Peer Revenue Past citations are no guarantee of future returns

Papers

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

All of arXiv

Monday 14 September

Hardware Architecture

LLM-enabled Behavior Driven Development Workflow for Formally Verified Hardware Designs

Recently, the use of Large Language Models (LLMs) for different tasks in the Electronic Design Automation (EDA) life-cycle has been studied extensively, but an integrated view is lacking. Specifications are the foundation of this life-cycle, but they suffer from ambiguity when written in natural language, which especially affects the quality of LLM output. Formal specifications mitigate these ambiguities, but they come with their own challenges. On the other hand, Controlled Natural Language (CNL) specifications can serve as a middle-ground, reducing ambiguity while retaining interpretability. In this work, we propose an integrated view on the use of LLMs for EDA and establish an LLM-enabled behavior driven hardware development workflow. We introduce and define Formal Verification Gherkin Scenarios (FV Gherkin Scenarios), unlocking CNL specifications as the foundation for formally verified hardware designs via Formal Property Verification (FPV). Experimental evaluation shows that our workflow is able to outperform other established LLM-based methods by 2.48x in functional correctness of generated Register Transfer Level (RTL) designs and by 2.54x in formal coverage of generated assertions for FPV.
Read the full abstract

Analysis of PDEs

Strong Solutions and Long-Time Behavior for One-Dimensional Compressible Navier–Stokes Equations with Thermodynamically Compatible Pressure Laws

Motivated in part by non-ideal equations of state such as the van der Waals law, we study the Cauchy problem for the one-dimensional full compressible Navier–Stokes equations with constant positive viscosity and heat conductivity and general thermodynamically compatible pressure and internal-energy functions. Assuming $e_\theta>0$, we obtain local $H^1(\mathbb R)$ strong solutions for initial data with positive specific volume and temperature, without sign conditions on the pressure derivatives. Near an equilibrium satisfying $p_v(\bar v,\bar\theta)<0$, sufficiently small $H^1(\mathbb R)$ perturbations generate global strong solutions with uniform thermodynamic bounds and time-integrable dissipation. For the full constitutive class, all first spatial derivatives decay in $L^2(\mathbb R)$ and the perturbation converges to zero in $L^p(\mathbb R)$ for every $2<p\le\infty$. We also identify a locally entropy-affine subclass of constitutive laws for which a Fourier–Duhamel argument establishes convergence to equilibrium in $H^1(\mathbb R)$. For more general thermodynamically compatible constitutive laws, we establish a conditional criterion for convergence in $H^1(\mathbb R)$, based on the space-time integrability of the nonlinear pressure remainder in entropy variables.
Read the full abstract

2609.15317 · 14 Sep · This version substantially extends the previous manuscript, which established local well-posedness and global existence. It adds long-time stability results, including decay of spatial gradients, conditional nonlinear H^1 convergence, and full H^1 stabilization for locally affine pressure laws in entropy coordinates. The title and author list have also been updated

High Energy Physics (Theory)

On the commuting charges and the Bethe ansatz for the quantised Kadomtsev-Petviashvili equation

We study the quantised version of the Kadomtsev-Petviashvili equation in the formulation proposed in [4]. We analyse two aspects. One is the conservation of the first higher charge $H_3$ (besides the total mass $H_0$, the momentum $H_1$ and the Hamiltonian $H_2$). We find the expression of $H_3$ and prove commutativity, provided we regularise a specific ultraviolet divergence. The result then does not depend on the choice of a regulator, under fairly general assumptions. We then study the Bethe ansatz in finite volume and obtain a general solution to the two-body problem. We have used AI supervised by us to aid the proofs.
Read the full abstract

Artificial Intelligence

Evaluation Metrics for Safe Reinforcement Learning

Safe reinforcement learning (RL) is commonly formalized as a Constrained Markov Decision Process (CMDP), in which an agent maximizes expected reward while keeping its expected cumulative cost below a specified safety bound. Existing safe RL benchmarks predominantly report whether an algorithm is safe on average, following this expectation-based guarantee. We argue that this convention is insufficient to reliably characterize an algorithm's true safety: it fails to capture how often and how severely the safety bound is violated, whether this holds consistently across tasks and safety bounds, and whether training-time behavior is representative of behavior of the final converged policy. Therefore, we introduce (i) evaluation metrics for safe RL that address each of these concerns and in addition allow for aggregation across tasks and safety bounds. We furthermore define (ii) a safety tier system to systematically categorize and compare algorithms in terms of safety and reliability at both training and for a final policy. Using this framework, we provide (iii) an empirical safety evaluation across multiple safety navigation tasks. Our results show that aggregate metrics, distributional reporting, and task- and safety bound-specific results each reveal information the other metrics cannot. We therefore recommend reporting all three jointly, rather than compressing this information into a single value, as is common practice. We provide SafeRLEval, an open-source evaluation suite to support the reliable characterization of safety in future safe RL research.
Read the full abstract

Machine Learning

The Universe of Universes: Benefit Yield Functions, Implosion Thresholds, and Infrastructure-Aware Optimization in Multi-LLM Systems

We introduce the Universe of Universes (UoU) framework, which treats the full ecosystem of major large language models (LLMs) as a structured retrieval corpus and proposes a compositional Automated Reasoning (AR) and Machine Learning (ML) architecture for cross-model retrieval-augmented generation. The central contribution is the formal characterization of the Benefit Yield Function (BYF), the marginal performance gain per additional model added to an ensemble, and the identification of the implosion threshold {\theta}*: the ensemble size at which BYF crosses zero and aggregate performance begins to degrade. Existing LLM ensemble and mixture-of-agents systems treat models as responders and aggregate outputs, but do not study performance as a function of ensemble size N across the full model universe. Benchmark research confirms performance plateaus at the individual model level; model collapse literature establishes that iterative training on AI-generated outputs degrades individual model distributions. Neither body of work formalizes the ensemble-level implosion threshold, models Epistemic Hereditary Drift (EHD) at the ecosystem level, or treats AI manufacturing velocity as a co-variable of {\theta}*. The framework has direct implications for DoD multi-model AI acquisition policy and the emerging science of testing AI-enabled systems.
Read the full abstract

Audio and Speech Processing

Reducing the Output-Mode Gap in Speech Language Models via Joint-Output On-Policy Distillation

Autoregressive generation of interleaved text and acoustic tokens is a common approach to spoken-response generation in speech large language models. Although this design enables streaming generation with explicit textual guidance, generated acoustic tokens become part of the context for subsequent text predictions. Given identical speech inputs, we observe markedly lower answer accuracy for the internal text generated in speech-to-text-and-speech (S2TS) mode than for speech-to-text (S2T) responses. We term this discrepancy the output-mode gap (OMG). To reduce OMG, we propose Joint-Output On-Policy Distillation (JO-OPD), which distills the model's stronger S2T policy into joint generation using student-generated S2TS trajectories. At each text position, the S2T teacher provides soft targets from a text-only projection of the student's preceding outputs, while the student predicts from the corresponding full interleaved history. A preservation objective further regularizes native non-text predictions. Experiments on Step-Audio-2-mini and Baichuan-Audio-Instruct reveal OMG across two interleaved generation architectures. On Step-Audio-2-mini, JO-OPD reduces OMG from 42.87 to 16.26 percentage points on Spoken-MQA and from 29.72 to 13.04 points on speech-rendered GSM8K, with little change in S2T accuracy and substantially larger reductions than matched SFT baselines. ASR-based evaluation further shows a 7.49-point improvement in spoken-answer accuracy on Spoken-MQA.
Read the full abstract

All authors Daxin Tan, Dehua Tao, Chengxi Deng, Hanlin Zhang, Xiao Chen

Cryptography and Security

Large Universe Subset Predicate Encryption with IND-CCA Security (with Constant-size Ciphertext and Keys)

Katz et al. (CANS'17) introduced Subset Predicate Encryption (SPE). This scheme is a generalization of broadcast encryption as it emulates the subset containment predicate in the encrypted domain. They proposed two selectively IND-CPA secure SPE constructions in the small universe setting. They also showed some black-box transformations of SPE to well-known primitives like WIBE and ABE to establish the richness of the SPE structure. Chatterjee and Mukherjee (RSA'19) proposed two SPE constructions in the large-universe setting. Their first construction achieved constant-size ciphertexts and secret keys, but it is proven secure in a restricted version of selective security. Although the second construction achieves adaptive security, the ciphertext size depends on the size of the data-attribute set. Furthermore, neither of these two constructions achieves CCA security. In this work, we propose the first large-universe CCA-secure subset predicate encryption with constant-size ciphertext and secret keys. We prove this construction achieves standard selective security under the standard subgroup decision problems. Finally, we transform our extremely efficient SPE into the first CCA-secure WIBE, WKD-IBE, etc., with constant-size ciphertexts and secret keys via black-box transformations.
Read the full abstract

Hardware Architecture

FlashGPU-sim: Enabling GPU Modeling for Modern Architectures and AI Workloads

As AI becomes increasingly ubiquitous, modern AI systems are shaped by a tight software-hardware co-design loop. Later GPUs expose features such as asynchronous data movement, tensor core pipelines, and fine-grained synchronization that high-performance kernels aggressively exploit, while emerging application behaviors increasingly influence the next generation of hardware design. Unfortunately, the latest open-source simulators for NVIDIA GPUs focus on architectures and software stacks from roughly six years ago. Therefore, they cannot support many state-of-the-art AI kernels generated by modern compiler stacks, e.g. Triton, or accurately model the hardware features they depend on. As a result, architects lack a credible platform for analyzing bottlenecks in this flywheel or evaluating design trade-offs for future AI systems. To bridge this gap, we present FlashGPU-sim, an open-source, execution-driven, cycle-accurate GPU simulator for modern AI workloads. FlashGPU-sim faithfully models modern hardware features such as asynchronous data movement, fine-grained synchronization, tensor-core execution, and distributed shared memory. A Triton extraction front-end allows direct simulation of optimized AI operators without manual porting, while multi-threaded execution makes large-scale software-hardware co-design practical. Across 131 workload configurations on RTX 5090, H100, and B200, FlashGPU-sim achieves a cycle-level MAPE of 5.24%, while multi-threaded simulation reaches a 7.86x speedup with 16 host threads. An H100 case study further demonstrates its utility for microarchitectural design exploration.
Read the full abstract

All authors Siying Yu, Yixun Hong, Guozhi Qiu, Jingci Liu, Feng Gu, Chenbo Geng, Zhengrong Wang, Chen Zhang, Bei Yu

Signal Processing

RT-VSS: In-House Ray Tracing for Vital Signs Sensing in FR3 Band

Vital Signs Sensing, enabled by the evolution of wireless communications, is inherently challenging as the chest's micro-Doppler signatures can be corrupted by multipath, body movement and noise. Additionally, the prospects of the Frequency Range (FR) 3 band, spanning 7-24 GHz, are compelling due to centimeter-scale wavelength improving phase sensitivity to chest motion, while its abundant bandwidth and rich multipath enhance temporal and spatial diversity. In this paper, we validate the feasibility of using cellular communications in the FR3 band to monitor a human's breathing and heartbeat in an indoor environment. Using Blender to model physiologically accurate vital-signs signals as ground truth and Sionna Ray Tracing simulations, we find that it is feasible to extract breathing and heartbeat waveforms at FR3 frequencies. Furthermore, we perform a comparative analysis of ISAC performance against representative sub-6 GHz (FR1) and mmWave (FR2) frequencies, indicating the utility of the study across the frequency bands.
Read the full abstract

All authors Rudranil Chattopadhyay, Zhuangzhuang Cui, Achiel Colpaert, Sofie Pollin

Computation and Language

When Agents Slow Down: Understanding LLM Agents' Test-Time Strategies via Elo-per-token Analysis

Large language model (LLM) agents allocate test-time compute adaptively as they revise solutions, use tools, explore alternatives, and decide when to stop. This test-time strategy makes it difficult to measure how agent performance scales. We study open-ended tasks that provide continuous scores for intermediate submissions, making progress observable throughout long trajectories. We propose Elo-per-token analysis, which tracks the best solution found at each token budget and uses a Bradley-Terry model to aggregate within-task orderings into Elo ratings across tasks with different score scales. We apply it to four general-purpose agents on four open-ended benchmarks, with sessions of up to 100M tokens, and to three feedback-driven LLM optimization harnesses in controlled single-task interventions. Independent sampling provides a theoretically characterized reference, for which Elo grows linearly with log compute. Against this reference, agents can initially convert tokens into Elo faster than independent sampling, but their marginal gains diminish and eventually fall below the reference. In contrast, the strongest historical human contestants improve superlinearly over contest time on shared AtCoder Heuristic Contest tasks, providing evidence of continual learning and substantial headroom after agents slow down. We define the scaling inflection point as the per-session budget where marginal Elo gains match the independent-sampling reference. Using this point as the per-session budget, we split 100M tokens across parallel sessions on FrontierCS Polyomino Packing, gaining +264 Elo over one long session and +355 over ten short sessions.
Read the full abstract

All authors Kaiyuan Liu, Qiuyang Mang, Bo Peng, Wenhao Chai, Hanchen Li, Shreyas Pimpalgaonkar, Luke Zettlemoyer, Alex Dimakis, Alvin Cheung

General Relativity and Quantum Cosmology

Deterministic denoising of long-duration gravitational-wave signal candidates with $\alpha$-(de)blending

We present a deterministic machine-learning method for denoising candidate signals from all-sky searches for continuous gravitational waves from rotating neutron stars. Using the time-domain $F$-statistic search pipeline, we post-process the resulting $F(f,\dot{f})$ candidate patterns with iterative $\alpha$-(de)blending, a deterministic diffusion-type generative model implemented as a U-Net, well suited to the non-Gaussian, correlated noise of the $F$-statistic output. Two models, trained on data based on simulated 6- and 12-day time-domain segments with software-injected signals similar to the LVK hardware injections, are tested by comparing deblended images to the library of expected signal patterns via the Structural Similarity Index Measure. In general the method recovers the correct sky-position-dependent pattern for signal-to-noise ratios $\rho \gtrsim 4$, moderately below typical all-sky detection thresholds, with recovery depending strongly on pattern morphology, demonstrating that deterministic diffusion-based denoising may serve as a consistency/veto tool ahead of the semi-coherent coincidence stage. We also discuss limitations and possible improvements of the method.
Read the full abstract

Computer Vision

Learning from Reliable Negatives: Confidence-Anchored Test-Time Adaptation for GUI Grounding

Graphical User Interface (GUI) grounding is essential for autonomous agents to map natural language instructions to precise screen coordinates. However, existing supervised fine-tuning and reinforcement learning methods are constrained by the high cost of annotation, creating a scalability bottleneck. In this paper, we introduce a label-free test-time training paradigm driven by two key insights: (1) confidence patterns in coordinate tokens are a better indicator than full-sequence confidence, and (2) in sparse GUI coordinate spaces, negative samples offer more reliable learning signals than potentially noisy positive ones. We first propose Confidence-Anchored Learning (CAL), which utilizes coordinate-token confidence to filter pseudo-labels and assign distance-based binary rewards. Building on this, we develop Confidence-Anchored Negative Learning (CANL), which exclusively optimizes the model using negative samples to bypass the risks of incorrect positive samples. Experimental results demonstrate that CANL-7B achieves 92.1% on ScreenSpot-V2. On more challenging ScreenSpot-Pro, CANL-7B reaches 33.8%, an 8.9% absolute improvement over the base model. Our findings establish coordinate-token confidence as a powerful alternative to manual annotations for scalable GUI agent development.
Read the full abstract

All authors Yizhou Liu, Fei Tang, Yuchen Yan, Zhengxi Lu, Songqin Nong, Tao Jiang, Wenhao Xu, Wenqi Zhang, Weiming Lu, Jun Xiao, Yongliang Shen

Mathematical Finance

The skew Brownian motion should not be used as a risk-neutral returns process: a well-posed skew-normal alternative

Return models for risk-neutral financial valuation based on skew Brownian motions (SBMs) have been introduced about twenty years ago, and have recently enjoying growing popularity. Unfortunately, the story behind their development is one of mistakes and erroneous interpretations, beginning from the foundational misrepresentations that the prevalent financial model is based on the Itô-McKean SBM – which, in fact, it is not. Besides, and more seriously, the clarification of \cite{rossello2012arbitrage} that price models with a local time in their returns, such as the SBM, are arbitrageable has been, by and large, ignored. In this paper, we try to clear the field from the confusions and misconceptions lurking in the standing option pricing literature on SBM, by exposing all the errors we could trace in the treatment so far. Recognizing however the potential of the SBM skew-normal marginals for risk-neutral valuation, as a positive contribution, we reformulate the putative SBM call pricing formula and show that, even if the SBM return model admits arbitrage, its option pricing formula does not. The correct Markovian SDE with skew-normal marginals is then identified, its strong well-posedness shown, and by exploiting the availability of closed formulae, an asymptotic analysis of the implied volatility surface is offered. En route to our conclusions, we obtain a novel normal/skew-normal stochastic dominance property of independent interest.
Read the full abstract

Machine Learning

When Correlations Mislead: Confounder-Aware Multi-View Urban Region Representation Learning

Urban region representation learning commonly combines heterogeneous data sources, such as mobility flows, points of interest, and land-use information, to support tasks including mobility analysis, public safety forecasting, and service demand estimation. Existing multi-view methods typically improve region embeddings by strengthening interactions across views. However, such methods often overlook view-specific regional structures and may propagate correlations induced by shared latent factors, which can reduce the stability of downstream predictions. To overcome this major limitation, we propose CURE, a confounder-aware framework for multi-view urban region representation learning. CURE first encodes each view with its regional graph structure, estimates a shared latent component, and then reduces its projected influence before cross-view interaction. A hierarchical graph-aware fusion module subsequently aggregates the residual view representations using local and global regional contexts Experiments on three real-world cities show that CURE improves predictive performance, remains robust under missing and noisy input views, and provides reliable cross-view integration through shared component separation and context-dependent view weighting.
Read the full abstract

All authors Sean Bin Yang, Ying Sun, Zongyi Xu, Tung Kieu, Jilin Hu, Bin Yang, Kristian Torp, Hua Lu, Torben Bach Pedersen

Optics

High-efficiency integrated laser on erbium-doped lithium niobate-on-insulator

Lithium niobate on insulator (LNOI) combines the outstanding optical properties of lithium niobate (LN) with strong optical confinement, scalable fabrication and high-density integration, making it a leading platform for integrated photonic chips. Recent advances in LNOI photonics have mainly centred on passive and electro-optic components, including couplers, waveguides, microcavities and modulators, whereas efficient on-chip laser sources remain insufficiently developed, limiting the realization of fully integrated LN photonic systems. Because LN is an indirect-bandgap material, lasing on LNOI generally relies on photoluminescence from rare-earth-ion doping, yet the conversion efficiency of doped LNOI lasers has remained low. By comparing LNOI microcavity lasers with fibre lasers and waveguide amplifiers, we identify the limited number of rare-earth ions participating in stimulated emission as a key factor responsible for inefficient pump utilization. Here we demonstrate an integrated Er-doped LNOI laser that combines high-quality, highly Er-doped LN, a large-diameter wide-microring resonator, a low-loss waveguide amplifier and bidirectional pumping. This architecture enables a slope efficiency of 16.91% at 1562 nm, exceeding 10% on the LNOI platform for the first time. Our results provide a route towards high-efficiency LNOI lasers for fully integrated photonic systems.
Read the full abstract

All authors Chunyu Zhang, Yuqi Zhang, Yiyang Zou, Xiaomin Wang, Binwen Niu, Cangsu Yuan, Tongxin Xue, Chunlin Zhu, Hongde Liu, Dahuai Zheng, Shiguo Liu, Fang Bo, Yongfa Kong, Jingjun Xu

General Relativity and Quantum Cosmology

Testing the equivalence principle with lunar laser ranging residuals from different planetary and lunar ephemerides

Lunar laser ranging (LLR) provides a sensitive test of the equivalence principle (EP) through a possible synodic perturbation in the Earth–Moon distance. In this work, we develop an independent LLR data-reduction framework and investigate this signature in the post-fit O–C residuals obtained with the EPM21, INPOP21, and DE430 planetary and lunar ephemerides. The same observation model and parameter-adjustment procedure are applied to the three ephemerides, allowing their residual-based results to be compared within a consistent framework. To account for the nonuniform distribution of LLR observations over the synodic angle $D$, the residuals are fitted with cosine and sine harmonics up to the third order. Signal-injection tests are also performed to quantify the attenuation of a possible $\cos D$ signal during the parameter adjustment. After correcting for signal attenuation, solar radiation pressure, and the thermal response of the lunar retroreflectors, the preferred EPM21, INPOP21, and DE430 solutions yield $\Delta(m_g/m_i)_{\rm EM}=(0.900\pm3.183)\times10^{-14}$, $(0.496\pm3.624)\times10^{-14}$, and $(2.669\pm7.048)\times10^{-14}$, respectively. The EPM21 and INPOP21 results are based on observations from 1970 to 2024, whereas the DE430 solution was obtained using observations up to 2016. All three estimates are consistent with zero within their uncertainties. This work demonstrates that residual-based analysis can provide an independent and transparent EP test, while also offering guidance for assessing sampling and ephemeris-dependent effects in future high-precision LLR studies.
Read the full abstract

All authors Jun Ke, Jie Luo, Qin Li, Yu-Jie Tan, Cheng-Gang Shao

Quantum Physics

Universality from Quadratic Angular-Momentum Operations: Exact Dynamical Lie Algebras and the Role of the Casimir Obstruction

Universality relies on combining operators with different algebraic properties, which depend on the physical structure of the Hilbert space encoding the information. Systems of identical symmetric particles—including bosons and symmetric collections of spin-$1/2$ particles—can be mapped to angular-momentum systems, where rotations supplemented by quadratic, spin-squeezing interactions provide universal control. We revisit universality in this framework and confirm that such linear and quadratic operations provide universal control on every fixed-particle-number subspace. At the abstract level, however, we show that the dynamical Lie algebra generated by all degree-at-most-two elements does not exhaust the full universal enveloping algebra of $\mathfrak{su}_{\mathbb{R}}(2)$, contrary to a stronger claim made previously: higher powers of the quadratic Casimir element cannot be generated by commutators and linear combinations. This obstruction does not conflict with fixed-sector universality because, under a fixed irreducible spin-$n/2$ representation, all powers of the Casimir element act as scalar multiples of the identity. We determine the resulting abstract dynamical Lie algebras exactly and show that the central-free and full degree-at-most-two algebras map, respectively, onto $\mathfrak{su}_{\mathbb{R}}(n+1)$ and $\mathfrak{u}_{\mathbb{R}}(n+1)$, establishing special-unitary and full-unitary universality. We further derive a practical degree-two universality criterion and discuss finite-dimensional SSRC universality.
Read the full abstract

High Energy Astrophysical Phenomena

Spectro-polarimetric study of the black hole X-ray binary Swift J1727.8-1613 in the JED-SAD framework

Context. Swift J1727.8-1613 is an X-ray binary observed during its 2023-2024 outburst, including bright and dim hard states with similar X-ray polarimetric signatures but fluxes differing by two orders of magnitude. Aims. We aim to reproduce the spectro-polarimetric properties of these two hard states. Methods. We modelled their spectra in the JED-SAD framework and computed the X-ray polarisation properties of the best-fit configurations with the MONK general relativistic radiative transfer code. Results. Both spectra are well reproduced within the JED-SAD framework, although the dominant radiative process differs: self-Comptonised bremsstrahlung in the bright state and self-Comptonised synchrotron in the dim state. In the bright state, the simulated polarisation fraction remains below the observed value of about 4%, while the polarisation angle differs from the observations by 90 degrees because of the JED's large vertical optical depth. In contrast, the dim state's simulated polarisation fraction is about 3% and its polarisation angle is perpendicular to the disk, both broadly consistent with IXPE measurements. The dim-state results are largely insensitive to the magnetic field configuration because of Faraday depolarisation. Conclusions. The dim hard state's spectro-polarimetric properties are well reproduced, whereas those of the bright state are not. We discuss a non-thermal particle distribution and electron scattering in an accretion-disk wind as possible explanations for the observed bright-state polarisation.
Read the full abstract

All authors R. Muhammad, P.-O. Petrucci, W. Zhang, J. Ferreira, R. Belmont, J. Malzac, G. Marcel, G. Matt, M. Parra, T. Bouchet, N. Barnier, M. Clavel, G. Henri

Strongly Correlated Electrons

Engineering Competing Fractional Topological States in a BHZ Superlattice

Scalar superlattices offer a route from simple topological bands to fractionalized states, while multicomponent systems raise the additional question of which fractional order is energetically selected. We show that a scalar superlattice applied to the Qi-Wu-Zhang (QWZ) Chern-insulator model produces an isolated C = +1 miniband with strongly improved Berry-curvature and quantum metric uniformity. Exact diagonalization of the projected interaction identifies a 1/3 fractional Chern insulator in this miniband. Restoring the time-reversed partner to form the Bernevig-Hughes-Zhang (BHZ) superlattice yields, at total filling 2/3, a balanced fractional quantum spin Hall (FQSH)-like state and a fully polarized 2/3 fractional Chern insulator. Their competition separates two aspects of fractional-state control: the finite-momentum intercomponent coupling governs the stability of the balanced FQSH-like state, whereas a uniform pseudospin anisotropy shifts the relative energies of conserved sectors and can switch the global ground state between the balanced and polarized states. These results establish a minimal lattice setting that connects topological-miniband reconstruction, fractional-state formation, and fractional-state selection, illustrating how distinct levels of control can be combined to navigate competing fractional topological states.
Read the full abstract

Materials Science

Benchmarking Machine-Learning Interatomic Potentials for Dynamical Stability in Inorganic Semiconductor Nanocrystals: A CdSe Case Study

Machine-learning interatomic potentials (MLIPs) enable nanosecond-scale atomistic simulations of inorganic semiconductor nanocrystals, but low errors on held-out configurations do not necessarily guarantee stable molecular dynamics. We benchmark five graph-neural-network MLIPs, SchNet, PaiNN, NequIP, Allegro and MACE, for dynamical stability in a chloride-passivated cadmium selenide nanocluster containing 149 atoms. The models were trained under harmonized conditions on 1,000 configurations and evaluated against 2,000 held-out configurations generated using density functional theory, considering force accuracy, computational efficiency, uncertainty and structural stability during 1 ns simulations at 300 K. Allegro produced the lowest validation force mean absolute errors, ranging from 12 to 14 meV per angstrom, whereas SchNet produced the largest, ranging from 84 to 110 meV per angstrom. This ranking did not predict dynamical robustness: NequIP remained stable for 1 ns without additional training data, whereas MACE became stable only after an ensemble-based active-learning procedure added 34 uncertainty-selected configurations. PaiNN and Allegro remained unstable after the addition of 100 and 95 configurations, respectively, and SchNet also failed to achieve stable dynamics within the tested augmentation budget. Under the benchmark hardware conditions, the machine-learning potentials required 6-66 ms per molecular-dynamics step, compared with approximately 20 s for density functional theory. These results show that held-out prediction errors alone are insufficient for selecting interatomic potentials for finite, surface-dominated nanostructures. Reliable deployment requires long-timescale dynamical testing combined with uncertainty-guided refinement, while the effectiveness and data efficiency of active learning remain strongly architecture dependent.
Read the full abstract

All authors M. Usman, M. Suleymanova, Z. U. Abideen, M. Fernández-Pendás, I. Infante

Astrophysics of Galaxies

Highly Ionized Oxygen in the Circumgalactic Medium of TNG100 Galaxies: Distributions, Thermal States, and Ionization Mechanisms

We investigate the origin of the highly ionized oxygen species, O VI, O VII, and O VIII, in the circumgalactic medium (CGM) of galaxies at redshifts $z \approx 0$, by post-processing the outputs of the Illustris TNG100 simulation alongside analytic modeling. Our computations are broadly consistent with observations, and we provide convenient functional fits for the predicted mean oxygen ion column densities as functions of halo virial mass. We determine the relative roles of electron impact collisional ionization versus photoionization by the metagalactic radiation field in producing the ions for halo masses ranging from $10^{10.5}$ to $10^{14.5}$~M$_\odot$. We also determine the thermal gas phases, virialized hot, intermediate temperature, or cool, within which the ions reside. If the halo virial temperature is below the gas temperature at which the ion fractions are maximal for collisional ionization equilibrium (CIE), then photoionization dominates the overall CGM ion mass and column density. When photoionization dominates, the ions are produced primarily in the shock heated virialized phase. If the virial temperature is above the gas temperatures at which the ion fractions are maximal for CIE, then collisions dominate the ionization. However, if the virial temperature is much larger than the ion CIE peak temperature a substantial portion of the ion mass may reside within the intermediate temperature phase as opposed to the hot virialized gas. For MW-mass halos, the O VI and O VII are produced collisionally, and the O VIII by photoionization, all three in the hot, virialized phase.
Read the full abstract

All authors Yossi Oren, Amiel Sternberg, Christopher F. McKee, Yakov Faerman

Machine Learning

Admissable: Training Reinforcement Learning Agents against Adversarial Missingness

In order to make Reinforcement Learning algorithms applicable in real world scenarios, safety must be ensured even under adverse operating conditions. In this work, we consider the challenge of adversarial feature missingness: a scenario in which an adversary occludes features from the agent's observation in order to reduce performance as much as possible. We formally define adversarial missingness for Reinforcement Learning and compare it to the related concepts of $\ell_\infty$-norm bounded adversarial perturbations and learning with missing data. We develop an adversarial training algorithm and show its effectiveness in increasing robustness against adversarial missingness on three MuJoCo benchmark environments. Compared to a baseline trained with random uniform missingness, our method achieves better robustness on all three tasks.
Read the full abstract

All authors Paul Stahlhofen, Luca Hermes, Tim Kochs, Markus Vieth, Barbara Hammer

Artificial Intelligence

Reason What Matters: Retrieval-Grounded Reasoning for Universal Multimodal Embeddings

Universal multimodal embedding (UME) maps multimodal inputs into a shared embedding space for diverse retrieval tasks. Recent methods improve embeddings through Chain-of-Thought (CoT) reasoning optimized with GRPO using retrieval rewards. However, existing methods overlook the mismatch bettween candidate-aware retrieval supervision and input-only CoT generation: (1)trajectory-level rewards convey retrieval outcomes without explicitly identifying the input-supported evidence that distinguishes the positive from hard negatives; (2) input-only generation cannot directly assess whether further reasoning improves retrieval, potentially producing redundant CoTs with substantial latency. To bridge this gap, we propose Reason What Matters (ReWAM), a retrieval-grounded framework that aligns candidate-aware supervision with input-only generation. Specifically, we introduce Retrieval-Aware Self-Distillation (RASD), which extracts privileged guidance from input-supported facts and evidence distinguishing the positive from hard negatives. Conditioned on this guidance, an on-policy self-teacher provides token-level feedback to refine credit assignment, directing policy updates toward retrieval-relevant reasoning grounded in the input. We further propose Retrieval-Adaptive Inference (RAI), which learns a retrieval-aware stopping criterion from prefix-level retrieval feedback. It stops redundant reasoning without candidate access and uses speculative decoding to further reduce CoT latency. Extensive experiments on MMEB-V2 and MRMR demonstrate that ReWAM achieves state-of-the-art retrieval performance while delivering up to 5x the inference throughput of competitive explicit-CoT UME methods. ReWAM thus enables high-quality retrieval through efficient input-only reasoning, making explicit CoT practical for corpus-scale multimodal retrieval. The code will be publicly available.
Read the full abstract

All authors Mingzhou Jiang, Peixi Wu, Hang Cheng, Yunhao Zhou, Biao Yang, Wei Yuan, Yun Li, Fan Yang, Wenwu Ou, Honghui He

High Energy Physics (Experiment)

Knowledge Distillation of a Normalising Flow for Real-Time Anomaly Detection at LHC Level-1 Trigger

We present a new strategy for unsupervised anomaly detection at the hardware-based first stage of event processing (Level-1 trigger) of the Large Hadron Collider (LHC). A normalising flow trained exclusively on Standard Model (SM) events provides a teacher anomaly score based on its exact negative log-likelihood, which is distilled into a compact neural network suitable for field-programmable gate array (FPGA) deployment. The teacher reaches state-of-the-art performance on a benchmark dataset, using a rigorous p-value-based definition of the anomaly score. A simple two-hidden-layer student reproduces the teacher performance across four beyond-the-SM benchmarks to within 0.1 percentage points in area under the receiver operating characteristic curve (AUC), while achieving a compression factor of approximately 325 times. Quantisation-aware training with PQuantML reduces the model to 8-bit precision with AUC changes below 0.06 percentage points. Compilation to Verilog firmware with Alkaid yields FPGA designs that require neither digital signal processors nor block random-access memory and achieve latencies of 27-52 ns. This modular pipeline decouples the expressiveness of the anomaly-detection model from real-time hardware constraints, enabling complex architectures to be used for model-agnostic searches for new physics at the full LHC collision rate of 40 MHz.
Read the full abstract

All authors Jaiman Abson, Florencia Canelli, Valentina Guglielmi, Roope Oskari Niemi, Maurizio Pierini, Chang Sun, Francesco Vaselli

Number Theory

Algebraic independence of the exponential and Weierstrass $\wp$-functions

Let $\Omega$ be a lattice in $\mathbb{C}$ with algebraic invariants and complex multiplication, let $\mathcal{E}$ be the elliptic curve associated with $\Omega$, and let $\wp$ be the Weierstrass function relative to $\Omega$. Set $ k:=\operatorname{End}(\mathcal{E}) \otimes_{\mathbb{Z}}\mathbb{Q}.$ We prove that if $t_1,\dots,t_s$ are $\mathbb{Q}$-linearly independent algebraic numbers and $p_1,\dots,p_n$ are $k$-linearly independent algebraic numbers, then the $s+n$ numbers \[ \mathrm{e}^{t_1},\dots,\mathrm{e}^{t_s}, \wp(p_1),\dots,\wp(p_n) \] are algebraically independent over $\overline{\mathbb{Q}}$. The proof uses the Tannakian description of the Lie algebra of the unipotent radical of the $1$-motive associated with these points.
Read the full abstract

Artificial Intelligence

Why LLM Agents Collapse Without Oversight: The Enforcement Gap as the Mechanism Behind Emergence World Failures

Binding the audit flag in Reflexion-style agents — without changing the auditor — reduces attack success rate substantially, reaching near zero on models whose flags parse cleanly. This single control-flow change exposes the enforcement gap: the controller receives a safety flag and executes anyway. Separating detection probability $p_d$ from enforcement probability $p_e$ establishes that $p_e \approx 0$ by default across every framework we tested, making detection quality formally irrelevant to security when enforcement is absent — a finding consistent with the spontaneous collapses recorded in unsupervised frontier-agent deployments~\citep{emergence2026}. Residual attack success concentrates where flags are unparseable or auditors leak; an RL-trained enforcement controller handles hedged and malformed verdicts that rule-based parsing cannot, cutting ambiguous-critique failure to a fraction of the rule-based baseline. Concurrent filtering and information-flow defenses address detection, not enforcement, leaving the binding constraint untouched. The Audit Enforcement Specification (AES) packages three concrete requirements that close each residue independently; each primitive is adoptable without redesigning the host framework, and no deployed framework currently implements any of them.
Read the full abstract

Artificial Intelligence

ProIQA: A Process-Based Framework for Fine-Grained Math Item Quality Assessment

Automatic Item Generation (AIG) is pivotal for personalized education, yet guaranteeing the pedagogical value of generated items remains a bottleneck. Existing Item Quality Assessment (IQA) methods typically rely on unscalable manual reviews or shallow stem-based metrics, failing to capture the reasoning process required for mathematical problem-solving. To bridge this gap, this paper proposes Process-based Item Quality Assessment (ProIQA), a process-aware framework for fine-grained quality assessment of math items. We first formulate IQA across three heterogeneous dimensions, including knowledge concepts, difficulty, and disciplinary competencies, under a unified process-aware perspective. Based on this formulation, we construct a process-enhanced IQA resource by augmenting original item data with structured reasoning trees derived from raw solutions. Technically, ProIQA leverages Large Language Modelsto construct hierarchical reasoning trees and employs Graph Neural Networks (GNN) to encode their topological dependencies and procedural semantics. The resulting solving representation is fused with stem semantics through a dual-view (“Stem + Solving”) architecture, enabling comprehensive assessment across learning objectives. Extensive experiments on K12 mathematical datasets show that ProIQA effectively captures process-oriented features, offering a scalable data-driven solution for evaluating AIG outputs in intelligent education systems.
Read the full abstract

All authors Junkai Tong, Mingjia Li, Haoran Chen, Yaoyu Jiang, Hanjie Ge, Yixuan Wang, Hong Qian

Functional Analysis

Fractional Volterra-type operators from Bergman spaces with two-sided doubling weights to Hardy spaces

For \(\omega\in\calD\), we give necessary and sufficient symbol conditions for the boundedness and compactness of the Riemann–Liouville family\(V^\varphi_{\alpha,\beta}:A^p_\omega\to H^q\) for every \(0<p,q<\infty\) and admissible \(\alpha,\beta>0\): without any additional weight assumption when \(\alpha\ge\beta\), and, when \(\alpha<\beta\), under the condition \(p(\beta-\alpha)<d_-(\omega)\), where \(d_-(\omega)\) is the critical reverse-doubling exponent. We prove that this inequality is exactly equivalent to the naturally shifted source weight retaining the two-sided doubling and tail geometry required by the reduction; an integrable logarithmic example shows that the endpoint fails. The extension from power weights is not formal, because the reduction replaces \(\omega\) by \(\omega(z)(1-|z|^2)^{p(\alpha-\beta)}\), and a negative shift may destroy even integrability. Combining moment-induced Littlewood–Paley theory for doubling Bergman spaces with an exact coefficient-multiplier comparison, we establish the required Riemann–Liouville transfer, including the finite-dimensional exceptional modes at positive integral shift orders. A finite-codimensional range decomposition and a fractional \(g\)-function reduction then reduce both questions to a single weighted area-map problem. Within the admissible parameter set, the effects of \(\alpha\) are absorbed by the shifted source weight and bounded model corrections, so the final symbol conditions depend on \(\beta\). The thresholds \(p=2\) and \(p=q\) yield pointwise, Carleson, tent-integral, and non-tangential maximal criteria. In the tent-integral range, boundedness already implies compactness.
Read the full abstract

Strongly Correlated Electrons

Twinning of domains and spin anisotropy in K$_5$Fe$_4$Ag$_6$Te$_{10}$

The Fe-based superconductors are derived from metallic parent compounds with nematic and stripe magnetic orders, which lead to two types of magnetic domains. Recently it was found that K$_5$Fe$_4$Ag$_6$Te$_{10}$ (KFAT), an Fe-based semiconductor, exhibits similar nematic and stripe magnetic orders, and is thus an analogue to the Fe-based superconductors in the limit of localized electrons. In this work, the superstructure and magnetic domains of KFAT are elucidated by fully mapping the reciprocal space using time-of-flight single crystal neutron diffraction. In KFAT, Fe and Ag atoms order to form a $\sqrt{5}\times\sqrt{5}$ superstructure containing $2\times2$ Fe blocks, which leads to two superstructure domains with identical main Bragg peaks but distinct superstructure peaks. Below $T_{\rm N}\approx35$~K, magnetic and nematic orders break in-plane rotational symmetry of the tetragonal $\sqrt{5}\times\sqrt{5}$ superstructure, and further give rise to two magnetic domains. These four equally populated domains account for the complex scattering pattern observed in our time-of-flight elastic neutron scattering measurements. Using polarized neutron scattering, we demonstrate a prominent spin anisotropy with an easy-plane spanned by the $c$-axis and the intra-block antiferromagnetic Fe-Fe bond direction. Such an anisotropy at ${\bf q}\neq0$ persists well above $T_{\rm N}$, accounts for the in-plane ${\bf q}=0$ magnetic anisotropy observed in uniaxial-strained KFAT, and offers an indicator for discovering similar piezomagnetic effects in other materials.
Read the full abstract

All authors Jiayu Guo, Hengyang Zhong, Dongsheng Yuan, Xuejuan Gui, Youzhe Chen, Nathan Giles-Donovan, Naomi Kawamura, Masaaki Matsuda, Yaohua Liu, Feng Ye, Rongyan Chen, Robert J. Birgeneau, Xingye Lu, Jincheng Wang, Yu Song

History and Overview of Mathematics

Math for AI safety: an invitation for mathematicians

Artificial intelligence threatens to outrun human understanding and control. New mathematics is needed to design AI that is legible, steerable, and cooperative with humanity. I organize this invitation by mathematical field, so you can turn straight to your own: logic and game theory for cooperation; probability for agency and world-models; algebra and representation theory for learned features; analysis and geometry for generalization and training dynamics. Each section ends with an open problem that is accessible to a working mathematician with no prior experience in AI safety.
Read the full abstract