.2\%$ of visual tokens, giving a $1.61\times$ end-to-end speedup, and it outperforms entropy-based APT, DART, and uniform lower-resolution dense adaptation at matched budgets; these gains depend on the conserved attention mass, whose removal costs $1.59$ Box AP. On a frozen Segmenter it also improves Cityscapes mIoU, and on frozen DeiT backbones it exceeds entropy selection at identical per-image token counts.
Read the full abstract

All authors Gaoxiang He, Bin Zhu, Bo Xie, Xiaotong Yu, Yi Chen, Hongfei Yan

Programming Languages

Symbolic Synthesis with Token-Level LLM Guidance

We propose a synergistic combination of classical program synthesis techniques and contemporary large language models (LLMs) to solve the problem of recursive program synthesis from first-order logic specifications. Our approach leverages the strengths of both paradigms: we use LLMs as a source of statistical bias to guide the search for a correct program. At the same time we use techniques from classical synthesis to short-circuit on semantically incorrect prefixes and to avoid sampling semantically equivalent responses. The statistical bias is derived from the next-token distribution of an LLM, which is highly context-dependent. We operate with the understanding that (1) there is some class of synthesis instances that can be discharged by small models and symbolic tools, and (2) every other instance can be decomposed into these instances. Therefore, our intention is to expand the reach of small, open-weight LLMs and symbolic synthesis techniques, not to compete with proprietary frontier models on every possible instance. These are different tools for different jobs. In our experimental evaluation, we find that our method solves more benchmarks than either small LLMs or classical synthesis techniques alone. Our method solves 30/32 benchmarks using only a Qwen2.5 model with 32 billion parameters. Qwen2.5 solves 0 benchmarks on its own, and only solves a few additional benchmarks when combined with traditional constrained decoding techniques.
Read the full abstract

Combinatorics

Non-log-concave $\tau$-polynomials of regular graphs

For any simple graph $G$ of order $n$, write its chromatic polynomial in the rising factorial basis as $P(G,x)=\sum_{j=0}^{n}(-1)^{n-j}c_j(G)x^{\langle j\rangle}$, where $x^{\langle j\rangle}=\prod_{i=0}^{j-1}(x+i)$ for $j\ge1$ and $x^{\langle 0\rangle}=1.$ Brenti (1992) introduced the $\tau$-polynomial of $G$ and its augmented version, defined by $$\tau(G,x)=\sum_{j=0}^n c_j(G)x^j~~~\text{and}~~~\overline \tau(G,x)=\sum_{j=0}^n j!c_j(G)x^j.$$ He asked whether $\tau(G,x)$ is always real-rooted and whether the coefficient sequence of $\overline \tau(G,x)$ is always log-concave. In this paper, we prove that for every integer $d\ge4$, there exists an integer $h_d\ge3$ such that $c_2(G)^2<c_1(G)c_3(G)$ for every connected $d$-regular graph $G$ of girth at least $h_d$. For such graphs, the coefficient sequences of both polynomials fail log-concavity at $j=2$, and thus both polynomials have nonreal roots by Newton's inequalities. This gives counterexamples to both questions among connected regular graphs of arbitrarily large girth.
Read the full abstract

Artificial Intelligence

But It Was True When I Checked: Compiling SHOP-Style Method Preconditions into Partially-Ordered HTN Planning

Method preconditions specify when a method may be applied in hierarchical task network (HTN) planning, but their meaning differs under two established semantics. The SHOP planners, like the classical HTN formalism of Erol, Hendler and Nau, require the precondition to hold immediately before the method's first action. HDDL, the input language of the International Planning Competition's HTN track, instead introduces an auxiliary action to check it; under partial order, unrelated actions may invalidate the precondition before the method starts. HDDL's designers adopted this weaker semantics because the first-action reading could not easily be compiled. Subsequent work has repeatedly identified its compilation under partial order as an open question, with existing transformations limited to total order. We give a compilation of the first-action semantics into standard partially ordered HTN planning without method preconditions using only two additional facts, thereby resolving a longstanding question in the HTN planning community.
Read the full abstract

Plasma Physics

Analysis of Solid Boron Injection for ITER

Since ITER will now operate with a full tungsten divertor and first wall, the ITER Organization is considering boron wall conditioning schemes to getter oxygen and reduce tungsten sputtering to avoid excessive radiation to ensure that ITER can reach and sustain high- performance plasma operation. ITER has adopted Glow Discharge Boronization (GDB) in deuterated diborane (B$_2$D$_6$) as its main technique, which can only be performed in the absence of magnetic fields, but has also requested a study to determine how effective solid boron injection (SBI), which can be performed during plasma operation, using either an Impurity Powder Dropper (IPD) or a boron pellet injector, might be as a risk mitigation to achieving Q=10. Solid impurity injection has been used for decades as a wall conditioning technique on numerous tokamaks and stellarators around the world. The study presented here begins to assess SBI on ITER by first quantifying the boron source term across a particle size scan from 1 $\mu$m to 5 mm diameter and an injection velocity scan from 5 – 200 m/s for several ITER plasma conditions: 1) Start of Research Operations (SRO) diverted L-mode, 2) Q=10 H-mode, and 3) low plasma current and low power limiter operation. Modeling of the ablation, transport, erosion and redeposition of solid boron is being carried out with the Dust Injection Simulator (DIS), the Neutral Gas Shielding (NGS), EMC3-EIRENE and WallDYN3D models. In the first steps presented here, the DIS and NGS models are compared for calculating the boron source term that is then used as input to EMC3-EIRENE and will later be used with WallDYN3D.
Read the full abstract

All authors Florian Effenberg, Joseph A. Snipes, Federico Nespoli, Robert Lunsford, Jonathan van Blarcum, Manni Jia, Heinke Frerichs, Alessandro Bortolon, Klaus Schmid, Tom Wauters

Neural and Evolutionary Computing

Neural Activity Synchronization-Based Fast Local Learnable Spiking Neural Network Architecture

Implementing intelligent learning systems on low- power edge devices is challenging due to the high compu- tational complexity, training latency, and energy demands of traditional Artificial Neural Networks (ANNs). Spiking Neural Networks (SNNs), biologically plausible and spike-driven, come as an alternative. However, SNN training commonly relies on backpropagation through time (BPTT), which is computationally expensive, inherently global, and unsuitable for battery-driven devices. This work proposes a novel local learning framework for multi-layered SNNs inspired by Neural Activity Synchronization (NAS). Each layer updates its own weights using locally com- puted gradients. No gradient backpropagation is needed. This significantly reduces computational complexity, training latency, and energy consumption. The proposed SNN model is trained and evaluated on the MNIST dataset, achieving a classification accuracy of 97.78%. The model is also analyzed with Gaussian and salt-and-pepper noise, demonstrating robust performance across different noise levels.
Read the full abstract

All authors Pranith Dutta, Sk Wasim Rajjak, Siddhanta Ray, Santlal Prajapati, Manobendra Nath Mondal

Image and Video Processing

Kinematic Total-Body PET Video Reconstruction via Hierarchical List-Mode EM and Validation-Adaptive Space-Time Regularisation

Kinematic PET is introduced as an imaging task focused on visualisation of early radiotracer delivery and rapid multi-organ motion without gating. The proposed kinematic video reconstruction method uses 33-ms frames (~30 frames per second), initialised with 3D expectation maximisation (EM) time of flight (TOF) list-mode reconstruction, followed by cycles of a 2-stage update. These 2 stages are i) a hierarchical EM (HEM) update using multiple temporal pooling levels, and ii) a validation-adaptive space-time (VAST) regularisation update. HEM combines frame-specific EM corrections with corrections estimated from higher-count temporal blocks, slowing noise amplification for the ultrashort frames. VAST mixes each 3D target patch in a frame with source patches sampled nearby in space and time. It selects the candidate that best matches the mean of the corresponding patches in the two adjacent frames, with the unchanged patch also eligible. Optional temporal block sum correction uses the initial EM temporal sum as a reference to reduce loss of spatial detail. HEM+VAST was evaluated with a dynamic digital phantom, and demonstrated on human FDG data acquired on a Siemens Biograph Vision Quadra (28 - 58 s post-injection). At a background spatial coefficient of variation of 10%, the temporal full width at half maximum of the fastest phantom bolus was 151 ms with HEM+VAST, compared with 86 ms for the ground truth. The video frame rate FDG images reveal vascular and organ detail, radiotracer bolus passage and respiration-related multi-organ motion. Kinematic PET could complement conventional dynamic PET by supporting the development of image-derived physiological biomarkers and motion estimation for quantitative reconstruction. Future work should compare HEM+VAST with prior 4D PET reconstruction methods at the same 33-ms frame duration and matched noise levels.
Read the full abstract

All authors Andrew J. Reader, Gary J. R. Cook, Paul K. Marsden, Sarah-May Gould, Alexander Hammers, Simon R. Cherry

Hardware Architecture

PRISM: Preference-based Reward and Uncertainty-Integrated Synthesis Minimization for RTL PPA Optimization

This paper introduces a budget-aware, multi-objective framework (called PRISM) for RTL optimization that improves power, performance, and area (PPA) while sharply limiting expensive commercial synthesis tool calls. Starting from a synthesizable baseline and functional testbench, the method iterates through LLM-driven rewrite generation, fast compile/simulation equivalence filtering, low-cost proxy feature extraction, and selective commercial synthesis tool (Synopsys Design Compiler (DC)) evaluation. To prioritize which candidates deserve high-quality synthesis, we train a preference-based reward model using pairwise ranking supervision derived from commercial synthesis tool outcomes, and combine ensemble uncertainty with a diversity-aware acquisition rule to avoid early bias and redundant evaluations. A key advantage is that PRISM maintains a best-tradeoff set G of non-dominated designs throughout the run: G preserves the best solutions found so far, is fed back into generation and refinement, and enables consistent, cumulative improvements rather than repeatedly restarting from the baseline. Across RTLLM v2.0 modules, the approach consistently yields post-synthesis gains under fixed commercial synthesis tool call budgets, and ablations confirm that reward-guided selection and candidate dropping are central to sample-efficient progress. PRISM efficacy is evaluated across multiple technology nodes (7 nm, 15 nm, and 45 nm) using identical pipelines without retraining.
Read the full abstract

All authors Armin Abdollahi, Negin Ashrafi, Mehdi Kamal, Massoud Pedram

Robotics

Floquet-Parameterized Koopman-MPC for Tracking Periodic Orbits

Designing real-time control for nonlinear systems to track periodic orbits can be challenging, especially for complex and coupled dynamical systems (e.g., tethered kites), where the true model parameters are hard to obtain. To address this, we propose a physics-informed Koopman learning approach that incorporates periodic-orbit stability directly into the learned linear predictor. The approach combines the underlying periodic structure of the dynamics along with periodic state-input trajectory data to learn physically consistent linear predictors (in lifted/observable space). We leverage Floquet theory, which studies the stability of periodic orbits in nonlinear systems, to obtain the periodic stability information of a given orbit via a monodromy matrix (i.e., the state transition matrix over an entire period). Subsequently, we impose open-loop monodromy constraints to learn Floquet-parameterized linear predictors using a least-squares Extended Dynamic Mode Decomposition (EDMD) formulation, thereby constraining the search space and aiding in Koopman learning. We then utilize these linear predictors to design a model predictive control (MPC) law for tracking periodic references. Finally, we evaluate the designed framework on a cartpole system under periodic forcing and compare it against a purely data-driven baseline.
Read the full abstract

All authors Kartik A. Pant, Vishnu Vijay, Yufei Zhu, Xintong Deng, Taewoo Nam, Inseok Hwang