Astrophysics of Galaxies
Precise C-N-O Abundances of Individual Galaxies in the First Billion Years from Ultra-Deep, Rest-UV-to-Optical JWST Spectroscopy
Raunaq Singh Rai, Guido Roberts-Borsani, Ryan Sanders et al. (+8 more)
We present ultra-deep JWST/NIRSpec $R\sim1000$ spectroscopy of nine
star-forming
galaxies at $z=6.0-8.3$ in the GOODS-N field, drawn from the Origins of Metal Enrichment in
Galaxies at
cosmic dAwn (OMEGA) survey. Paired with ancillary ultra-deep G140M observations, the G395M integrations from OMEGA yield continuous rest-frame UV-to-optical coverage and the full suite of diagnostics needed for precise characterisations and measures of the chemical abundances of these systems. Crucially, the auroral [O\,{\scriptsize III}]\,$\lambda$4363 line is detected in every
galaxy, enabling the determination of electron temperatures and direct oxygen abundances of $12+\log(\mathrm{O/H})=7.14-8.18$, independent of strong-line calibrations that dominate systematics at these
redshifts. We couple direct oxygen abundances with measurements of rest-frame UV emission-line strengths and ratios to infer the physical properties of our sample. Most importantly, we trace the relationship between oxygen abundance and both C/O and N/O abundance ratios with high fidelity, demonstrating robustly that $z=6.0-8.3$
galaxies follow the local trend in C/O vs. O/H, and that inferences regarding N/O are sensitive to the choice of rest-frame UV (\niv] or \niii]) or optical ([\nii]) tracers. This pilot study demonstrates the power of combining ultra-deep JWST/NIRSpec $R\sim1000$ rest-frame UV and optical spectroscopy, which now requires statistical samples to uncover the nebular properties of the full population of $z\geq 6$
star-forming
galaxies.
Read the full abstract ▾
All authors Raunaq Singh Rai, Guido Roberts-Borsani, Ryan Sanders, Alice E. Shapley, Leonardo Clarke, Pascal A. Oesch, Daniel Schaerer, Rui Marques-Chaves, Romain A. Meyer, Emma Giovinazzo, Shreya Karthikeyan
Machine Learning
Deep Epistemic Value Functions for Optimistic Exploration
Leander Diaz-Bone, Marco Bagatella, Jonas Hübotter et al. (+1 more)
Principled exploration in reinforcement learning requires an
agent to quantify its epistemic uncertainty and act to resolve it. Uncertainty over the value function provides a natural signal for exploration, yet existing deep approximations remain brittle and perform inconsistently. The central challenge is therefore to scale these ideas robustly. We conduct a systematic empirical study of how epistemic uncertainty is represented, propagated, and optimized in deep epistemic value functions, and uncover distinct failure modes along each of these axes. These findings motivate DEVOTE, a model-free reinforcement learning
algorithm that controls how uncertainty generalizes beyond observed data, stabilizes its temporal propagation, and preserves adaptation to the resulting non-stationary exploration objective. Across reward-free exploration and challenging continuous-control tasks, DEVOTE reaches novel states more effectively and achieves higher task return than strong model-free and model-based exploration baselines. These results provide evidence that deep epistemic value functions are a promising path toward scalable, principled exploration.
Read the full abstract ▾
All authors Leander Diaz-Bone, Marco Bagatella, Jonas Hübotter, Andreas Krause
Signal Processing
On the Universal Approximation Capability of Stacked Intelligent Metasurfaces
Davide Dardari
Wave-domain processing is an emerging paradigm in which signal-processing operations are partially shifted from the digital domain to the electromagnetic (EM) domain to reduce complexity, energy consumption, and latency in next-generation wireless systems. Stacked intelligent metasurfaces (SIMs) are a promising technology for wave manipulation. Although numerous studies have addressed the numerical design of SIMs for specific tasks, no theoretical results have yet established whether a prescribed SIM architecture can approximate an arbitrary complex linear mapping, i.e., whether it possesses a universal approximation capability. To address this question, this paper establishes fundamental
theorems that provide necessary and sufficient closed-form conditions under which a given SIM architecture can serve as a universal approximator of complex linear transformations.
Read the full abstract ▾
Quantum Physics
Induced Coherence in Quantum and Classical Interferometry: Origin and Control
Balakrishnan Viswanathan, Ravi Kamal Pandey, Leena Barshilia et al. (+3 more)
The wave-particle duality that manifests in Young's double-slit interferometry satisfying the well known Englert-Greenberger-Yasin inequality has been shown to obey a universal triality relation involving coherence, predictability and rather surprisingly
entanglement. In the present work, we explicitly demonstrate the persistence of the triality relation in the
quantum induced coherence interferometer and the classical configuration recovers the well known duality inequality. Furthermore, we explore the origin of coherence and
entanglement, in the induced coherence interferometer, and means to control them. We show that path indistinguishability (or intrinsic indistinguishability of the sources) in our setup plays a complementary role in influencing both interference and
entanglement. We bring out manifestly the nature of coherence in both classical and
quantum versions of induced coherence interferometers.
Read the full abstract ▾
All authors Balakrishnan Viswanathan, Ravi Kamal Pandey, Leena Barshilia, Devendra Kumar Mishra, Sibasish Ghosh, Prasanta K. Panigrahi
Combinatorics
Maximizing the discrepancy between zero forcing parameters relative to graph order
Andrew McKay, Allie Ray and Jonathan Valliere
Zero forcing is a process described by a color change rule on the vertices of a graph. In this paper, we maximize the discrepancy between various zero forcing parameters relative to graph order. First, we find an upper bound on the difference in cardinality between minimal zero forcing sets (sets containing no proper zero forcing subset) of maximum and minimum size, and we show that this bound is sharp for an infinite family of graphs. Furthermore, we derive an upper bound for the discrepancy between the maximum and minimum propagation times of the minimum zero forcing sets of any graph, showing this bound is sharp for an infinite family of graphs.
Read the full abstract ▾
Computation and Language
Beyond Token Scale: Chunk-Level Sparse Autoencoders for Reliable Semantic Feature Discovery
Xu Wang, Yifan Yang, TingHao YU et al. (+1 more)
Sparse autoencoders (SAEs) expose features that help us understand and steer
language models, but faithful reconstruction does not guarantee informative concepts.
Token-level objectives reward lexical and formatting details alongside semantic content, all competing for a limited sparse budget. We introduce a family of chunk-level SAEs that encode mean-pooled activations over chunks, each a contiguous span of
tokens: Mean-Chunk reconstructs the observed chunk, Cross-Chunk predicts an independently processed neighbor, and Joint-Chunk combines both targets. These designs separate the effect of a larger observation unit from that of predicting information shared across passages. With matched training data, chunk-level SAEs remain powerful interpretability tools while learning reliable semantic features that capture high-level concepts and respond selectively to relevant content. Their strengths are complementary: Mean-Chunk improves high-level feature discovery,
reasoning detection beyond surface cues, and steering; Cross-Chunk leads document
retrieval and classification transfer while producing selective, persistent features. Changing what an SAE sees and predicts yields reliable semantic features for more meaningful tasks. We demonstrate their practical value through gains across downstream tasks such as
retrieval,
reasoning detection, and steering.
Read the full abstract ▾
All authors Xu Wang, Yifan Yang, TingHao YU, Difan Zou
Classical Analysis and ODEs
Jackson and Bernstein Inequalities with Explicit Weight Dependence for Band-Limited Wavelet Approximation in Weighted Besov Spaces
Hung-Lung Lin and Kai-Cheng Wang
We prove direct and inverse estimates for band-limited biorthogonal wavelet systems in weighted Besov spaces with Muckenhoupt weights. On functions recovered from their wavelet expansion along the dyadic tree of the unit cube, the error of approximation by a prescribed number of terms decays at a rate set by the smoothness-to-dimension ratio, and this exponent cannot be increased. Under a multiresolution condition, Bernstein inequalities hold at a single scale and on every multiresolution subspace, and the linear approximation spaces are identified with weighted Lebesgue–Besov intersections. All constants are explicit; the weight enters only through displayed powers of the Muckenhoupt characteristic.
Read the full abstract ▾
Instrumentation and Detectors
A high-flux electron detection system to measure non-linear Compton scattering at LUXE
Antonios Athanassiadis, Oleksandr Borysov, Florian Burkart et al. (+16 more)
This paper presents the development and testing of a high-flux electron detection system designed to measure the energy spectrum of more than $10^9$ electrons from
laser-electron collisions at the LUXE experiment, planned at DESY. The system is intended to enable precision measurements of non-linear Compton scattering in the strong-field regime of
quantum electrodynamics. It combines a scintillating screen observed by a camera system with a spatially segmented detector based on the Cherenkov effect. Prototype tests were conducted at two accelerator facilities, where the performance of the system was evaluated under realistic beam conditions. The results demonstrate the feasibility of the detector concept, characterize its current performance and limitations, and provide input for the optimization and final design of the detection system to be used in the experiment.
Read the full abstract ▾
All authors Antonios Athanassiadis, Oleksandr Borysov, Florian Burkart, Hannes Dinter-Habermeier, John Hallford, Beate Heinemann, Louis Helary, Luke Hendriks, Ruth Jacobs, Max Kellermeier, Willi Kuropka, Jenny List, Frank Mayet, Gudrid Moortgat-Pick, Evan Ranken, Stefan Schmitt, Ivo Schulthess, Thomas Vinatier, Matthew Wing
Complex Variables
Dual slice functions and dual quaternionic polynomials
Giulio Binosi and Caterina Stoppato
This work constructs a new function class over the
algebra of dual quaternions: dual slice functions, which include one-sided dual quaternionic
polynomials. The peculiar properties of dual slice functions therefore provide new tools to understand factorization of dual quaternionic
polynomials, which was extensively studied in recent years for its applications to mechanism science. Dual slice functions are strictly related to the well-established theory of slice and slice regular functions over quaternions and dual quaternions.
Read the full abstract ▾
Machine Learning
Reward-Aligned Reweighting for On-Policy Distillation
Haofeng Xu, Junwei Su, Lansong Diao et al. (+2 more)
On-policy
distillation (OPD) trains a student
language model with dense feedback from a stronger teacher on student-generated trajectories. Yet standard OPD weights
token-level
distillation terms uniformly, implicitly treating local teacher preference as a proxy for correction utility. A decision's task value, however, depends on how the student completes the subsequent
reasoning. This mismatch can cause imitation to suppress viable student strategies or reinforce paths the student cannot reliably execute. Verified trajectory outcomes provide complementary evidence about continuation quality, but do not directly identify the utility of individual decisions. We introduce Reward-Aligned Reweighting for On-Policy
Distillation (R$^{2}$-OPD), which uses outcome agreement and the magnitude of teacher–student disagreement to continuously reallocate teacher supervision. It gives reward-aligned corrections greater relative influence while retaining dense feedback, moving beyond uniform imitation and hard filtering. Our analysis formalizes the mismatch between local teacher preference and student continuation value and establishes sufficient conditions for reallocation to improve first-order task progress over uniform OPD. Across seven mathematical
reasoning benchmarks, R$^{2}$-OPD achieves the highest average accuracy among the compared training methods in both cross-size and same-size
distillation. It outperforms standard OPD on all seven
benchmarks, with average gains of 3.5 and 2.4 percentage points for 1.7B and 4B students, respectively. An extension to code generation yields an average gain of 1.6 percentage points over standard OPD. These results highlight outcome-guided supervision allocation as an effective way to translate dense teacher feedback into stronger student performance across model scales and task domains.
Read the full abstract ▾
All authors Haofeng Xu, Junwei Su, Lansong Diao, Wenchao Zhou, Chuan Wu
Robotics
Inspection-SPARS: Task-Oriented Sparse Roadmaps for Inspection Planning
Adir Morgan, Oren Salzman and Kiril Solovey
Inspection planning seeks a minimum-length collision-free
robot tour that observes a given set of points of interest (POIs). Sampling-based methods reduce this continuous problem to a graph inspection planning (GIP) problem over a discrete roadmap, which is then solved using combinatorial solvers. Dense roadmaps capture diverse inspection viewpoints and motion shortcuts, and thus admit higher-quality solutions, but they induce large combinatorial search spaces on which state-of-the-art GIP solvers struggle to find good solutions within practical time budgets. Roadmap sparsification—restructuring a dense roadmap into a compact representation that preserves connectivity and path lengths—can alleviate this burden. However, existing sparsification approaches are either agnostic to the underlying inspection task, or strive to ensure coverage of the POIs without accounting for the quality of the resulting inspection plan. We present Inspection-SPARS, which is, to our knowledge, the first inspection-roadmap sparsifier with POI coverage and path-quality guarantees relative to the dense roadmap. To this end, we generalize the SPARS framework, a popular task-agnostic sparsifier, from purely
geometric criteria to task-oriented ones, introducing an inspection-aware vertex admission mechanism that treats POI coverage as a first-class sparsification criterion alongside connectivity and path quality. Experiments in realistic 3D environments show that Inspection-SPARS reduces vertex and edge counts by 4-8x while preserving coverage, allowing the GIP solver to compute tours up to 25% shorter than with the dense roadmap or state-of-the-art inspection roadmap. More broadly, Inspection-SPARS shows that sparsification can be made task-aware without sacrificing guarantees on solution quality.
Read the full abstract ▾
Artificial Intelligence
MechBench: Can AI Scientific Agents Discover Mechanisms Beyond Phenomenal Laws?
Zihan Yu, Jiadong Zhang, Jialin Cheng et al. (+2 more)
Scientific discovery requires not only recovering mathematical laws that describe observable behavior, but also identifying the mechanisms that generate them. Existing
benchmarks for symbolic regression and scientific
agents primarily evaluate phenomenal-law recovery, leaving mechanism discovery largely untested. We introduce MechBench, a
benchmark that explicitly separates these two capabilities. Each task is defined by a mechanistic model, a structured set of scientifically meaningful relations whose joint consequences entail an observable phenomenal law, while
agents receive only observational data and scientific context. We evaluate mechanism recovery through mechanism probes, which query internal scientific consequences that cannot be inferred from the phenomenal law alone. To reduce reliance on memorized textbook mechanisms, we construct unfamiliar variants through controlled, scientifically interpretable mutations of canonical mechanisms, and screen for mechanistic indistinguishability to exclude ambiguous instances admitting comparable competing mechanisms. Experiments across representative scientific
agents reveal a substantial phenomenal–mechanism recovery gap: for Codex with GPT-5.6-sol, phenomenal-law accuracy reaches 35.00% on the Core-set while mechanism accuracy is only 13.75%, with mechanism recovery failing in 64.29% of cases where the phenomenal law is correctly recovered. The gap widens as mechanisms become increasingly mutated, and even providing the correct phenomenal law leaves mechanism recovery below 50%. These results reveal a substantial generalization gap in mechanistic
reasoning and establish mechanism discovery as a distinct challenge beyond recovering observable scientific laws.
Read the full abstract ▾
All authors Zihan Yu, Jiadong Zhang, Jialin Cheng, Jingtao Ding, Yong Li
Machine Learning
One Proposal for Every Margin: Zero-Shot Amortized Sequential Importance Sampling for Binary Matrices
Ruishuo Chen, Weijia Li, Xun Wang et al. (+3 more)
In ecology, psychometrics, and the analysis of social and financial networks, binary
matrices are often analyzed conditional on their observed row and column sums, which restricts the problem to a finite sample space of
matrices with the same margins. Two fundamental problems are to count this space and to sample uniformly from it. Sequential importance sampling (SIS) addresses both with independent weighted samples and an unbiased count
estimator, but its efficiency depends critically on the proposal distribution. Existing proposals are analytically designed, and their accuracy can vary substantially with the margins. We show that the ideal SIS proposal, under which every weight equals the count and the variance vanishes, is exactly the policy of a generative flow network (GFlowNet) with unit reward on every
matrix that has the given margins. We therefore propose MarginFlow, a framework that turns the design of the proposal into a learning problem and amortizes it across margins by exploiting their self-similarity. Every partial
matrix is itself an instance with reduced margins, so one set
transformer that reads the remaining margins serves every margin. We train MarginFlow on a pool of 1904 margins and evaluate it zero-shot on 1190 held-out margins, synthetic and real, from $3\times3$ to $870\times6$. On 1187 of the 1190 margins it matches or beats the best of 31 analytically designed configurations, chosen post hoc for each margin, and its median effective sample fraction is 99.8%. On the 56 margins where that best loses more than one nat of effective sample size, MarginFlow wins every one and raises the median effective sample fraction from 10.3% to 94.1%.
Read the full abstract ▾
All authors Ruishuo Chen, Weijia Li, Xun Wang, Yu Chen, Leheng Cai, Longbo Huang
Mesoscale and Nanoscale Physics
Superfluid Spin Transport in the Van der Waals Antiferromagnet CrCl$_3$
Peisen Yuan, Xiaomin Guo, Vincent Flynn et al. (+3 more)
Over the past decade, the quest for
spin superfluidity has moved to the forefront of spintronics, driven by the promise of phase-gradient-driven, ultra-low-loss
spin transport. However, experimental investigations remain limited, primarily due to the lack of suitable material systems. Here, we report on the discovery and control of a superfluid
spin transport in the easy-plane van der Waals
antiferromagnetic (AFM)
insulator CrCl$_3$ by a nonlocal device structure. Combining nonlocal
magnon transport measurements with theoretical modelling, we demonstrate that
spin superfluidity emerges in CrCl$_3$ under canted AFM
spin configurations, where it gives rise to ultra-long range (around 90 $\mu$m), weakly decaying
spin transport. We also provide direct evidence that strong
magnetic fields and elevated temperatures suppress the superfluid state, restoring the rapid exponential decay with distance of incoherent
magnons. These findings underscore the potential of
spin superfluidity in two-dimensional
magnetic insulators and establish CrCl$_3$ as a promising platform for energy-efficient, long-distance
spin transport in next-generation spintronic applications.
Read the full abstract ▾
All authors Peisen Yuan, Xiaomin Guo, Vincent Flynn, Benedetta Flebus, Fèlix Casanova, Luis E. Hueso
Machine Learning
Improving Generative Model Self-Training with Geometrically Modified Outputs
Patrick Batsell, Thomas Walker and Richard Baraniuk
Self-training generative models - the continued improvement of a model using its own outputs - is becoming increasingly important as high-quality training data becomes scarce. However, naively finetuning on model-generated samples leads to degradation through model collapse and the model autophagy disorder. Negative-guidance self-training methods turn this degradation into a useful signal, using a model finetuned on its own outputs to guide the original model toward improved generation. Existing methods, however, take the negative signal in standard model outputs as given. We instead ask whether this signal can be explicitly strengthened. We introduce Geometrically Modified Outputs (GMOs), which reweight the singular values of the generator's input-output Jacobian to increase the influence of its leading singular directions. This
geometric modification amplifies the mode-seeking behavior and distortions of standard outputs, providing a stronger and more targeted negative signal for self-training. Across a range of one-step generative models, GMOs consistently improve the performance of negative-guidance methods, including Neon and SIMS, compared with using standard model outputs.
Read the full abstract ▾
Quantum Physics
How fast can a parent estimate the value of their children? A quantum algorithm for stochastic games
Marien Raat, Merlin Incerti-Medici, James R. Wootton et al. (+2 more)
We give a
quantum algorithm for
stochastic $2$-player games. Their game trees are the expectiminimax trees of classical game search: $m$
adversarial layers alternate with $m$ chance layers, every node has $deg$ children, and the leaf values lie in an interval of length $N$. The
algorithm estimates the value of the game to root mean square error $\epsilon$ with \[ O\!\left( K^{D}\, deg^{\frac{D}{4}} \, \frac{N}{\epsilon} \, D^{4D} \log\left( \frac{deg N}{\epsilon} \right)^{4D} \right), \qquad D = 2m \] queries to the leaf values, where $K$ is an absolute constant. If we treat the depth $D$ as a constant as is sometimes done in the literature, the bound is $\widetilde O(deg^{m/2}\epsilon^{-1})$ against the classical $\widetilde O(deg^{m}\epsilon^{-2})$. So the quadratic speedup in the branching factor known for
adversarial trees and the quadratic speedup in the accuracy known for a single expectation both survive when we nest one inside the other. We use a derandomised multilevel Monte Carlo
estimator for chance layers and a coherent binary search for
adversarial layers. The composition is done by introducing a conversion between root-mean-square and uniform error guarantees, and a composition to nest it with
quantum mean estimation. We state the
algorithm for
stochastic games, but it is valid for the following conditions: the value of a parent is a Lipschitz function of the values of its children and the value at a chance vertex is linear in its children, however the speedup is preserved only if the deterministic step has a
quantum subroutine with sublinear query complexity.
Read the full abstract ▾
All authors Marien Raat, Merlin Incerti-Medici, James R. Wootton, Evert van Nieuwenburg, Daniel Bultrini
Materials Science
High-Throughput Imaging of Degradation-Inducing Microscopic Impurities in Perovskite Solar Cells
Sofiia Kosar, Anil R. Pininti, Vladyslav Hnapovskyi et al. (+6 more)
The scalable fabrication of high-quality, large-area perovskite thin films is hindered by microscopic inhomogeneities, particularly residual compositional
impurities formed during processing. Identifying these
impurities, understanding their impact on device operation, and enabling their rapid detection are essential for upscaling perovskite
solar cells (PSCs). Here, nano-
Fourier transform infrared spectroscopy combined with high-resolution optical and scanning probe techniques was used to identify detrimental
impurities in wide-bandgap perovskite films relevant to tandem
solar cells. Photo-stress experiments revealed that degradation of the perovskite layer initiates at
impurity-perovskite interfaces, demonstrating that
impurities act as failure nucleation sites. Leveraging these insights, a rapid, non-invasive, and high-throughput framework based on high-resolution reflected light microscopy and machine learning-supported image analysis was developed, enabling detection and quantification of harmful
impurities in as-prepared films within seconds. Films with higher
impurity density show accelerated early degradation, establishing this parameter as an early-warning metric for stability screening. Extending this framework to degradation tracking further reveals coupled photo- and thermo-chemical contributions to
impurity-mediated instability. Overall, this work establishes a practical chemically-validated diagnostic imaging framework for rapid pre-screening of perovskite films to enable stable and scalable PSCs.
Read the full abstract ▾
All authors Sofiia Kosar, Anil R. Pininti, Vladyslav Hnapovskyi, José P. Jurado, Subhashri Mannar, Lorenzo Mardegan, Anand S. Subbiah, Frédéric Laquai, Stefaan De Wolf
Instrumentation and Detectors
Event-by-event track imaging of charged particles with an ultrafast plenoptic camera
Till Dieminger, Marco Barbieri, Vincenzo Berardi et al. (+8 more)
Modern elementary particle detectors often require high-resolution three-dimensional particle tracking in large dense active volumes, which is difficult to achieve in standard devices without an extremely fine
segmentation. In recent work, a plenoptic camera instrumented with a single-
photon avalanche-diode array, called PLATON, demonstrated to image individual visible
photons generated in monolithic scintillating material and reconstruct their three-dimensional origin from a single-event exposure, opening a route to high-spatial-resolution detection of elementary particles in unsegmented scintillator. However, those tests aimed to address
photon sources of essentially zero
geometric extent. In this work we close the gap to the tracking regime. First, we calibrated the PLATON prototype using light tracks in an organic plastic scintillator induced by a femtosecond infrared
laser via two-
photon absorption and characterised its capability to reconstruct
photon-starved 3D lines with a novel
algorithm, achieving good agreement with custom Monte Carlo simulation. Finally, the PLATON prototype was exposed to a test beam of pions at the CERN
Proton Synchrotron. For the first time, a plenoptic detector system was successfully used to detect and reconstruct the 3D tracks of elementary particles from 2D
photon-starved images on an event-by-event basis marking a transformative approach to high-resolution measurements.
Read the full abstract ▾
All authors Till Dieminger, Marco Barbieri, Vincenzo Berardi, Saúl Alonso-Monsalve, Kodai Kaneyasu, Paul Mos, Tim Weber, Johannes Wüthrich, Claudio Bruschini, Edoardo Charbon, Davide Sgalaberna
Operating Systems
Argus: Agentic, Reference-Calibrated, Tree-Guided, System-Software-Level Bottleneck Localization
Vlad-Petru Nitu, Harsh Songara, Konstantinos Sgouras et al. (+3 more)
Operating system (OS) code can account for a substantial share of CPU execution time. First, as application logic is offloaded to heterogeneous accelerators (e.g., GPUs), the CPU increasingly acts as an orchestrator, spending cycles in driver calls, data movement, and synchronization rather than in application code. Second, workloads such as serverless functions frequently invoke OS services. At the same time, the OS is a complex codebase spanning many subsystems (e.g.,
memory management, networking), making it hard to localize the specific code path responsible for a slowdown. Existing profilers expose measurements that require interpretation(e.g., perf and Intel VTune) or can perturb short operations when extensively instrumented (e.g., ftrace). Diagnosing OS bottlenecks can therefore require repeated kernel instrumentation and manual interpretation. We introduce Argus, an agentic LLM-based profiler that produces instrumentation code and autonomously reasons over potential OS-level bottlenecks. Argus integrates two key mechanisms: (i) a calibration methodology that involves collecting a measurement from an idle system and using it as a reference point to discover potential bottlenecks, and (ii) a tree-based data structure that represents the different OS execution paths, improving the
agent's bottleneck localization accuracy. Argus aims to identify a specific kernel code path rather than stop at a subsystem-level diagnosis. In two case studies, we employ Argus to autonomously discover bottlenecks present in the
memory management subsystem caused by (i) a THP aggressor co-running with other applications, and (ii) applications that incur different types of page faults. Argus produces 19 times fewer incorrect deep-path diagnoses than the strongest evaluated LLM-based baseline, which lacks reference calibration, while preserving low time-to-diagnosis (approximately 31 s)
Read the full abstract ▾
All authors Vlad-Petru Nitu, Harsh Songara, Konstantinos Sgouras, Spiros Galanopoulos, Konstantinos Kanellopoulos, Onur Mutlu
Computer Vision
ReVA: A Scene-Centric Dataset Beyond Repetition for Remote Sensing Video Question Answering
Zhen Yao, Likai Wang, Yuming Yang et al. (+9 more)
Multimodal Large
Language Models (MLLMs) have demonstrated remarkable advances in remote sensing. However, existing remote sensing
multimodal reasoning benchmarks exhibit two critical limitations: they rely on (i) template-driven questions, which causes repetitive questions; and (ii) static images that fail to capture the inherent temporal nature of drone/
UAV videos. This leaves systematic evaluation of remote sensing
video reasoning largely unexplored. To address this gap, we introduce ReVA, a new
dataset for remote sensing
video question answering, designed to assess spatiotemporal, scene-centric, and
reasoning-oriented capabilities of MLLMs. ReVA comprises 2,438 drone
videos spanning 18 cities worldwide (580K frames) and 22K high-quality question-answer pairs across 11 challenging QA tasks. We develop a semi-automatic annotation pipeline that leverages Text LLMs and MLLMs for question-answer generation with human verification. We evaluate 23 proprietary and open-source
Video LLMs on ReVA, exposing fundamental limitations of current models. These findings position ReVA as a critical
benchmark toward better remote sensing
video understanding and temporal
reasoning capabilities for real-world deployments. Our code and
dataset are available at: https://github.com/zyaocoder/ReVA
Read the full abstract ▾
All authors Zhen Yao, Likai Wang, Yuming Yang, Zhihao Zheng, Bo Lang, Qiuyu Tang, Jialu Sheng, Jingqi Xu, Yuehai Yang, Jumal Barker, Xiaowen Ying, Mooi Choo Chuah
Cryptography and Security
Detecting False Data Injection and Unstable Operation in Smart Grid via System-Aware Graph Boundary Learning
Emad Efatinasab, Denis Donadel, Mirco Rampazzo et al. (+1 more)
Cyber-physical power systems increasingly rely on data-driven tools to detect instability and support reliable grid operation. However, reliable stability prediction is difficult when unstable operating configurations are rare, sensitive, or unavailable during model development, since collecting such data safely and at scale is often impractical. At the same time, False Data Injection (FDI) attacks can manipulate reported system parameters to trigger false instability alarms or conceal unsafe operation, and such threats in Decentral Smart Grid Control (DSGC) systems remain largely unexplored. These challenges are rarely addressed jointly, leaving a gap between stability prediction and attack detection that this work aims to close. In this paper, we introduce StarGNN, a graph learning framework that learns the stable operating region exclusively from clean stable configurations and uses a single abnormality score to flag reported configurations that should not be trusted as evidence of safe operation, covering both genuine instability and unseen FDI manipulations. Each configuration is represented as a producer-consumer
star graph and processed by a role aware graph neural network, with physics constrained pseudo-negatives generated by perturbing reaction time and price response parameters standing in for the unavailable unstable and attack data. A single threshold, calibrated only on held-out stable data, is used without task or attack specific adjustment. Evaluated on nine unseen FDI scenarios, StarGNN detects between 0.780 and 0.973 of attacks on stable configurations and retains a post attack instability recall between 0.972 and 0.999 on unstable ones, with 0.899 recall against a stronger adaptive attacker, showing that stable-only boundary learning can support both stability prediction and attack detection without access to genuine unstable labels or attack samples during training.
Read the full abstract ▾
All authors Emad Efatinasab, Denis Donadel, Mirco Rampazzo, Chuadhry Mujeeb Ahmed
Machine Learning
An RL View of OPD: Least Square Policy Distillation for Sample-Efficient LLM Reasoning
Shangzhe Li, Yuxiao Yang, Tianrun Yu et al. (+4 more)
We study on-policy
distillation (OPD) through the lens of reinforcement learning, establishing a connection between the reverse-KL objective in OPD and KL-regularized policy optimization. Building on this connection, we introduce Least-Square Policy
Distillation (LSPD), an RL-inspired framework that brings optimistic exploration and off-policy data reuse from value-based RL into policy
distillation. LSPD preserves policy diversity through exploration while improving rollout efficiency by repeatedly learning from previously collected trajectories. Our theoretical analysis connects LSPD to optimistic value-based learning and shows that its idealized formulation achieves a sharp $\tilde{\mathcal O}(\log K)$ regret bound under online exploration. Empirically, LSPD consistently outperforms existing
distillation baselines across six mathematical
reasoning benchmarks and diverse teacher-student settings, with average gains of +1.59 points in Avg@16. Remarkably, through Pass@k evaluations up to k=64, we found that LSPD better preserves policy diversity by achieving stronger performance as k grows. Its fully off-policy variant achieves comparable performance to vanilla OPD using only the first 25% of rollout batches. Together, these results provide an RL perspective on OPD that offers both a principled interpretation and a practical route toward more effective and rollout-efficient
language model
distillation.
Read the full abstract ▾
All authors Shangzhe Li, Yuxiao Yang, Tianrun Yu, Kaixiang Zhao, Xiaoyun Wang, Taylor W. Killian, Weitong Zhang
Computer Vision
SolveEdit: Benchmarking Visual Problem Solving in Generative Models
Wenjie Shu, Yexin Liu, Harold Haodong Chen et al. (+9 more)
Machine intelligence is often evaluated through abstract
reasoning problems, yet many real-world problems are visual, such as arranging objects, repairing layouts, or tracing routes. Solving these problems requires understanding a scene, inferring what must change to achieve a goal, and realizing that change without disturbing unrelated content. However, existing
benchmarks mainly evaluate perception, generation, or explicitly specified transformations, leaving goal-driven visual problem solving underexplored. To bridge this gap, we introduce SolveEpIT, a
benchmark for visual problem solving through scene transformation. Given an image and a goal, a model must infer a valid transformation from the request, the scene, or a visually expressed rule, then execute it while preserving unrelated content. SoLvEEDrr contains 2,728 cases. Atomic transition contracts specify required and protected conditions, enabling SoLvEScoRE to measure completion and unintended changes without a single reference output. The strongest evaluated model achieves only57.0% SolvEScore. We further introduce SolveEdiT-PLAN, a two-stage visual planner that instantiates the transition before generation. Under matched single-generation evaluation, it improves SoLvEScoRE by 9.1 points on average across three tested generators, including a gain from 57.0% to 71.6% for GPT-Image-2, without modifying the editor.
Read the full abstract ▾
All authors Wenjie Shu, Yexin Liu, Harold Haodong Chen, Xuerui Qiu, Zehan Wang, Yidi Zhang, Yizhan Chen, Zunwei Wang, Minghao Liu, Qi Chen, Harry Yang, Xiaogang Xu
High Energy Physics (Phenomenology)
Magnetic thresholds and spin-resolved spectral moments of $\phi$ and $K^{*0}$ in a hadronic model
Lang Yu, Tao-Tao Sui and Xinyang Wang
We study the
spin-resolved spectral functions of $\phi$ and $K^{*0}$ in a thermal hadronic model with a constant
magnetic field. We construct the rest-frame retarded propagators by combining real
vacuum matching with exact Landau levels, connection contacts, and thermal scattering terms. The resulting stable-daughter calculation is checked against independent collision kernels and published expressions in common limits. We find that closed transverse thresholds give rise to narrow peaks with eV-scale detunings, whose areas must be resolved in the calculation of small
spin spectral moments. We test the leading threshold estimate for the peak area by an independent integration of the full kernel. Using positive compact daughter spectra, we then investigate how a finite spectral spread changes the local threshold structure. The disappearance of zeros of the real part of the inverse propagator is controlled by the spread relative to the detuning, whereas the weight integrated over a finite patch can remain nearly unchanged. We relate this redistribution to smooth accepted moments through a weighted identity with a bounded Taylor remainder, and evaluate the complete-window response to three endpoint replacements, including the patch complement. For the stable-daughter $\phi$ spectra with a fixed smooth weight, the response at the two smallest sampled fields is negative and compatible with $b^2$ scaling. Hard-window results instead oscillate in sign, and pseudoscalar mixing substantially changes the smooth response coefficient. These results quantify the sensitivity to thresholds and spectral projection; microscopic daughter widths and source-matched experimental
spin alignment require additional input.
Read the full abstract ▾
Machine Learning
Structured Latent Modeling for Supervised Multimodal Information Decomposition
Wanting Huang, Sanvesh Srivastava and Weiran Wang
Multimodal prediction relies on diverse forms of evidence: information repeated across modalities, cues specific to a single source, and complex cross-modal dependencies that emerge only when inputs are considered together. While recent methods promote richer interactions, they lack a principled way to isolate these target-relative contributions within learned continuous representations. We introduce a framework that applies contrastive or masked objectives at intermediate layers, coupled with source-wise invertible normalizing flows and a supervised, low-rank latent variable model. This architecture explicitly factorizes the joint distribution into shared task-relevant variation, modality-specific predictive variation, and task-irrelevant dependence. Drawing connections to prior
multimodal learning assumptions, our approach evaluates how modalities independently and jointly contribute to the target. Ultimately, this framework unites intermediate representation learning with structured likelihood-based guidance, offering a practical latent-variable lens for characterizing continuous
multimodal interactions. Empirically, we demonstrate the effectiveness of our approach across diverse
multimodal benchmarks, showing robust improvements in predictive performance.
Read the full abstract ▾
Artificial Intelligence
SRHarness: A Harness for Agentic Symbolic Regression
Zihan Yu, Shixuan Zhou, Hao Huang et al. (+2 more)
Recent agentic symbolic regression approaches increasingly rely on large
language models to analyze data, select scientific operations, and refine hypotheses over long search trajectories. In such systems, performance depends not only on the underlying model and search strategy, but also on the runtime infrastructure that supports scientific search. We introduce SRHarness, a domain-specific harness for agentic symbolic regression built around three mechanisms: composable scientific actions that provide a common interface over raw, transformed, and candidate-derived quantities; persistent scientific state that retains evaluated hypotheses and exposes compact model-facing views; and trajectory lifecycle management that coordinates continuation, branching, restart, and termination. On LLM-SRBench, SRHarness consistently improves both numerical generalization and symbolic recovery under matched LLM backbones. With DeepSeek-v4-flash-0731, it achieves 93.69% symbolic accuracy on LSR-Transform, compared with 62.16% for SR-Scientist, and retains 72.97% accuracy on an anonymized variant that removes scientific descriptions and variable semantics, versus 39.64% for SR-Scientist. Under the same DeepSeek-v4-flash-0731 backbone, SRHarness also substantially outperforms Codex (72.97% vs. 20.72%) and reaches performance comparable to Codex with GPT-5.5, while simply providing Codex with the same scientific tools does not reproduce this advantage. These results show that effective agentic symbolic regression depends not only on models or tools, but also on structured runtime support for organizing scientific actions, accumulated hypotheses, and long-horizon search.
Read the full abstract ▾
All authors Zihan Yu, Shixuan Zhou, Hao Huang, Jingtao Ding, Yong Li
General Relativity and Quantum Cosmology
Kasner critical scaling in Einstein–Maxwell–scalar black holes with nonlinear collapse and scalarization
Chihang He and Chao Liu
A critical scaling exponent of $1/2$ was numerically observed for the Kasner parameter of asymptotically flat scalarized black holes and conjectured to be universal [Phys.\ Rev.\ Lett.\ 136, 251402 (2026)]. In the same static, spherically symmetric Einstein–Maxwell–scalar setting, we show that this exponent is not universal. We develop a rigorous mathematical framework and construct couplings realizing exponents $1/(2m)$ for any integer $m\ge1$ by tuning higher-order terms at fixed quadratic coupling parameter $\mu$. Nonlinear interior collapse produces a spacelike Kasner singularity with $\beta_{\rm K}\propto\varepsilon^{-1}$, where $\varepsilon$ is the value of scalar field at the event horizon. An exterior branch satisfying $q/q_\star-1\propto\varepsilon^{2m}$, therefore, yields $\beta_{\rm K}\propto|q/q_\star-1|^{-1/(2m)}$, where $q$ is the charge-to-mass ratio and $q_\star$ its critical value. We also determine the prefactor explicitly. This recovers the $1/2$ law and establishes a hierarchy including $1/4$ and $1/6$.
Read the full abstract ▾
Machine Learning
Physics-Guided Conditional Diffusion Model for Rare Event Synthesis and Diagnosis for the Water-Gas Shift Reaction
Md Abrar Rafid Siddique, Bibek Aryal and Qiugang Lu
As the world moves towards sustainable energy sources, hydrogen (H2) can be treated as an eco-friendly alternative to fossil fuels due to its high energy density and zero carbon emissions. The water-gas shift (WGS) reaction is a widely used industrial process for hydrogen production by converting carbon monoxide and steam into hydrogen and carbon dioxide. However, occurrences like severe fouling, catalyst deterioration, and thermal runaway can hamper the reaction kinetics/process safety and decrease the yield of H2. These incidents are rare, and gathering process data under such abnormal conditions is challenging. In this work, we propose a physics-guided conditional
diffusion model to generate realistic rare-event trajectories for the WGS reaction. The proposed model integrates a conditional denoising
diffusion probabilistic model (CDDPM) with governing laws of the reaction to generate physically consistent process trajectories. The conditioning features allow the model to produce high-quality synthetic profiles for rare-event domains that are typically beyond the training regimes. The generated rare-event trajectories then augment the raw
dataset for a balanced distribution between normal and abnormal conditions. We further propose a hazard score to assess the risk severity of the operating condition based on the operating trajectory. Deep learning models are trained with the augmented
dataset to diagnose the health status of the reaction. Simulation results show that the proposed physics-guided
diffusion model outperforms data-driven models in terms of the quality of synthetic data and diagnosis performance for rare events.
Read the full abstract ▾
Quantum Physics
Electron decoherence in cylindrical holes and circular apertures
Cruz I. Velasco and F. Javier García de Abajo
The coherence of free electrons sets fundamental limits on the resolution and contrast of phase-sensitive electron microscopy because coherence is lost whenever the electron leaves a distinguishing excitation in its environment. We develop a quantitative theory of fast-electron decoherence in two canonical geometries: a cylindrical hole drilled through a realistic metal and a circular aperture in a thin perfectly conducting film. By combining an electromagnetic Green-tensor formulation with the fluctuation–dissipation
theorem while fully retaining retardation, we obtain the decoherence probability and elastic phase as functions of the electron trajectory, temperature, and material response. For the cylindrical hole, we obtain a closed-form, azimuthally resolved expression that separates the dependences on path position, temperature, and conductivity. In the high-conductivity and high-temperature limits, this result reduces to a universal expression that is linear in temperature and independent of both conductivity and electron velocity. For the circular aperture, which we solve using a cylindrical-wave modal expansion, the energy-loss probability diverges as $1/\omega$ at low frequency, whereas the decoherence probability remains finite and reaches a maximum when the aperture radius is comparable to the relevant electromagnetic wavelength. Because the interaction is spatially localized, aperture-induced decoherence is generally weaker than that produced by a cylindrical hole, except for sufficiently large apertures at high temperature. Finally, we show that, for sufficiently large holes, decoherence can dominate the spatial broadening of a focused electron probe and must therefore be incorporated into the design of coherent electron-beam instruments.
Read the full abstract ▾
Computer Vision
Sprout: Building Dynamic Memory While Reasoning for Agentic Video Understanding
Wei Chen, Xuanyu Zheng, Yancheng Long et al. (+6 more)
Long
video understanding relies on
video memory to overcome the context limits of
multimodal large
language models. Existing methods follow a build-then-
reasoning pipeline:
memory is built offline for the entire
video, then reasoned over as a static source. In practice a long
video is shared by several questions, and this pipeline is costly at both ends: with few questions, building
memory for the whole
video costs far more than answering them; with many questions, the
memory is never updated, so what is learned while answering questions is lost to the next question. To alleviate these, we introduce Sprout, an agentic framework that builds
memory while
reasoning: a temporal tree that sprouts detailed nodes as questions are answered. The
agent watches the
video segment by segment at a low frame rate, stopping when the current question can be answered, remembers each segment as a coarse node of the tree, and revisits key intervals at a higher frame rate to refine the tree with the recovered details. Once a segment is recorded as text, its
video input is removed from the context history, while the original
video remains reachable through the
video tools. The
memory tree and prior question–answer records persist across questions, so the
memory is online and dynamic: built from the first question onward and updated by every question thereafter. We find that replacing accumulated
video inputs with textual
memory substantially reduces context usage while maintaining accuracy, with slight improvements in some settings. Across
benchmarks on three models, Sprout achieves competitive or improved accuracy relative to representative offline
memory methods, with no upfront construction stage and lower context cost per question.
Read the full abstract ▾
All authors Wei Chen, Xuanyu Zheng, Yancheng Long, Haoyang Xu, Kaiyu Jiang, Bin Wen, Tingting Gao, Han Li, Long Chen