Peer Revenue Play credits. Real papers.

Papers

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

All of arXiv

Wednesday 9 September

Signal Processing

Sequential Transmit Covariance Optimization for Wireless Tracking Exploiting Observation History

This paper studies a multiple-input multiple-output (MIMO) radar tracking system, where a multi-antenna base station (BS) aims to track the location of a moving target over multiple time slots based on the observed echo signals, initial prior probability density function (PDF) of the target state, and state evolution model. By exploiting the realized observation history in the past time slots, the BS sequentially updates the predictive state information and designs the transmit covariance matrix before collecting the current echo observation. Considering a Gaussian random-walk model for the target location and a Gauss-Markov model for the complex radar cross-section (RCS) coefficient, we propose an effective method to characterize the predictive PDF conditioned on the realized observation history via Gaussian approximation. Based on this, we derive the conditional posterior Fisher information matrix (PFIM) for the target state, and further characterize the conditional posterior Cramér-Rao bound (PCRB) for the mean-squared error (MSE) in estimating the target's location state as an explicit expression of the transmit covariance matrix. Next, we formulate the sequential transmit covariance matrix optimization problem to minimize the conditional PCRB for each time slot. Despite the non-convexity of the problem, we obtain the optimal solution via the Schur complement technique. Numerical results show that the proposed design effectively exploits the realized observation history, achieves a lower conditional PCRB than the benchmark schemes, and improves tracking accuracy over time.
Read the full abstract

Software Engineering

If It's Not Buggy, Don't Fix It: On the Dynamics of Iterative Bug-fixing with LLMs

Large language models (LLMs) have become ubiquitous in software development, with LLM-based automated program repair tools increasingly used during code review. In this report, we explore the iterative blind use of LLMs as bug-fixers. Across multiple models and repair environments, we find that LLMs consistently claim to detect bugs in entirely bug-free programs while the rate of repair of buggy programs is less than that of the damage to correct programs. We also explore the long-term dynamics of this iterative process, and find that this frequently reaches a pseudo-bug-fixing cycle where the same changes are added and removed again ad infinitum. Lastly, via mechanistic probing, we unveil the existence of a steering vector which controls the editing propensity, suggesting that LLMs have an internal representation of “buggy code", and that this representation is what is falsely activated to induce pseudo-bug fixing. These results provide insight towards the dynamics of fully autonomous bug-fixing systems, as well as stopping conditions under ambiguous goals.
Read the full abstract

Computation and Language

ProbPlug: A Plugin Uncertainty Network for Reliable Confidence in LLM Binary Classification

Large language models (LLMs) have achieved strong performance across a broad range of classification settings, yet the reliability of their predictions remains a major obstacle to deployment in high-stakes scenarios. Although confidence estimation for LLMs has been widely studied, confidence calibration for LLM-based classification remains underexplored. We introduce ProbPlug, a lightweight confidence estimation framework for LLM-based binary classification, which predicts whether an output is correct using internal token features extracted from a frozen LLM. ProbPlug employs a self-attention module to aggregate hidden representations and can be integrated into the original inference pipeline without modifying the base model. Experiments across multiple tasks involving both text-based and multimodal large models show that ProbPlug provides more reliable confidence estimates, improves classification performance with negligible additional overhead, and exhibits strong generalization across tasks. These results indicate that ProbPlug serves as a practical solution for confidence estimation in LLM-based classification. Our code is publicly available at Github.
Read the full abstract

All authors Jianzong Wang, Chuhang Liu, Botao Zhao, Zuheng Kang, Xulong Zhang, Xiaoyang Qu, Junqing Peng, Zhiewei Ye, Yayun He

Quantitative Methods

ADMET-EvO: a self-evolving scientific agent for sustained research across heterogeneous tasks

Scientific agents can move beyond automated model building by using accumulated evidence to revise both their questions and experimental strategies. The challenge is sustaining this adaptation across heterogeneous tasks without overfitting decisions to internal validation. Absorption, distribution, metabolism, excretion and toxicity (ADMET) prediction provides a demanding setting across diverse assays, datasets and chemical domains. We therefore developed ADMET-EvO, an evidence-gated agent that formalizes endpoints, generates falsifiable hypotheses and tests interventions across data, feature and model axes. It carries supported, rejected and inconclusive outcomes forward to guide each new cycle. Across the 22-task Therapeutics Data Commons (TDC) ADMET benchmark, ADMET-EvO achieved the highest task-normalized score of 96.77. Evidence-guided selection reduced cumulative fitting time by 72.2% within a predefined non-inferiority margin. It also formalized 43 toxicity-related tasks and constructed endpoint-specific predictors. Together, these results show how ADMET-EvO can accumulate evidence, revise its strategy and expand its research scope over time.
Read the full abstract

All authors Yiling Zhou, Yilin Wang, Jianmin Wang, Heqin Zhu, Zirui Wang, Chang-yu Hiesh, Kejun Ying, Jiaqi Wang, Yuzhi Xu, Tingjun Hou, Odin Zhang

Dynamical Systems

On the dynamics of self-consistent transfer operators for interval maps with singularities

We study the dynamics of self-consistent transfer operators associated with mean-field coupled systems whose unit dynamics is given by an interval map with finitely many singularities of cusp type. We show that, when the coupling strength is sufficiently weak, such an operator has a unique fixed point in a suitable cone of the associated Banach space, and that this fixed point attracts the orbit of every density in the cone at an exponential rate in the corresponding norm.
Read the full abstract

Operator Algebras

Regularity for Subfactors

We study three naturally associated notions of regularity for a unital inclusion of von Neumann algebras $\mathcal{B} \subseteq \mathcal{M}$: regularity, groupoid regularity, and unitary regularity. We first observe that regularity and groupoid regularity are unconditionally equivalent by showing that the standard normaliser and the groupoid normaliser of $\mathcal{B}$ in $\mathcal M$ generate the same linear span. The central question addressed in this work is whether regularity implies unitary regularity. For arbitrary von Neumann subalgebras, we demonstrate that this implication fails dramatically; we construct explicit examples to show that every factor (of type $\mathrm{I}_\infty$, $\mathrm{II}_1$, $\mathrm{II}_\infty$, and $\mathrm{III}$) contains a regular von Neumann subalgebra that is not unitarily regular. In stark contrast, we prove that the behaviour changes completely in the setting of subfactors. For subfactor inclusions across all these types, regularity always implies unitary regularity, and thus all three notions of regularity coincide.
Read the full abstract

Cryptography and Security

Understanding the Security Boundary of Obfuscation-based On-Device LLM Protection

Trusted Execution Environments (TEEs) offer a promising mechanism for safeguarding the intellectual property of on-device Large Language Models (LLMs). To overcome the inherent computational bottlenecks of TEEs, existing TEE-Shielded LLM Partition (TSLP) methods apply efficient obfuscation schemes to computationally intensive layers, offloading them to external GPUs while retaining only lightweight operations within the TEE. Although a growing body of TSLP-based approaches has emerged, these defense mechanisms remain largely heuristic. Consequently, some methods are proven vulnerable to certain specialized adversarial attacks designed to exploit their specific architectural implementations. To overcome the limitations of these heuristic designs, this paper addresses a fundamental research question: can we establish common primitives to unify representative prior methodologies, characterize the security boundary of their compositions, and systematically extend them? To this end, we formalize a set of obfuscation primitives, defined as dual-tuples of linear computations satisfying specific algebraic properties. We demonstrate that the matrix-level weight transformations of several representative efficient TSLP frameworks can be expressed as compositions of these primitives; consequently, the canonical form of these primitive compositions, denoted as \priorboundary, defines the security boundary of this primitive family. We then expose the vulnerabilities of \priorboundary through a novel primitive-guided attack methodology, \sysattack, demonstrating a shared vulnerability in several prominent TSLP methods published in top-tier venues, such as ArrowCloak (Security'25), TSQP (S&P'25), and LoRO (NeurIPS'25). Finally, we introduce two novel obfuscation primitives and integrate them with existing constructs to formulate \sysdefense, extending the prior security boundary \priorboundary.
Read the full abstract

All authors Hanyi Zhou, Chenyang Li, Yuanzhe Pang, Ke Xu, Mingwei Xu, Zhuotao Liu

Instrumentation and Detectors

Development of Ultrafast and Radiation-Hard GAGG for the Next-Generation of High-Energy Physics Calorimeters

The evolution of High Energy Physics (HEP) toward future collider experiments with High Luminosity (HL), such as the HL-LHC, requires the development of scintillating materials that combine high density, excellent radiation hardness and an ultrafast response. While Cerium-doped Gadolinium Aluminum Gallium Garnet (GAGG:Ce) offers a very high light yield and resilience to irradiation, its typical decay time of approximately 50-60 ns may lead to pile-up effects in high-rate environments. In this paper, we report on the development and multi-stage characterization of various accelerated GAGG compositions optimized for timing performance and grown by Crytur. By taking advantage of divalent co-doping to engineer the scintillation kinetics, we achieved an effective decay time (tau_d,eff) down to 5.5 ns while maintaining a competitive light yield of several thousand photons per MeV. Laboratory characterization demonstrates that the time resolution under gamma-ray excitation is comparable to commercial GAGG, while the time resolution measured with 120 GeV pions reaches performance levels comparable to state-of-the-art LYSO:Ce,Ca. After a 1 MGy proton irradiation campaign the material retains most of its optical transmission. The results confirm that this ultrafast GAGG composition is a viable candidate for the next generation of HEP calorimetry and timing detectors.
Read the full abstract

All authors Louis Roux, Loris Martinazzoli, Julie Delenne, Philipp Roloff, Ondřej Zapadlík, Jan Polak, Jan Havlíček, Silvia Sýkorová, Martin Nikl, Pavel Boháček, Christophe Dujardin, Etiennette Auffray

Functional Analysis

Littlewood subordination for de Branges–Rovnyak spaces

In this paper, we study composition operators that act between different de Branges-Rovnyak spaces. Our main results are suggested by a paper of Mashreghi and Shabankhah concerning composition operators between model spaces. We also answer several related open questions posed by Dellepiane and Seco. To prove these results, we apply reproducing kernel Hilbert space methods, Sarason's approach to composition operators as integral operators, and Aleksandrov-Clark measures.
Read the full abstract

All authors Emmanuel Fricain, Andreas Hartmann, William T. Ross, Dan Timotin

Logic in Computer Science

Policy Change for Treelike Monitors

We study the policy change problem that arises in the runtime verification of long-running systems. The online monitors typically used in this context are generally treelike, in that they maintain substates that monitor subformulae of the target policy. We consider when and how the policy can be changed while the monitored system is running by only exploiting the information stored in the monitor's state. This is relevant, for example, to account for new system functionality or changes in regulatory requirements. We formally define the policy change problem in a general setting, independent of any specific (treelike) monitor implementation. We then show that policy change for past-time metric temporal logic (pMTL) is decidable but has tight non-primitive recursive lower and upper bounds, while with discrete-time semantics it is EXPSPACE-complete.
Read the full abstract

Computation and Language

Data-Centric Post-Training for Financial Reasoning: Mining, Distillation, and Verifiable Learning

Financial text, textbooks, and question-answer pairs are abundant, but only a small fraction is directly usable for reasoning-focused post-training. Existing QA pairs often lack explicit reasoning, sufficient context, or reliably verifiable answers, while textbooks must first be transformed into synthetic training examples. We present a data-centric pipeline that constructs complementary corpora by mining open-source reasoning traces, distilling financial instruction data, and generating knowledge-graph-guided question-answer pairs from financial educational material. After semantic deduplication, three lightweight sequence classifiers select finance-relevant examples, reject under-specified questions, and identify tasks suitable for reinforcement learning with compact rule-based verifiers. For model adaptation, we study supervised fine-tuning and reinforcement learning, while self-distilled fine-tuning and post-training model merging are used to prevent the loss of financial capabilities already present in the starting model. We evaluate the adapted language models using FINESSE-Bench, reporting aggregate performance and changes relative to their starting checkpoints. Across the selected comparisons, ordinary SFT reduces FINESSE-Bench accuracy by 3.2-4.0 percentage points, whereas self-distilled SFT improves over the corresponding starting models by 1.0-2.8 points. Equal-weight merging recovers 3.0 points over its SFT parent and finishes 0.9 points above the original model; GRPO on hard tasks adds 0.4 points after self-distilled SFT or 3.0 points when applied directly to verifiable tasks. These results show that retention-aware adaptation can improve financial reasoning without the regressions observed after ordinary SFT.
Read the full abstract

All authors Zhirayr Hayrapetyan, Andrei Kalmykov, Denis Kokosinskii, Dmitry Stanishevskii, Dmitry Zmitrovich

Machine Learning

Storage-Scalable Progressive Semantic Communication via Knowledge-Base Reuse

Existing knowledge-base-assisted semantic communication schemes commonly adopt either single knowledge-base quantization (SKBQ) or multi-knowledge-base residual quantization (MKBQ). SKBQ incurs limited storage overhead but has restricted quantization capacity, whereas MKBQ supports progressive refinement by assigning an independent knowledge base (KB) to each stage, causing the KB storage to grow linearly with the transmission depth. To address this problem, we propose storage-scalable knowledge-base reuse quantization (SSKBQ), which reuses a compact set of KBs across multiple residual refinement stages and thereby decouples the number of transmission stages from the number of maintained KBs. A stage-aware residual supervision mechanism is further introduced to regularize intermediate quantized representations and encourage progressive refinement. Experimental results demonstrate that KB reuse provides an effective solution to the storage scalability problem while maintaining competitive progressive reconstruction performance.
Read the full abstract

All authors Heng Zhu, Ye Liu, Kun Zhu, Feifei Song

Quantum Physics

Multiparameter quantum bounds for entanglement-assisted aperture synthesis

Entanglement-assisted optical interferometry promises diffraction-limited imaging over baselines where light cannot be physically combined, but existing theory treats only a single visibility on one two-station baseline. This work formulates the multiparameter estimation problem for an M-station array imaging an extended scene. A reduction lemma shows that the multimode quantum Fisher information (QFI) equals the mean photon number times the QFI of a single delocalized photon, collapsing the problem to a finite-dimensional one. The local photon-number superselection rule (SSR) erases all phase information from the bare state, while shared entanglement restores a fraction f(r)=r/(1+r) of the QFI on a baseline supplied with r pairs – a factor shown to be achievable, since the reduced two-mode state of any pair is exactly the two-station weak-thermal state. The QFI matrix and its mean Uhlmann curvature reveal numerically that a point source is measurement-compatible, whereas for every extended-source model examined it is incompatible, the Holevo bound exceeding the symmetric-logarithmic-derivative (SLD) bound by up to ~74%. An explicit collective receiver – a global mode-sorting (quantum Fourier transform) measurement – then attains a weighted variance within ~7% of the SLD bound and roughly an order of magnitude below an explicit pairwise receiver, and a noisy-resource advantage threshold gives the break-even baseline (~20 km for near-term parameters) beyond which repeater-distributed entanglement beats direct transmission. Finally, the optimal allocation of a finite entanglement budget is a convex program with a closed-form proportional (square-root-law) solution favoring low-visibility baselines. These results recast quantum aperture synthesis as collective multiparameter estimation and give concrete design targets, illustrated on the CHARA array.
Read the full abstract

Representation Theory

Derived equivalences between diagram categories of finite posets

We study universal derived equivalences between diagram categories indexed by finite posets. Starting from a construction of Ladkani, we give an intrinsic criterion for determining when a finite poset admits a decomposition to which this construction can be applied. This leads to the notion of an admissible cut, formulated entirely in terms of the order structure of the poset. Our main result therefore provides a method for producing, from a given finite poset admitting such a cut, a new poset that is universally derived equivalent to it. The construction is reversible once the partition is retained, so that the original mixed order relations can be recovered from the transformed poset. As applications, we show that every finite poset of height at most one is universally derived equivalent to its opposite, give a criterion characterizing source-to-sink transformations at minimal elements, and recover the universal derived equivalence of all orientations of a finite tree through sequences of such local transformations. These results are also applied to persistence modules indexed by finite posets and to extensions obtained by adjoining further finite parameters.
Read the full abstract

Instrumentation and Methods for Astrophysics

Enabling New Discoveries with Machine Learning

The next generation of telescopes such as the Square Kilometre Array and the Vera C. Rubin Observatory will produce enormous quantities of data, too large for traditional analysis techniques. Machine learning has proven invaluable in handling massive data volumes and automating tasks traditionally done by human scientists. The question is, can machine learning also automate scientific discovery? Recent advances in the field have made this feasible but only when combined with expert human input through active learning. Astronomaly is a publicly available framework for active anomaly detection in astronomy. It learns to make recommendations of interesting sources to a user, allowing scientists to quickly uncover rare and new classes of objects in large datasets. In this paper, we review some of the discoveries made using Astronomaly, discuss the latest deep learning-based approaches and explore the critical role of the cyber-human interface in unveiling cosmic mysteries hidden in massive astronomical datasets.
Read the full abstract

Machine Learning

A Trust-Network-Based Federated Learning Framework for Multi-Center Aging Clock Prediction

Aging clocks quantify biological aging and help characterize individual health status. What protein interactions are important for accurate aging clocks, and are they zeroth-order or higher-order? Addressing these questions requires learning from large molecular datasets distributed across medical centers, where privacy constraints prevent centralized data sharing. Federated learning offers a natural solution but faces four challenges in this setting: limited local sample sizes, sparse and directional inter-center trust, the need to retain discriminative age prediction while supporting interpretation, and model drift and forgetting under heterogeneous cross-center data. We propose TNFL, a trust-network-based federated learning framework that progressively propagates models along directed pairwise trust relations without centralized aggregation. TNFL combines an age-aware mixture-of-experts model with generative replay to preserve previously learned information and reduce forgetting and drift. Experiments across multiple molecular datasets show that TNFL enables effective aging-clock prediction with limited local data, provides interpretable age-dependent prediction patterns, and maintains stable performance across interaction orders. To investigate the biological questions, we analyze TNFL-identified pairwise protein interactions and their higher-order organization through functional and network analyses. The identified interactions repeatedly form coordinated higher-order subnetworks spanning multiple aging-related biological systems, with several proteins recurring across subnetworks. These findings suggest that TNFL captures molecular relationships beyond isolated pairwise associations and reveals coherent higher-order biological organization associated with aging.
Read the full abstract

All authors Chunxu Zhang, Bo Li, Wenliang Wang, Yang Liu, Di Jiang, Yuan Huang, Yo-ichi Nabeshima, Akinori Yamamura, Bo Yang, Qiang Yang

Combinatorics

On $S$-packing total colorings

In this paper, we generalize the concept of packing total coloring by introducing a new concept called the $S$-packing total coloring. For a graph $G$ and a non-decreasing sequence $S=(a_1,a_2,\ldots)$ of positive integers, an $S$-packing total coloring of $G$ is a mapping $c: V(G)\cup E(G)\rightarrow \{1,2,\ldots\}$ such that for any two distinct elements $A,B\in V(G)\cup E(G)$ with $c(A)=c(B)=i$, the distance between $A$ and $B$ is at least $a_i+1$. The smallest integer $k$ such that $G$ admits an $S$-packing total coloring using $k$ colors is called the $S$-packing total chromatic number of $G$, denoted by $\chi_S^{''}(G)$. For any sequence $S$, we establish general lower and upper bounds for $\chi_S^{''}(G)$, and characterize all graphs $G$ with $\chi_S^{''}(G)\in\{1,2,3\}$. Furthermore, we investigate $S$-packing total chromatic numbers of complete bipartite graphs, as well as infinite and finite paths and cycles.
Read the full abstract

All authors Jasmina Ferme, Jaka Hedžet, Petra Melicharova, Daša Mesarič Štesl

Statistical Mechanics

Infinite ergodic theory and functional statistics of non-confined Feller process

We study additive observables of a non-confined Feller process characterized by a non- normalizable stationary probability density and return times with infinite mean. The process is equivalent, after a deterministic rescaling, to a squared Bessel process, but we focus here on a question: the spatial structure of its local-time field. We show that the local time admits a fac- torization in the long time limit. Its spatial profile is governed by the non-normalizable stationary density, whereas its temporal fluctuations are controlled by a single Mittag-Leffler random ampli- tude. As a consequence, normalized spatial correlations of the local time converge to a universal constant independent of the two observation levels. Occupation times of finite intervals follow as corollaries and display Darling-Kac fluctuations. In contrast, non-integrable power observables exhibit self-similar squared-Bessel-type limits rather than Mittag-Leffler statistics. This spatial- field perspective provides a unifying framework for infinite ergodic theory under state-dependent multiplicative noise.
Read the full abstract

Computers and Society

Beyond Training: A Feasibility Taxonomy for Inference-Time AI Governance

Compute governance today is a governance of training: the thresholds, reporting requirements, and frontier-AI regimes now in force attach to training compute and treat the trained model as the regulatory unit. That picture is incomplete: capability increasingly migrates to the deployment stage through inference-time scaling, agentic scaffolding, and compression onto consumer hardware. This paper asks which mechanisms are available once the regulatory object shifts from the training run to the inference call. We develop a feasibility taxonomy of twenty inference-time mechanisms across monitoring, verification, and enforcement, each rated on a four-point readiness scale against a documented four-vendor evidence base. We then stress the taxonomy against a two-dimensional adversary model (three capability tiers crossed with four adversary roles) and map each mechanism to four governance scenarios (domestic regulation, bilateral or multilateral coordination, industry self-regulation, and compute-marketplace governance). Fifteen of the twenty mechanisms have commercial technical substrates in production today, although governance-grade assurance and adversarial robustness vary substantially. The adversary analysis shows that this readiness holds only against a cooperative deployer and a low-to-medium-capability user: no mechanism rates adequate against a high-capability state-level deployer, and fine-tuning removes the model-internal components of the enforcement cluster, although platform-external controls can persist. A substitution analysis connects the taxonomy to a companion hardware paper as a conditional substitution principle describing when inference-stage and hardware-stage mechanisms provide comparable regulatory coverage under stated conditions. A second-rater reliability check on a random subset of the readiness ratings returned a quadratic-weighted Cohen's kappa of 0.74.
Read the full abstract

Cryptography and Security

Distributed and Private Textual Data Synthesis from Embeddings

We revisit differentially private (DP) text synthesis in the realistic setting of distributed users, where privacy concerns preclude a trusted curator with access to raw user texts. Existing DP text synthesis pipelines are designed for a trusted, centralized curator and often cannot be deployed in distributed settings due to unrealistic trust and access assumptions; when adapted naively, they require repeated, tightly synchronized user participation and incur significant overhead. To address this gap, we propose a DP–cryptography co-design for textual data synthesis that requires no trusted curator and requires only lightweight user participation. Our approach has two optimized components. First, we design a distributed-friendly DP synthesis algorithm that releases a one-time DP summary in an embedding space: it identifies frequent semantic regions and releases their DP centroids, enabling training-free, non-iterative offline text synthesis. We further introduce semantic support protection, which ensures the released summary avoids semantic neighborhoods of infrequent texts, reducing the risk of exposing rare user data. Second, we develop a custom secure protocol that implements this algorithm over distributed user data, enforcing end-to-end DP guarantees without requiring a trusted curator. On four benchmarks, we achieve utility comparable to the state-of-the-art centralized DP synthesis method.
Read the full abstract

All authors Ergute Bao, Hongyan Chang, Ali Shahin Shamsabadi, Ting Yu, Xiaokui Xiao

Chemical Physics

Modeling electrolytes in nanopores by Monte Carlo simulations and the Bazant–Storey–Kornyshev model

We study electrolyte-filled cylindrical nanopores through Monte Carlo (MC) simulations and the Bazant–Storey–Kornyshev model, for different ionic valencies, sizes, and bulk concentrations, pore radii and surface charge densities. Our model accounts for finite ion size through a Stern layer and we use de~Souza and Bazant's mechanical equilibrium principle to derive a boundary condition for the outer Helmholtz plane. For 1:1 electrolytes and moderate surface charge densities, we find that both the classical Poisson–Boltzmann–Stern (PB–Stern) and our Bazant–Storey–Kornyshev–Boltzmann–Stern (BSKB–Stern) model fit MC data well—for 2:1 and 3:1 electrolytes, the BSKB–Stern outperforms the PB-Stern model. Conversely, the BSKB–Stern model poorly fits MC data in other scenarios. First, BSKB–Stern does not capture drying and apparent like-charge attraction in 3:1 electrolytes and small surface charge densities. Second, BSKB–Stern only fits MC data for finely-tuned ionic diameters; for other ionic diameters, the ionic charge densities show oscillations or extended near-surface regions caused by ionic packing and strong Coulomb interactions, not captured by the BSKB–Stern model.
Read the full abstract

All authors Nader Nekoubin, David Fertig, Dezső Boda, Mathijs Janssen

General Relativity and Quantum Cosmology

Black hole and wormhole branches in gravitational decoupling

Minimal Geometric Deformation (MGD) applied to a static Schwarzschild black hole seed generates a single decoupler function $h(r)$, obtained by solving the $\theta$-sector field equations together with an equation of state. Once $h(r)$ is fixed, the resulting one-parameter family is controlled by the coupling strength $k$ through $F(r;k)=1+k\,h(r)$. We show that, whenever the deformation develops a simple outermost root that crosses the seed horizon, the same fixed decoupler leads to two mutually exclusive branches associated with different global completions: on one side of the critical coupling the deformed metric preserves the seed horizon as a black hole, whereas on the other side the root $r_*>2M$ lies in the exterior and cannot be interpreted as an interior modification of the black hole geometry. We prove that this root forces a loss of Lorentzian signature on the interval $(2M,r_*)$, so that no smooth extension of the exterior metric through the seed horizon $r=2M$ exists once $r_*$ lies outside it. Within the static, spherically symmetric class considered here, the corresponding smooth Lorentzian completion is a two-ended wormhole obtained by excising $(2M,r_*)$ and doubling the region $r\geq r_*$ across the minimal sphere $\mathcal T=\{r=r_*\}$. No topology change of any single spacetime is claimed or required: $k>k_c$ and $k<k_c$ simply correspond to two different, non-diffeomorphic manifolds, and Lemma~1 below shows that the metric itself dictates which of the two is the admissible completion for a given $k$. We compute the second homology group of both completions explicitly, $H_2(\Sigma_{\rm BH},\mathcal H)=0$ for the black hole exterior relative to its horizon and $H_2(\Sigma_{\rm WH})\cong\mathbb Z$ for the completed wormhole manifold, giving a discrete invariant that distinguishes the two branches.
Read the full abstract

All authors Francisco Tello-Ortiz, Y. Gomez-Leyton, Vitalii Vertogradov, Jean Baez Cuevas

Computational Geometry

A 2.37332-Competitive Algorithm for Online Square Packing with Gravity

We consider online packing of axis-parallel squares into a unit-width strip under the Tetris and gravity constraints: An incoming square must be lowered from above along a monotonic downwards path until it reaches support from below. Fekete, Kamphans, and Schweer [Algorithmica, 2014] gave an algorithm with asymptotic competitive ratio $34/13\approx2.6154$ in this model. We present $\mathrm{AsymmetricSlots}$, a recursive algorithm based on splitting each slot into a wide and narrow subslot. The proof uses a local charging argument: squares that are large relative to its associated slot pay for the height they create with their own area, while smaller squares are balanced between the two subslots and may use a bounded temporary credit. For a suitable split parameter $p^\star$, we prove $\mathrm{AsymmetricSlots}{p^\star}(\sigma)\le 2.37332 \operatorname{OPT} (\sigma)+O(1)$ for every input sequence $\sigma$. Additionally, we show that the same framework gives an algorithm with asymptotic competitive ratio $O(\kappa)$ for rectangles of aspect ratio at most $\kappa$, and a matching $ \Omega(\kappa)$ lower bound shows that the dependence on $\kappa$ is asymptotically optimal. For the square algorithm, we give a lower bound of $2$ on its asymptotic competitive ratio.
Read the full abstract

Quantum Physics

In situ characterization of a photon-subtraction device via heralding counts and homodyne detection

Photon subtraction is one of the most important techniques for generating non-Gaussian optical states and constitutes a key resource for quantum information processing and quantum metrology. The practical performance of a photon-subtraction device is primarily determined by the transmissivity of the beam splitter and the quantum efficiency of the heralding detector. Accurate knowledge of these parameters is therefore essential for assessing the quality of the generated non-classical states. In this work, we propose an experimentally feasible in situ scheme for the simultaneous estimation of these two parameters using only the measurement data produced during the operation of the device. Our protocol combines the click statistics of an on/off heralding detector with homodyne measurements performed on the transmitted mode of the beam splitter when fed by a displaced squeezed state. Within the framework of classical multi-parameter estimation theory, we derive the corresponding Fisher information matrix and investigate both joint and sequential estimation strategies. For simultaneous measurement of parameters, we evaluate the sloppiness of the underlying statistical model and analyze its dependence on the measured quadrature, probe photon number, squeezing fraction, beam splitter transmissivity, and detector efficiency. Our analysis proves that an appropriate choice of the homodyne quadrature substantially reduces parameter degeneracy and enables efficient simultaneous estimation. Furthermore, we show that the joint estimation strategy consistently provides a lower estimation bound than the sequential estimation approach over a broad range of experimentally relevant parameters.
Read the full abstract

Machine Learning

A Systematic Evaluation of Molecule Generation Models for De Novo Drug Design: From Benchmarks to Practical Insights

Molecule generation has emerged as a powerful computational tool for de novo drug design, enabling the exploration of chemical space beyond the limits of conventional virtual screening. The field has progressed rapidly, driven by advances in molecular representations, generative architectures, and target-aware modeling strategies. However, existing reviews typically address specific model families or application scenarios in isolation, rather than offering an integrated perspective on how these components collectively form a coherent generation workflow. In this review, we present a comprehensive evaluation of molecule generation models for de novo drug design, covering 82 methods across five deep generative frameworks, including recurrent neural network (RNN)- and Transformer-based models, variational autoencoders (VAEs), generative adversarial networks (GANs), flow-based models, and diffusion models. We first summarize widely used benchmarks and molecular representations, and then examine the methodological principles underlying both general and pocket-conditioned generation. A central contribution of this work is a systematic synthesis and comparative analysis of reported performance across commonly used benchmarks and evaluation metrics. We also summarize representative experimentally validated case studies. Looking ahead, we discuss future directions in standardized 3D data, interaction-aware generation, receptor flexibility, and multi-objective molecular design, with the aim of improving the reliability and experimental relevance of molecule generation. All collected benchmark resources, evaluation metrics, and model references are provided in a publicly accessible repository at https://github.com/JacklinGroup/molecule-generation-review.
Read the full abstract

All authors Xinrui Xu, Xueer Wang, Dan Luo, Sisi Yuan, Xuan Lin

Analysis of PDEs

Secular Instability and the Turning-Point Principle for Rigidly Rotating Viscous Stars

We study axisymmetric stability of rigidly rotating viscous stars modeled by the free-boundary Navier–Stokes–Poisson (NSP) system. The unstable index of the linearized NSP generator, counted with Riesz algebraic multiplicity, equals the Morse index of the augmented energy at fixed mass and total angular momentum. We prove an abstract Kelvin–Tait–Chetaev theorem for damped gyroscopic equations with finite negative stiffness index, requiring no compactness assumptions and no spectral gap at zero; the stiffness kernel may be infinite-dimensional. In the NSP application, the viscous dissipation is degenerate: its axisymmetric kernel, generated by rigid axial translation and rigid rotation, is projected out before applying the abstract theorem, while conservation laws and a Routh reduction transfer the resulting instability index back to the full NSP generator. For slowly rotating branches with fixed total angular momentum, viscous instability begins at the continuation of a first nondegenerate spherical mass maximum. By contrast, for every sufficiently small nonzero angular momentum, the corresponding Euler–Poisson star remains axisymmetrically spectrally stable on an interval beyond this maximum. Thus dissipation can destabilize a rotating star before the corresponding inviscid star becomes unstable. This separation of the stability thresholds results from viscous redistribution of angular momentum, which removes the inviscid constraints on its material distribution while preserving its total value.
Read the full abstract

Astrophysics of Galaxies

The JWST Emission Line Survey (JELS): Multi-wavelength properties of Paschen line-emitters at Cosmic Noon

Context: Narrowband JWST/NIRCam selection isolates star-forming galaxies in well-defined redshift slices, while ALMA and archival far-IR/submm data reveal obscured growth. We combine these to measure Paschen-line SFRs, infer nebular extinction from EAZY Av, and probe FIR-to-submm emission via weighted-median stacking, to constrain galaxy growth at cosmic noon. Aims: We characterize stellar, nebular, and FIR properties of Paschen-selected galaxies in COSMOS, in two ensembles at z~1.50 (Pa-alpha) and z~2.65 (Pa-beta), quantifying short- vs long-timescale SF and dust attenuation, and constraining obscured SFR and gas masses. Methods: We select robust narrowband-excess samples from the JELS F466N/F470N mosaic (Pa-alpha: 77, Pa-beta: 31 sources), remove AGN via X-ray/radio matching, and assemble multiwavelength photometry (JWST, HST, Spitzer, Herschel, SCUBA-2, AzTEC, ALMA). Photo-z/stellar properties come from EAZY/STARDUST; line SFRs use recent Paschen calibrations, with nebular extinction from EAZY Av. Results: EAZY fits show stellar masses log(Mstar/Msun)~7.4-10.3 and modest SFRs (0.1-27 Msun/yr). Line SFRs are 0.1-22.0 and 0.7-47.9 Msun/yr for Pa-alpha/beta. Most sources show short-term bursts (line SFRs exceed SED SFRs), weakly decreasing with mass. Nebular attenuation correlates with mass and SFR. Stacking gives no >3sigma detections beyond MIRI. The weighted-median SED gives SFR_UV,med = 0.48+-0.11 and 3.92+-0.47 Msun/yr for Pa-alpha/beta. ALMA Band 4 gives the tightest constraints: median 3-sigma limits SFR_IR<53 (Pa-alpha) and <44 Msun/yr (Pa-beta); gas mass <10^11.2 (Pa-alpha) and <10^10.6 Msun (Pa-beta). Conclusions: Paschen-selected samples show widespread short-timescale star formation, but FIR/sub-mm data give only upper limits on obscured SFR/ISM mass. Deeper spectroscopy and ALMA continuum data are needed to constrain obscured SF and gas reservoirs.
Read the full abstract

All authors Malte Brinch, Edo Ibar, Luis Ossa-Fuentes, Corey Pirie, Philip Best, R. K. Cochrane, Kenneth Duncan, James Dunlop, Sophia Flury, Catherine Hale, Jorryt Matthee, Derek McLeod, Joshua Selfridge, David Sobral, Harry Stephenson, J. P. Stott, A.M. Swinbank

Optimization and Control

Observability and controllability for the Schrödinger equation on polyhedra

We consider the observability and controllability for the Schrödinger equation on polyhedra. We show that for the Schrödinger equation on convex polyhedra in $\mathbb{R}^d\,(d\geq2)$, when the initial data possess $H^s\,(s>d/2)$ regularity, observability and controllability hold, with the control region being an arbitrary nonempty open neighborhood of the singular set of the boundary, which is the same as the control region for eigenfunctions. Our proof of observability is a concrete adaptation of the method in the Burq–Zworski black box control work, where the observability inequality is proved via a resolvent estimate. Hence, before proving observability, we present an observability resolvent estimate on polyhedra. The idea of the proof of this resolvent estimate comes from the case of eigenfunction concentration on polyhedra considered by Cekić–Georgiev–Mukherjee, which uses semiclassical measures and relies closely on the dynamical properties of the billiard flow in polyhedra. Finally, by a standard HUM argument, we show that observability implies controllability.
Read the full abstract

Computer Vision

LinearMask-GS: Stable-Mask Importance Pruning for Compact 3D Gaussian Splatting

3D Gaussian Splatting (3DGS) enables real-time novel view synthesis but produces millions of primitives through adaptive densification, leading to significant storage overhead. Learned-mask pruning methods such as LP-3DGS address this by assigning each Gaussian a learnable mask to identify and prune redundant primitives. However, we identify a limitation of this paradigm: the steep slope of the Gumbel-Sigmoid activation drives mask values to the extremes within the short mask-training window, before the importance ranking has stabilized, producing a sharply bimodal distribution from which that ranking can no longer be reliably recovered. We propose LinearMask-GS, which replaces Gumbel-Sigmoid with a linear increment activation that keeps mask values in a mid-confidence regime throughout mask training, producing a stable, unimodal mask distribution whose ranking tracks importance. On Mip-NeRF 360, our method achieves 3.6x and 1.6x Gaussian reductions over 3DGS and LP-3DGS, respectively, while maintaining or improving rendering quality. For outdoor scenes, it yields a 1.6x reduction (from 2.18M to 1.36M) with notable gains in PSNR (+0.38 dB), SSIM (+0.025), and LPIPS (-0.029).
Read the full abstract