Peer Revenue Now accepting positions

Papers

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

All of arXiv

Monday 9 November

Machine Learning

Do Time-Series Foundation Models Forecast Index-Futures Returns? A Preregistered Evaluation of Models, Metrics, and Input Representations

We ask whether the apparent forecasting skill of frozen time-series foundation models (TSFMs) on intraday index-futures returns is incremental skill. In a pre-registered evaluation on E-mini Nasdaq-100 (NQ) and E-mini S&P 500 (ES) futures, a common panel of five zero-shot models (Chronos-Bolt, Chronos-2, TimesFM 2.5, Moirai 2.0 and TiRex), plus separate experiments with Kronos, is compared with a no-change forecast at 5-120 minutes, using native prices, log returns, volatility-standardised returns and a volume clock as inputs. On a blind set opened once, 0 of 160 model-representation-horizon-market cells beat no-change, and a Hansen SPA test gives p = 1.000 on both markets; a later retrospective hold-out, scored as a descriptive check, gives 0 of 120. Input representations change forecast magnitude but not accuracy, and a transformation-aware ridge regression does as well. Several standard measures nevertheless suggest skill. Price-level correlation exceeds 0.9996, but no-change scores higher. Within-day RankIC across forecast origins rewards a trivial revert-to-mean rule (+0.35 at 30 minutes); we derive the expected within-day covariance under a random walk in closed form, and a sign-randomised null reproduces about 95% of the observed value. One model beats HAR for 5-minute volatility, but not the EWMA its input already contains. On strongly planted synthetic signals the models recover part of the signal at 5 minutes, and recovery deteriorates at 30 minutes. Benchmark-level success of TSFMs should therefore not be assumed to carry over to low-signal domains such as intraday returns.
Read the full abstract

Machine Learning

Visible Tails: Characterizing Adaptation to Unknown Moments in Heavy-Tailed Bandits

Heavy-tailed bandit learners must tell a poor arm from an arm whose favorable rewards have not yet appeared. We characterize this obstruction through a scale of one-sided visibility conditions on the optimal reward law: the mean hidden in a rare event is at most a constant multiple of a retained central moment. A quantile-trimmed UCB index that is given the visibility constants attains arm-specific moment rates at every exponent, with no lower bound on the exponent. Conversely, every strategy with a fixed exploration budget has polynomial regret on each law that violates the corresponding condition. For the variance benchmark we determine the order of both budget constants in the visibility level. The frontier depends on the largest exponent the benchmark promises, and the resulting classes are strictly nested. The same index has logarithmic expected regret when rewards are integrable, with a coefficient that can be exponential in the inverse gap.
Read the full abstract

Robotics

BioMARS 2.0: a robotic biologist for formulating and adapting biological experiments

Biological experimentation is not reducible to a single optimization problem, but requires decisions at multiple levels of abstraction: translating scientific knowledge into experimental procedures, adapting culture conditions to measured phenotypes and selecting interventions that test biological hypotheses. Automating experimental execution does not, by itself, determine which experiments should be performed next or how they should change in response to biological feedback. Here we introduce BioMARS, a human-in-the-loop multi-agent robotic system that connects knowledge-guided experimental decision-making with robotic execution and biological feedback. At the procedural level, BioMARS generated literature-grounded protocols and traceable action plans, improving expert-reviewed protocol quality scores across four model backends. At the parametric level, BioMARS used image-derived phenotypes from robotic retinal organoid culture to select subsequent experimental conditions, with higher reported scores in later rounds of the three-round experiment. At the intervention level, BioMARS nominated drug-repurposing candidates from biomedical evidence and advanced six candidates to wet-lab testing by researchers. Together, these studies demonstrate feedback-driven robotic biological experimentation within a framework that supports decisions about experimental procedures, culture conditions and biological interventions.
Read the full abstract

All authors Yibo Qiu, Zan Huang, Yilin Qiao, Ziyang Ling, Yi Wang, Zhengyang Lu, Hangke Peng, Haoliang Ye, Ao Tian, Shu'ang Sun, Handi Liu, Yuqing Long, Mohan Yang, Kan Ding, Zhenzhen Zhou, Di Wu, Yu Xiong, Yaxuan Han, Jun'ao Song, Haoran Zhang, Ronald X Xu, Mingzhai Sun

Solar and Stellar Astrophysics

An assessment of the Weak Field Approximation

The Weak Field Approximation (WFA) is widely used to infer magnetic fields from solar spectropolarimetric observations. However, its validity and sensitivity to atmospheric conditions are not fully understood. We investigate the reliability and limitations of the WFA using Generalized Response Functions (GRFs), which quantify the sensitivity of the inferred magnetic field to atmospheric perturbations. We derive analytical expressions for the WFA GRFs and compute them from synthetic Stokes profiles generated with the SIR and DeSIRe radiative transfer codes. We consider photospheric and chromospheric lines under different magnetic field strengths and inclinations. The inferred magnetic field depends on the atmospheric structure, field inclination, and wavelength range. For photospheric lines, the WFA increasingly departs from the actual field strength well before reaching the saturation value, due to the loss of linearity between Stokes V, Q, and U and the intensity profile derivatives. The transverse-field inference is particularly sensitive to the wavelength range, with line wings providing more reliable estimates. For chromospheric lines, the core and wings probe different atmospheric layers, with the wings providing reliable estimates of the photospheric field. The WFA has a more limited range of validity than commonly assumed. Its reliability depends on field strength, inclination, and wavelength selection, particularly for the transverse component. Chromospheric line wings can also provide useful information on the photospheric magnetic field. These findings support the original perturbative formulation of the WFA.
Read the full abstract

All authors D. Arroyo Caballero, J. C. del Toro Iniesta, D. Orozco Suárez, A. L. Siu-Tapia

Machine Learning

Good Views Are Not Enough: Global Counterexamples for Gaussian-Regularized JEPAs

Good views do not suffice to guarantee scene invariance or uniqueness up to rotation at the global optima of Gaussian-regularized JEPAs. We give globally certified counterexamples showing that the uniqueness premise of the Gaussian extension in Good Views fails for specified soft objectives. Gaussian marginal matching can reward retaining independently resampled observation variation, and this benefit can exceed its prediction cost. In a four-state model, comparison against the entire scene-only class proves that every global minimizer splits a scene and admits an equal-value partner outside its rotation orbit. In a 512-dimensional model with a complete scene basis, a spectral gap, and exact population whitening, pooled Epps-Pulley regularization still forces every global minimizer to retain nuisance and yields positive cross-modal alignment error even after the best rotation when modality nuisances are independent. Better scene alignment can also accompany worse prediction. With matched 512-label readout budgets and 1% observation noise, whitened prediction JEPA aligns better to the scene than next-token prediction but incurs 0.2872 nats/token higher linear-readout test NLL. A Bayes-risk decomposition distinguishes information lost through the observation channel from the fitted readout's difficulty using retained information. These results connect three determinants of learned representations: what views share, what regularization favors, and what a prediction task can use.
Read the full abstract

Robotics

LHNav: Long-Horizon Humanoid Navigation via Multirate Coordination and Physics-Grounded Action Data

We study humanoid navigation across multiple destinations under long-horizon language instructions. This task is challenging because robots must maintain progress across successive goals while continuously producing natural, stable whole-body motion through confined spaces. However, task decomposition in existing methods does not by itself ensure progress throughout execution, while waypoint-based navigation does not guarantee natural, route-faithful whole-body motion. Our key idea is to organize navigation into three complementary roles operating at different time scales: a Brain for task reasoning, a Cerebellum for local motion generation, and a Brainstem for whole-body execution. We realize this idea in LHNav, an asynchronous system that uses progress and arrival feedback to coordinate navigation goals, VLA motion-token prediction, and whole-body control. To train the Cerebellum for natural, route-faithful motion, we develop a physics-grounded data engine that improves physical tracking of motion references along collision-aware routes and pairs the resulting demonstrations with rendered visual observations. Together, these designs allow task planning to proceed alongside continuous movement and extend single-goal navigation to successive destinations with natural, route-faithful whole-body motion. In simulation, our navigation VLA outperforms OmniNav and DualVLN in single-goal success and motion smoothness, with fewer collisions and falls. Using a VLA trained only on single-goal demonstrations, LHNav achieves a 75.56% success rate on long-horizon navigation tasks. Deployment on the Unitree G1 further demonstrates single-goal navigation and zero-shot long-horizon task execution in real-world environments, including layouts with repositioned objects. Project website: https://robbyant-research.github.io/LHNav/.
Read the full abstract

All authors Xueyang Wang, Zehong Shen, Dingkun Wei, Huaijin Pi, Changpeng Wang, Xing Zhu, Hujun Bao, Yujun Shen, Xiaowei Zhou

Instrumentation and Methods for Astrophysics

Particle Swarm Optimization Strategies for Lensed Gravitational Waves: Parameter Estimation, Source Classification, and Accelerated Sampling

The detection and parameter estimation of lensed gravitational wave (GW) signals are crucial for astrophysics and cosmology. Strong lensing introduces extra parameters, rendering the search high-dimensional, multi-modal, and strongly nonlinear. We apply Particle Swarm Optimization (PSO) to the search, parameter estimation, and classification of lensed compact binary coalescence GWs within a four-detector coherent network. Using the restricted 2-PN waveform as a template and a model-independent double-image lens parameterization, we assess PSO on both model-matched 2-PN signals and template-mismatched IMRPhenomD signals. Results show effective recovery of chirp mass, mass ratio, sky location, and lens parameters in the 2-PN case, with mass biases appearing under waveform mismatch. A three-class classification pipeline combining 12 PSO runs, majority voting, and a flux ratio threshold achieves 86.19% accuracy in distinguishing lensed GWs, unlensed GWs, and noise. We further propose PSO_GAUSS_MIX method, which constructs a Gaussian mixture prior from PSO estimates for nested sampling, yielding speedups of 1.38x in the restricted 2-PN cases while preserving posterior accuracy. These results establish PSO as an efficient optimizer for lensed GW searches and an effective prior constructor for Bayesian parameter estimation.
Read the full abstract

All authors Weiqing Yu, Meng Zhang, Xiaobo Zou, Xiao Guo

Optimization and Control

Ride-time inequalities for the event-based formulation of the dial-a-ride problem

The paper investigates the model structure of the event-based mixed-integer programming formulation of the Dial-a-Ride Problem. We find that the design of events allows introducing new valid inequalities on the ride times. Time progression is modelled using Big M constraints which lead to a weak LP relaxation and deteriorate the solution process. The newly introduced valid inequalities deal with this issue and allow reducing the computation times by up to 99% for benchmark instances (a-X* instances) in comparison to the state-of-the-art event-based model formulation. To evaluate the potential of the new valid inequalities, we increase the time windows on the benchmark set by another 15 minutes and evaluate the model on the same instances with a time-related objective function where the Big-M-influenced time variables directly occur in the objective function. Even for the large time windows only two instances could not be solved to optimality within two hours (final gaps of at most 0.8%). With the time-oriented objective the largest average computation time with our valid inequalities was 74 seconds.
Read the full abstract

High Energy Physics (Theory)

Rotating the Replica: Twisted Rényi Coefficients from CFT Operator Counting

We show that twisting vacuum Rényi entropies by a spatial rotation reveals conformal field theory data governing the asymptotic counting of spinning operators. For a ball in four dimensions, the logarithmic coefficients of these angular-momentum-twisted Rényi entropies are determined by rotating-sphere thermal data to all orders in the local derivative expansion, including the Weyl-anomaly contribution. We test this relation in free theories and in strongly coupled, large-$N$ $\mathcal N=4$ Yang–Mills theory using holography.
Read the full abstract

Disordered Systems and Neural Networks

Spatial effects of Kuramoto modeling on the power-grid frequencies

Spatial coupling and heterogeneity can generate complex power-grid frequency dynamics. We investigate the contributions of spatial structure, intrinsic nonlinear dynamics, and external driving using the Kuramoto model (KM), with and without uncorrelated noise, and compare its predictions with real-grid measurements. We analyse one month of 100 ms frequency measurements from 16 Hungarian transmission-grid sites and simulate the KM on a 387-node transmission network and two-dimensional lattices. The measured power spectra exhibit two scaling regimes, with exponents of approximately 1.63 and 2.23. One-second frequency increments decorrelate rapidly, while correlations of their local inverse variance decay slowly, initially as a power law, $\tau^{-0.26(2)}$, up to 15 minutes. A superstatistical Gaussian mixture based on measured local variances captures much of the non-Gaussian increment distribution, but not the rare extreme events. We also estimate the non-stationary long-time frequency autocorrelation exponent. KM simulations on low-dimensional networks exhibit algebraically decaying autocorrelations and power spectra, consistent with the measurements and the Wiener-Khinchin relation in the steady state. On the heterogeneous transmission network, partial synchronization and frequency locking produce persistent spatial patterns and multi-peaked frequency distributions. However, neither deterministic KM dynamics nor independent Gaussian driving reproduces the measured spectra and increment statistics simultaneously. Moreover, empirical frequency fluctuations are largely shared across substations, whereas simulated fluctuations mainly reflect relative node motion. These findings suggest that spatially coupled nonlinear dynamics contribute to observed frequency fluctuations alongside external driving, including 15-minute power dispatches.
Read the full abstract

All authors Kristóf Benedek, Michelle T. Cirunay, Sangita Dutta, Attila Balogh, Bálint Hartmann, Géza Ódor

Machine Learning

One-Shot Federated Graph Learning via High-Fidelity Proxies and Transferability-Guided Collaboration

One-shot Federated Graph Learning (OFGL) has emerged as a communication-efficient paradigm for federated graph learning under strict bandwidth constraints, compressing collaboration into a single transmission round. Nevertheless, this extreme setting suffers from a fundamental Single-Transmission Bottleneck, which limits both the quality and utility of exchanged knowledge. First, while existing methods attempt to distill private graphs into shareable proxies, their aggressive one-shot compression inherently triggers a fidelity crisis. This compression distorts local graph topology and semantics across clients, leading to externalized knowledge underexpression. Second, although the server aggregates these proxies to update global models, current integration schemes suffer from transferability agnosticism. The server cannot reliably identify cross-client collaboration potential, thus resulting in integrated knowledge underutilization. To tackle these dual challenges, we propose FedHFT, an effective One-Shot Personalized Federated Graph Learning (OPFGL) framework that addresses this bottleneck from two perspectives, namely High-Fidelity Knowledge Externalization and Transferability-Ware Knowledge Integration. We use Fidelity-Driven Proxy Refinement to preserve client-side knowledge fidelity, and Transferability-Steered Personalized Collaboration to enable adaptive server-side collaboration. The superiority of FedHFT is validated through extensive experiments on both homophilic and heterophilic graphs. The code is available at https://github.com/Jumpy-Juice/FedHFT.
Read the full abstract

All authors Jia Jiang, Zihan Tan, Wenke Huang, Bin Yang, Mang Ye

Robotics

Compact and Wide-Force-Range Magnetorheological Hand Exoskeleton for Kinesthetic Feedback in Teleoperation

Hand kinesthetic feedback is crucial for restoring and enhancing human perception of force and displacement in robotic hand teleoperation. Wide perception range of force feedback is vital in rendering objects of different stiffness, especially for high-stiffness rendering. However, existing approaches struggle to achieve wide range of force feedback while maintaining compact and wearable design. This paper presents a hand exoskeleton system employing magnetorheological (MR) brakes to provide a wide range kinesthetic feedback during teleoperation. The MR brake provides a braking torque ranging from 0.01 to 0.918 Nm, while maintaining a compact structure with a total mass of only 39.8 g. The proposed exoskeleton can render fingertip forces up to 21 N while maintaining back-driving force of 1.01 N. Experimental results show that the system allows users to discriminate object stiffness using kinesthetic feedback alone. In a tele-grasping task, when both conditions achieved a 100% success rate, kinesthetic feedback led to a 28.2% reduction in object contact force compared to visual feedback only, enhancing object safety and operator immersion. Questionnaire results further indicate that kinesthetic feedback improves reliability, realism, and user confidence during teleoperation.
Read the full abstract

All authors Xianlong Mai, Rui Yu, Rui Chen, Daniele Leonardis, Wenbo Li, Domenico Chiaradia, Antonio Frisoli, Shuaishuai Sun

Computer Vision

RAPA: Ray-Aware Physics Augmentation for Robust Camera-LiDAR Calibration

Camera-LiDAR calibration is essential for reliable perception in robotic and autonomous systems. Most target-based calibration methods rely on checkerboard observations, but accurate checkerboard feature extraction from LiDAR point clouds remains challenging due to measurement sparsity. Existing approaches either require manual annotation of 3D corners or depend on geometric feature extraction methods, both of which can become unreliable when checkerboard features are insufficiently observed. In this paper, we formulate checkerboard localization as a 2D detection problem using cylindrical LiDAR reflectivity images. To overcome the limited availability of labeled training data, we propose ray-aware physics augmentation (RAPA), which generates diverse synthetic samples from a single checkerboard observation. The generated data are used to improve detector robustness and generalization under varying sensing conditions. Across four independently captured indoor and outdoor locations, the proposed synthetic augmentation exceeds the strongest evaluated image-space augmentation baseline by 16.1-43.1 percentage points in mAP@0.5. Compared with training-free target-based methods on 242 observations, RAPA yields the largest number of valid target localizations, with 96.4% of its returned detections yielding valid target poses. These reliable localizations support downstream camera-LiDAR calibration across all four evaluated locations. Overall, RAPA addresses a key bottleneck in target-based calibration by enabling automated LiDAR-side target localization with only a single annotated target observation.
Read the full abstract

All authors Kyung-Rin Oh, Ji-Hoon Kim, TaeJoon Ha, Jae-Sang Hyun

Artificial Intelligence

Toward Sustainable Agentic AI: Carbon Tradeoffs Across Iterative Self-Refinement and Tool-Augmented Coding

Agentic language model systems can improve performance either by iterating on their own outputs or by using external tools in grounded environments. While these strategies can reduce error rates and substitute for larger models in some settings, their environmental cost remains relatively underexplored. We present two complementary empirical studies that estimate the energy and carbon footprint of agentic AI across settings of increasing complexity. Study~1 evaluates iterative verbal self-refinement: GPT-4.1-mini at $k = 1$–$5$ steps versus a GPT-4.1 single-pass baseline on two MATH benchmark splits. Study~2 evaluates tool-augmented agentic coding: GPT-4.1-mini, Claude Opus~4.6, and GPT-5.3-codex on SWE-bench Lite, where agents operate inside Docker-isolated environments with access to code execution and file editing tools. In Study~1, $k=3$ refinements close a 16-percentage-point accuracy gap to within 1~point, but nearly triple the carbon footprint relative to the cheap single-pass baseline while remaining below the GPT-4.1 single-pass baseline under the current token-based carbon scenario. In Study~2, GPT-5.3-codex achieves the highest resolution rate in this sample (81.2%), ahead of Claude Opus~4.6 (68.9%) and GPT-4.1-mini (25.0%), but resolved-task-normalized efficiency remains mixed: GPT-4.1-mini is lowest on estimated carbon and cost per resolved task, while GPT-5.3-codex delivers substantially higher task success. All carbon estimates cover output-token inference only; prompt tokens, which exceed 99% of tokens in Study~2, and tool execution are not charged, so the figures are lower bounds on the full agentic footprint.
Read the full abstract

Computation and Language

Lost in Rounding: bf16 Recurrent State Makes Hybrid Reasoning Models Think Longer

Hybrid reasoning models interleave attention with recurrent mixers such as Mamba-2 and Gated DeltaNet, whose state is cached between decoding steps. Prior work showed that storing this state in FP16 or INT8 with round-to-nearest (RTN) makes Mamba-2 and GDN hybrids more verbose, and that stochastic rounding (SR) fixes it. We study the other 16-bit format that serving engines expose, bf16, on six open hybrid reasoning models and 590 competition-math problems. With bf16 RTN state, five of six models reason longer (geometric-mean length 1.07x to 1.27x; also under greedy decoding in the two models tested) and hit the 32K token cap more often. On Qwen3.5-9B, cap hits rise from 118 to 223 and accuracy falls from 85.6% to 70.8%. On LiveCodeBench, both models tested also reason longer and lose 12.6 and 13.4 points of pass@1. Rerunning the capped items with a 128K budget halves the accuracy gap but does not close it (93.6% vs. 86.6%): bf16 runs that never finish are twice as common, and 85% of them are exact repetition loops. On Nemotron-Nano-9B-v2, by contrast, 30 of 31 extra caps finish at 128K, so the cost there is mostly extra tokens (exploratory). Same-width bf16 SR removes the length increase in four of five affected models, restores Qwen3.5-9B to 87.6% and brings code length and pass@1 near FP32 levels, so the failure is mostly rounding bias, not missing resolution. With an integer-hash random generator, SR on Qwen3.5-9B matches FP32 on cap hits and accuracy (113 vs. 118; 85.6% for both) at the memory and decoding speed of bf16 RTN. In the Nemotron-Nano-v2 lineage, SR on 25% of heads (an intermediate-decay band) removes the effect, while random heads and the top teacher-forced-KL octiles do not; in Nemotron-3 and Qwen3.5-9B the effect is spread across heads. We recommend that evaluations of hybrid models report the state dtype, cap rate and output length alongside accuracy.
Read the full abstract

Quantum Physics

Modular and Looping Quantum Photonic Processing Unit for Quantum Computing

Photonic processors are emerging as promising platforms in the race towards the realization of practical universal quantum computers. Several architectures with universal capabilities have been proposed for their implementation via a photonic platform, each with their own set of benefits and drawbacks. In this work we propose an architecture for a universal photonic quantum computer. The proposed architecture is modular, allowing for a high degree of flexibility and scalability, and relies on a loop-based structure, significantly reducing the number of physical components required to implement it experimentally. We prove that our architecture is universal, deterministic, and compatible with the quantum circuital model, as well as reprogrammable and capable of adaptive measurements. Furthermore, we showcase its functioning via numerical simulation on specific examples, by using machine learning techniques on a set of fundamental linear and nonlinear optical components to obtain some of the building blocks used in the architecture. The obtained results can be used to highlight the benefits and the drawbacks of the proposed architecture, as well as the route for future developments towards its physical implementation.
Read the full abstract

Machine Learning

CARE-PO: Controllable Tool-Call Efficiency for LLM Agents

External tool calls are a major running cost of an LLM agent, while its training objectives align almost exclusively with task completion. In a step-level diagnosis of Qwen3.5 trajectories at two model sizes on GAIA, a long-horizon tool-use benchmark, we find that most contain removable calls, at a similar rate at both sizes. Some methods add an efficiency reward to the objective to reduce the call count, but introduce problems of their own: the anchor they price calls against is computed from the policy's own rollouts (e.g., the group's shortest correct rollout) and so moves with the policy; training can drift toward a collapsed policy that issues no tool call at all, a degeneracy seen in published reports and in our 27B baseline; and a training run delivers a single operating point, so the user cannot trade efficiency against capability. We present CARE-PO (Controllable Anchored Relative Efficiency Policy Optimization) to mitigate these problems. The efficiency reward prices calls against a reference budget anchored to the task's own difficulty and fixed before training. Injecting a successful trajectory from the training data into each online GRPO rollout group and charging for evidence a trajectory never gathers ease the drift toward a collapsed call count. Shaping the cost curve by an effort tag in the prompt yields an efficiency and a capability operating point from a single training run. On GAIA, five of the six operating points across three model sizes improve on their base model in success rate and call count at once. On four held-out multi-hop benchmarks the result is size-dependent: at 27B and 4B success rate is held or raised at a lower call count, while at 9B both operating points trade success rate for fewer calls. The call count remains well above zero throughout training at all three sizes.
Read the full abstract

All authors Xiaoyang Ye, Xiao Yao, Hairui Zhu, Yiying Yang, Bei Gan, Jinliang Lu, Tengjin Weng, Yuan Yao, Wenhao Jiang

Audio and Speech Processing

SCLAP: Temporal structural alignment improves localization in long-form audio

Audio-text models such as Contrastive Language-Audio Pretraining (CLAP) typically pool frame and token sequences uniformly into global embeddings. This discards temporal structure and impairs both retrieval and localization, especially for short events in long recordings. We introduce SCLAP, a sequence-level contrastive framework that aligns audio frames and text tokens using monotone one-dimensional optimal transport. Its learned marginals and transport plan provide fine-grained temporal supervision from clip-level captions alone. On short-form audio (i.e., Clotho and TACOS datasets with 15-30s recordings), SCLAP remains competitive for retrieval while improving zero-shot localization over CLAP and Text-aware Attention Pooling. On long-form audio (i.e., CASTELLA dataset with up to 300s recordings), SCLAP improves retrieval performance and its localization advantage grows as the event-to-context ratio decreases. At inference, SCLAP requires no optimal transport computation since retrieval can use pooled embeddings and localization can use inexpensive frame-text cosine scores.
Read the full abstract

Cosmology and Nongalactic Astrophysics

CMB Bounds on Primordial Black Holes via Radiation Capture: Linear Perturbations

We investigate the gravitational interaction between primordial radiation, consisting of photons and ultra-relativistic neutrinos, and primordial black holes (PBHs), and study its imprint on the anisotropies of the cosmic microwave background (CMB). Assuming that PBHs are immersed in the cosmic radiation background, we introduce a covariant energy-momentum transfer four-vector to describe the gravitational capture of radiation by PBHs. We extend the background-level framework of M. Farhang and S. M. S. Movahed [Phys. Rev. D 114, 083533 (2026)] to an inhomogeneous Universe at linear order by consistently modifying the perturbed continuity and Euler equations to account for both energy and momentum transfer. We use the imprint of this interaction on the CMB to place novel constraints on the PBH abundance, $f_{\rm pbh}$. We show that these constraints are up to two orders of magnitude tighter than those obtained from background-only analyses and exhibit a strong dependence on the PBH mass. Using Planck temperature and polarization data, we constrain the PBH fraction to $f_{\rm pbh} \lesssim 0.05-0.002$ for $M$ in the range $10^{15}-10^{17} M_\odot$. Furthermore, sensitivity forecasts for a cosmic-variance-limited survey with $\ell_{\rm max}=7000$, assuming a fiducial $f_{\rm pbh}=0$, yield projected bounds of $f_{\rm pbh} \lesssim 10^{-2}-2\times 10^{-4}$ over the mass range $10^{13}-10^{17} M_\odot$.
Read the full abstract

Artificial Intelligence

System Optimizations for Cost-Efficient Agent Harnesses

AI agent harnesses play the critical role of providing LLMs with an interface to a computing environment. Recent studies show that AI agents' task-execution accuracy depends mostly on the LLM used, while harness-side variances such as system prompts or tools have minimal impact. Rather, such harness-side variances show a larger impact on the cost of agent executions, achieving cost reductions up to 2.6 times on the same model and task. To further explore these harness-side cost optimization opportunities, we analyze the cost accumulation patterns across various agent traces and identify system opportunities to reduce token use and LLM calls without modifying the model or affecting task accuracy. Based on this analysis, we design Agency, a harness interposition platform that provides a unified abstraction and allows easy and portable implementation of harness-side system optimizations on existing off-the-shelf harnesses such as Claude Code or OpenAI Codex. Using Agency, we implement proof-of-concept optimization techniques based on our analysis and achieve cost reductions up to 35% across various real-world benchmarks on existing harnesses without compromising accuracy.
Read the full abstract

All authors Jongseok Park, Jiaxiang Yu, Eric Zhou, Tony Hong, Sunga Kim, Yusuf Afifi, Ian Wong, Ion Stoica, Alvin Cheung

Solar and Stellar Astrophysics

Enhanced Mid-Photospheric Heating Around a Solar Pore

We present depth-stratified atmospheric reconstructions of a solar pore observed with the Sunrise Chromospheric Infrared SpectroPolarimeter (SCIP) aboard Sunrise~iii. Using spatially coupled NLTE spectropolarimetric inversions, we characterize the thermal, magnetic, and velocity structure of the pore and its surroundings from the photosphere to the chromosphere over approximately two hours. The inversions recover the canonical photospheric properties of pores: a cool umbra with strong magnetic fields dominated by the vertical component. Most photospheric velocities are subsonic, but localized downflows near the pore boundary reach transonic or potentially supersonic values around $\log\tau_{500}=0$. At higher atmospheric layers, the thermal contrast between the pore and its surroundings decreases substantially. We also detect localized mid-photospheric temperature enhancements forming a patchy, ring-like structure in the wings of the Ca\,ii 8542\,Å line, without a clear continuum counterpart. The temperature excess reaches approximately $300$\,K and corresponds to an excess thermal-energy density of up to $170$\,J\,m$^{-3}$. These regions coincide with an enhanced transverse magnetic-field component and the expanding magnetic canopy surrounding the pore. The structure is detected in all three rasters, although the cadence does not establish whether individual patches persist or repeatedly re-form. While the heating mechanism cannot be uniquely identified, the inferred magnetic topology supports an association with processes operating in the expanding canopy. These results demonstrate the value of pores for investigating the coupling between magnetic topology, convection, and atmospheric energy transport.
Read the full abstract

All authors Ana Belén Griñón-Marín, Adur Pastor Yabar, Jaime de la Cruz Rodríguez, Sami K. Solanki, Andreas Lagg, Achim Gandorfer, Jose Carlos del Toro Iniesta, Yukio Katsukawa, Pietro Bernasconi, Thomas Berkefeld, Alex Feller, Tino L. Riethmüller, Alberto Álvarez-Herrero, Masahito Kubo, H. N. Smitha, David Orozco Suárez, Bianca Grauf, Michael Carpenter, Alexander Bell, Valentín Martínez Pillet, Francisco Javier Bailén, Julian Blanco Rodríguez, Juan Sebastián Castellanos Durán, Edvarda Harnes, Johannes Hoelken, Francisco A. Iglesias, Ryohtaroh T. Ishikawa, Yusuke Kawabata, Takuma Matsumoto, Takayoshi Oba, Azaymi L. Siu-Tapia, Hanna Strecker, Dušan Vukadinović, Hirohisa Hara, Toshifumi Shimizu

Artificial Intelligence

SCALE: Small-Model-First Concept Abstraction with LLM-Guided Experience Internalization

Autonomous agents must continually acquire reusable knowledge from experience without relying on costly large language model (LLM) reasoning for every decision. However, existing experience-based approaches often struggle to transform accumulated observations and reasoning into compact knowledge that lightweight models can independently apply. This paper proposes SCALE, a small-model-first framework that integrates selective LLM reasoning, experience-driven concept abstraction, and knowledge internalization. A lightweight model handles routine predictions, while prediction errors and accumulated cross-episode evidence selectively trigger LLM reflection. The LLM analyzes relevant historical experiences, generates positive and negative counterfactual examples, and proposes reusable concept hypotheses. Independently validated concepts are progressively incorporated into hierarchical neural representations, enabling lightweight models to exploit increasingly abstract knowledge. Meanwhile, mastered experiences are excluded from active memory while their original records remain available in a complete archive. Experiments across ten random seeds show that historical recall improves held-out accuracy from 87.11% to 98.28%, hierarchical concept knowledge increases difficult-case accuracy from 55.33% to 66.47%, and abstract archiving reduces active experience records from 6,000 to 1,528.
Read the full abstract

Audio and Speech Processing

Discriminator-free adversarial domain adaptation for robust speech classification

The performance of speech classification systems such as speech emotion recognition or pathological speech detection can degrade substantially when there is a domain shift between training and test data. This shift can arise due to e.g., differences in recording setup, acoustic environment, speaker, language, or speaking style. Adversarial domain adaptation addresses this problem by encouraging representations that are less sensitive to domain-specific differences. However, many adversarial approaches rely on training a separate domain discriminator, which can be unstable and difficult to train. In this paper, we propose to use a discriminator-free adversarial learning network (DALN) for unsupervised domain adaptation in speech classification tasks. DALN reuses the task classifier as an implicit discriminator and aligns domains through the nuclear-norm Wasserstein discrepancy on classifier outputs. We extensively evaluate the performance of DALN on intra-lingual, cross-lingual, and noisy adaptations for speech emotion recognition and pathological speech detection. Extensive experimental results show that DALN generally achieves the best performance across the considered tasks and domain shifts, outperforming discriminator-based domain adaptation approaches.
Read the full abstract

Artificial Intelligence

Joi-Bench-Task: Benchmarking Task Coordination in Full-duplex Voice Agents through Interactive Simulation

Full-duplex voice agents enable natural, hands-free interaction, allowing users to refine requests, redirect ongoing work, and seek clarification while the agent is speaking. However, existing voice-agent benchmarks insufficiently isolate coordination between task execution and conversation. We introduce a benchmark that assesses this coordination by examining whether agents track what is being executed, what has been completed, what has been communicated, and what remains to be reported as the conversation evolves. We propose an interactive user simulator that introduces configurable interventions at designated timing in tool execution and spoken result delivery, supporting controllable and reproducible evaluation conditions. We evaluate five leading commercial voice agents along three dimensions: execution responsiveness, coordination under request changes, and coordination under interruptions. Agent evaluations reveal persistent coordination failures. Under request changes, the highest result completion rate is only 55%, indicating difficulty for current voice agents adapting to new requests while preserving earlier obligations. After user interruption, four of the five agents exhibit average information omission rates of at least 23%. These findings highlight task coordination as a central challenge for current voice agents in complex, evolving conversations.
Read the full abstract

All authors Xiangheng He, Mingjun Zhao, Zhehao Fan, Zhengjia Li, Xiaopeng Sun, Qingqing Xiao, Haolan Chen, Yu Xu

General Economics

Does fare communication format matter? Evidence from a randomised experiment in public transport

Highly concentrated demand in public transport places pressure on infrastructure and reduces service quality. Pricing instruments-such as discounted off-peak single fare tickets-are a widely used tool for rescheduling demand across time. Yet no study has directly examined whether the price communication format of single fare tickets affects travel behaviour in terms of departure-time rescheduling. To fill this gap, we exploit a randomised stated preference survey conducted in Switzerland based on participants' actual journeys (n = 1,082), assigning participants to one of three price formats: absolute amounts, absolute deviations, or percentage deviations from their current fare. Estimating a multinomial logit model with price-format interaction terms, we find no significant effect of communication format on aggregate price sensitivity. However, for low-fare trips, percentage-based communication substantially amplifies price sensitivity relative to absolute amounts or absolute deviations. Moreover, when prices are shown as percentage deviations, price sensitivity increases for small price spreads and decreases for large price spreads, with a crossover point corresponding to a discount depth of approximately 47%. In summary, these findings suggest that percentage-based price communication has the potential to reinforce the demand-shifting effect of off-peak discounts, particularly for low-fare products and for discounts shallower than approximately 47%.
Read the full abstract

Machine Learning

Forecast Value Add: Reward-Based Action Selection over Probabilistic Demand Forecasts

Strong machine-learning ensembles capture much of the trend, seasonality, and known events in demand, yet they can still understate peaks, miss moving events, or mistake temporary changes for shifts in demand level. We address this gap by framing forecast refinement as a reward-driven decision problem. For each forecast period, we let an agent choose among values already produced by the ensemble: its base forecast, distribution-free bounds, quantiles, and a bias-corrected estimate. We evaluate these choices against context-matched historical periods, accounting for calendar position, recurring and moving events, trend, price, and promotions. We then apply a structural seasonal guard to constrain adjustments and record the model, agent, and planner contributions to every published value. We evaluate our approach across five retail, consumer-goods, and direct-to-consumer deployments over nine months. Our refined forecasts reduce mean WMAPE by 6.3% compared with the stacked machine-learning forecast for every customer, with gains reaching 16.7%, and achieve the lowest error in 35 of 45 customer-months. Compared with statistical forecasts, our approach reduces mean error by 18.7%. These results demonstrate that context-aware, auditable action selection can add forecast value beyond a strong ensemble.
Read the full abstract

Applied Physics

Analytical Framework Relating Capacitance-Frequency Measurements and Hysteresis in Perovskite Solar Cells

Mobile ion migration strongly influences the electrical response and current-voltage hysteresis of perovskite solar cells. However, quantitative determination of ionic transport parameters typically requires transient measurements, extensive numerical simulations, and/or fitting procedures involving multiple uncertain parameters. Here, we present an analytical model that establishes a direct relationship between capacitance-frequency (C-f) characteristics and ionic transport properties in perovskite solar cells, while accounting for interlayer effects. The model describes the transition from the ion-screened depletion capacitance at low frequencies to the geometric capacitance at high frequencies and demonstrates that the characteristic transition frequency provides a direct measure of the dielectric relaxation time for ions. Furthermore, the model establishes a quantitative connection between C-f characteristics and scan-rate-dependent current-voltage hysteresis, enabling prediction of the scan-rate window in which maximum hysteresis occurs. Drift-diffusion simulations show excellent agreement with the analytical predictions over wide ranges of ion density, ion mobility, transport-layer mobility and doping concentration. Experimental measurements confirm the predicted correlation between capacitance and hysteresis. By providing a direct, physically transparent, and computationally efficient route for extracting ionic transport parameters from routine impedance measurements, this work offers a practical tool for quantifying ion migration and understanding dynamic phenomena in perovskite solar cells.
Read the full abstract

All authors Manjeet Kumar, Changzeng Ding, Mathias Nyman, Yongqi Sun, Chang-Qi Ma, Ronald Österbacka, Oskar J. Sandberg

Artificial Intelligence

OGDA: Ontology-Grounded Data Agents for Reliable Analysis over Heterogeneous Data Lakes

Full-pipeline data analysis over real-world data lakes requires LLM agents to locate and integrate information across large collections of tables with different schemas. Existing data agents follow task-specific exploration, which exhausts the context window in massive lakes. As the context fills, the agent's reasoning degrades, leading it to locate and process incorrect data files. These errors propagate into the final answer. To overcome these limitations, we introduce Ontology-Grounded Data Agent (OGDA), which utilizes a compact ontology to reduce exploration overhead and constrain data processing. OGDA constructs the ontology through agentic hypothesis testing. At each step, a proposer agent induces schemas, concepts, and relations from raw sources, and a verifier checks each proposal against the underlying data. During task execution, the ontology serves as a representation of the data lake for the agent to explore. Every data-processing action is constrained by a task graph, which the agent can extend only after observing the semantic meaning of the data. Across KramaBench and DiscoveryBench with three LLM backbones, OGDA outperforms the strongest baselines by up to 10.82 points. These results suggest that explicitly modeling the data lake is a key step toward reliable end-to-end data agents.
Read the full abstract

All authors Boyi Zhang, Hong Ting Tsang, Yisen Gao, Qiao Xiao, LI Yufei, Haoyu Huang, Zhongwei Xie, Jiaxin Bai, Yangqiu Song

Robotics

Motion Belief Action: From a Visual Scene Map to Physical Decisions

A robot that works among people must judge what it cannot see: whether an object is still where it was last seen, and whether a pull on a door or drawer is worth its risk. Maps that track object change hold nothing about force, maps of physical properties answer a risk question at a fixed risk level, and vision-language-action models keep no explicit estimate of hidden physics. We propose Motion Belief Action (MBA), a decision layer on the scene map. From how often people move an object and how much force it needs, how much load it holds and whether it is blocked, MBA chooses under a force limit to look again, nudge, open, brace or decline, with its risk set by the task's costs. On real activity in 9 kitchens, it predicts whether an object is still in place better than FreMEn and Perpetua, and decides when to look again at lower cost than these and the persistence filter, because about a quarter of placements end in a pick-up within 30 s. On real pulls of doors and drawers in 17 locations, made by a person with a force-sensing gripper, the map's force estimates decide better than a part-type average. Setting the risk from the costs also beats a published risk rule at its fixed level. In simulation, with the force limit at the training parts' median force, MBA raises SmolVLA's and Octo's scores by declining pulls; most declined door pulls would fail, but on drawers more would have opened, and a part-type average gets about half of this gain. In both, a published risk rule set to the risk these costs imply is MBA's rule for a single pull and decides as well in the main settings. A link between opening and holding force, a probing nudge and an action prior for camera tracking do not help in the main settings, and over the real-pull sweep the nudge costs a little.
Read the full abstract

Software Engineering

LiveWorkspace: Externalizing Visual Understanding as Persistent Shared State for Screenshot-to-Code Agents

Screenshot-to-Code requires coding agents to sustain visual understanding across code generation, execution, and revision. We introduce LiveWorkspace, a visual coding harness that externalizes this understanding as Shared Visual State (Shared VS), an implementation-independent representation of visual hierarchy, node attributes, and layout relations. LiveWorkspace constructs and retains Target from screenshots and recovers Current from browser execution in the same schema. Structured Diff and implementation bindings guide code modifications; user edits update Target and reuse the alignment loop. Using GPT-5.2 for both methods, LiveWorkspace improves end-to-end Avg by 1.8 points on Design2Code and 2.0 on ScreenBench over a control with matched model-call counts and stagewise budget limits. Further experiments demonstrate reuse of constructed Shared VS across coders and improvements in iterative repair. LiveWorkspace also achieves 93.0% edit success across 100 controlled editing cases. These findings support treating visual understanding as a persistent shared working state throughout code generation, execution, and modification.
Read the full abstract

All authors Shijun Xiao, Zonghui Yang, Xinrong Chen, Hao Liu, Wensheng Zhang, Yongqiang Tang