Combinatorics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 31 August
In a preceding article, we introduced stable symmetric series encoding simultaneously the normalized conjugacy classes of all symmetric groups. The same rational series remain stable for the Jack-deformed Goulden–Jackson product. We investigate their two-parameter Macdonald analogue. Starting from the $(q,t)$-deformed class product (dual to the coproduct diagonal on the $J$-basis), we construct the unique infinite series whose multiplication realizes any shifted Macdonald eigenvalue. In contrast with the classical and Jack cases, this transform is no longer multiplication by a fixed explicit series. We identify it with a composition of a Cauchy multiplication, the integral nabla operator, and a simple diagonal operator. We then determine the series realizing the Nazarov–Sklyanin operators $A^{(k)}$, derive a single generating series for all column partitions, and compare our construction with the Macdonald characters and Theta operators of Ben Dali and D'Adderio.
Materials Science
Fine structure of the M-center in Si
Color centers in silicon offer great possibilities for scalable quantum technologies. The M-center, proposed to originate from a carbon-hydrogen complex, offers telecommunications-band emission and a paramagnetic ground state similar to the T-center. Here, we report on photoluminescence lines in the vicinity of the M-center and investigate their properties, including their dependence on temperature, implantation fluence, implantation isotope, and annealing temperature. Three emission lines are observed that are blue-shifted by 1.1, 2.8 and 3.6 meV relative to the 761 meV zero phonon line of the M center, where the 2.8 meV line exhibits negative thermal quenching and is therefore proposed to originate from a second excited state of the M-center. The remaining blue-shifted emission lines, together with two additional red-shifted (4.7 and 6.4 meV) emission lines display normal thermal quenching, and are unaffected by an isotope shift of the implanted carbon atom ($^{12}$C versus $^{13}$C implantation) and implantation fluence. Thus, they likely arise either from other defects with similar emission energies or from a perturbed configuration of the M-center.
Read the full abstract
Computer Vision
Camera trap classification with deep learning under ground truth uncertainty
Supervised deep learning methods enable the rapid processing of ecological image data, but depend on a costly annotation process. Consequently, training labels are commonly derived from volunteer citizen science projects. However, disagreement among volunteers introduces uncertainty in the "ground truth" data that are assumed to be correct for model training and validation. Using two datasets containing camera trap images with associated volunteer and expert classifications, we investigated the effects of training under higher ground truth uncertainty. We observed improved overall test accuracy, particularly for images that were more difficult for volunteers. Species-level accuracy also generally improved, but generalisation to a different dataset did not. The benefits of ground truth uncertainty were enhanced by pre-training on ImageNet. Pre-training also reduced the number of training epochs required; further reductions in computational cost, but not gains in accuracy, resulted from additional pre-training on other camera trap images. With unbalanced training data, we still observed a clear benefit of increased ground truth uncertainty for overall accuracy, especially on difficult images. Class imbalance improved accuracy for common species, reduced rare species accuracy, and changed patterns of misclassification to more closely resemble mistakes made by volunteers. Our findings have implications for applying deep learning across ecological image types with multiple labels. Practitioners can improve accuracy, especially on difficult examples, by including moderate levels of label disagreement during training and using models pre-trained on general image data. In addition to improving the use of citizen science-derived labels in model training, our study suggests avenues for more effectively integrating human and deep learning classifications in combined workflows. (abridged)
Read the full abstract
Numerical Analysis
Model reduction of port-Hamiltonian systems via neural networks
In this paper, we consider structure-preserving model reduction of port-Hamiltonian (pH) systems which extend classical Hamiltonian systems with dissipation and an input-output port. These pH systems are often used in multi-physics systems, as the interconnection of one or more \pH systems results again in a pH system. If particularly the system matrices associated with the interconnection and/or dissipation of a pH system are state-dependent, then the evaluation of standard reduced-order models (ROMs) may depend on the dimension of the original full-order model, resulting in high computational costs. To circumvent these high costs, we propose to use structure-preserving neural networks. In particular, we perform two steps: (1) we use the generalized manifold Galerkin projection to project the pH system onto the reduced space; then (2) we train a neural network to learn the map from the reduced-order state to the reduced-order interconnection and dissipation system matrices. To ensure that the resulting ROM is again a pH system, the architecture of the neural network is chosen such that the skew-symmetry and positive semi-definiteness of the reduced-order systems matrices are maintained. In a numerical example, we consider a nonlinear mass-spring-damper system with state-dependent system matrices. The numerical results show that the proposed method achieves a significant computational speed-up compared to the original \ROM with comparable accuracy.
Read the full abstract
Differential Geometry
Ancient cylindrical flows with fast or slow convergence
In a recent breakthrough, Bamler-Lai proved the mean-convex neighborhood conjecture in all dimensions by showing that any nontrivial ancient asymptotically cylindrical mean curvature flow is – up to splitting Euclidean factors – either a translating bowl, or an ancient oval, or a translating oval-bowl. In this paper, we provide a short alternative argument for two of the three scenarios. Specifically, using more elementary/traditional methods, we show that if the convergence to the round cylinder is fast then the solution is a bowl times a Euclidean factor, and if the convergence is slow then the solution is an ancient oval. Moreover, the present paper also yields a new proof of the mean-convex neighborhood conjecture for neck-singularities that substantially simplifies and streamlines the approach from our prior work joint with Hershkovits (Acta '22) and Hershkovits-White (Inventiones '22).
Read the full abstract
Quantum Physics
Photon-efficient quantum repeater chains via hyperentanglement-assisted purification
Linear quantum repeater chains based on Werner-state purification (the BBPSSW protocol) and entanglement swapping (the BDCZ scheme) are fundamental to entanglement distribution in quantum networks. However, they operate under a stringent operation reliability threshold and rely on resource-intensive recurrence purification rounds, each consuming two entangled pairs to probabilistically produce one. In the literature, hyperentanglement has been proposed to exploit multiple degrees of freedom (DOF), such as polarisation and spatial modes, to encode independent entangled states within a single photon pair. This has led to the definition of DAEPP (DOF-Assisted Entanglement Purification Protocol), which we propose to integrate with the BDCZ scheme, resulting in a chain protocol that we call DAHR (DOF-Assisted Hyperentanglement Repeater). The DAEPP step distils the fidelity of a DOF by consuming other(s). In this work, we propose and analyse a DAHR variant which integrates DAEPP at every segment of an end-to-end path combined with BDCZ. We derive a closed-form end-to-end fidelity recursion that embeds single-segment DAEPP into the BDCZ scheme and give a strict resource lower bound for any BDCZ baseline utilising BBPSSW purification to match DAHR's per-segment effective fidelity. At a representative asymmetric operating point informed by prior experiment, we show numerically that matching DAHR's single-photon-pair performance requires two to three rounds of BBPSSW purification. Additionally, below a critical operation reliability, no amount of BBPSSW rounds matches DAHR's one DAEPP round performance.
Read the full abstract
Artificial Intelligence
SkillZip Pro: Execution-Aware Dynamic Compression of Progressively Loaded Skills for Self-Evolving Agents
Production agent skills are directory bundles, not isolated prompts. The root is loaded at activation; references, schemas, scripts, assets, and nested subskills are loaded only when an execution path needs them. Compressing only the root misses most deployment cost and may move branch-specific details into the always-loaded context. Flattening instead destroys progressive-loading boundaries. We introduce \method, an evaluation-free compressor for complete, progressively loaded skill bundles. It leaves the agent harness unchanged and emits an ordinary directory. The method combines two safeguards. First, it compresses across files, removing content from a reference or subskill when the root or a declared environment contract already provides it. Second, it preserves routing, so every required file and directly callable entry remains reachable after rewriting. Users can configure \method along two independent axes. One-Shot mode rebuilds the full bundle; Continual mode reuses state and applies Zip-on-Write after each evolution patch. Persistent compression rewrites the shipped bundle to reduce storage and runtime context. Transient compression keeps that bundle byte-identical and builds a task-specific view, reducing only per-run context after build cost. Entry contracts mark private, public, and conditional resources; a multi-entry audit preserves standalone public subskills. On a production content-moderation skill evaluated by our industrial multi-round harness, \method removes \hl{38%} of skill bundle tokens and \hl{10.4%} of end-to-end per-run tokens with no quality loss, while an unprotected 71% configuration loses up to 26 accuracy points to one-sided false positives. On a multi-entry bundle, \method effeciently reduces token cost while near-perfectly preserving every route and public entry.
Read the full abstract
General Relativity and Quantum Cosmology
Gravitational-Wave Echoes from Layered Compact Objects: A Double-Shell Model
Layering is a ubiquitous feature of astrophysical objects. Motivated by the fact that physical black holes retain the layered structure of their progenitor stars when viewed in the time concepts synchronizable with the clock of an outside fixed-position probe, we investigate linear perturbations and gravitational-wave (GW) echoes from a compact object composed of two concentric thin shells. Compared with the single-shell case, the double-shell structure introduces an extra barrier in the effective potential and partitions the wave propagation space into four coupled effective cavities. As the mass ratio of the inner shell increases, new spectral peaks enter from the high-frequency side of the echo spectrum; the second and later peaks shift toward higher frequencies; and the lowest-frequency peak first shifts toward lower frequencies and then returns to its $q=0$ position. We call this variation pattern spectral-peak queueing (SQ). Its existence suggests that GW echoes can be used as probes for the internal structure of compact objects under consideration.
Read the full abstract
Quantum Physics
Neutral atom quantum computing
Neutral atom qubits are one of the leading approaches for implementation of a large scale quantum computer. The original proposals for neutral atom qubits were formulated more than 25 years ago, with first demonstrations of a universal gate set following 10 years later. In the last few years the performance and scale of neutral atom qubit arrays has developed at a rapid pace leading to demonstrations of quantum algorithms, and logical qubits for fault tolerant error correction. This contribution reviews the physics of the neutral atom approach, surveys current capabilities, and provides an outlook for future progress.
Read the full abstract
Computer Vision
PixelIR: Fidelity-Perception Decoupling via Pixel-Space Image-Residual Flow Matching for Efficient One-Step Real-World Super-Resolution
Real-world image super-resolution (Real-ISR) aims to preserve structures supported by the degraded observation while reconstructing perceptually realistic details. However, existing Real-ISR methods largely optimize fidelity and perceptual quality within a shared network, causing the two objectives to interfere throughout training and making their balance difficult to control. Recent one-step methods reduce sampling steps, yet often inherit both this coupled optimization behavior and the expensive high-resolution backbone of their multi-step predecessors. We argue that efficient Real-ISR requires not only a shorter sampling trajectory, but also specialized modeling of faithful reconstruction and perceptual detail synthesis. Based on this insight, we propose PixelIR, a fidelity-perception decoupling framework built upon pixel-space image-residual flow matching. PixelIR first learns an image flow that maps the degraded observation to a faithful reconstruction. Then, a residual flow synthesizes the missing perceptual details from noise without repeatedly relearning or overwriting the complete restoration solution. We further distill the teacher into a deployment-oriented one-step student within a coarse-to-fine pyramid architecture. Extensive experiments show that PixelIR achieves leading PSNR, SSIM, and LPIPS on both RealSR and DRealSR. The final model completes pixel-space restoration in a single evaluation with only 32.9M parameters, 89.7G MACs, and 8.5ms latency, demonstrating a strong practical fidelity-perception-efficiency balance.
Read the full abstract
Instrumentation and Methods for Astrophysics
Radio measurements of air showers with the IceCube-Gen2 surface prototype station at the Pierre Auger Observatory
The design of the IceCube-Gen2 observatory, proposed as a next-generation extension of IceCube, includes a surface array consisting of scintillators and radio antennas. In addition to several such detectors already deployed at IceCube's surface array at the South Pole, a complete prototype station including three SKALA antennas has been operating at the Pierre Auger Observatory for several years. This setup has been used to successfully observe radio signals from air showers, demonstrated through coincident observations with the Auger Surface Detector. In this contribution, we present an updated analysis of these radio signals, and compare them to CoREAS simulations using the reconstruction from the Auger Surface Detector as input.
Read the full abstract
Representation Theory
Kashiwara–Nakashima tableaux, Gelfand–Tsetlin patterns, and quantum symmetric pairs
Kashiwara–Nakashima tableaux and Gelfand–Tsetlin patterns of orthogonal type are famous as combinatorial models of the finite-dimensional irreducible representations of the special orthogonal Lie algebras $\mathfrak{so}_N$. The former is constructed based on the representation theory of quantum groups, especially the theory of crystals, while the latter based on the branching rule for $(\mathfrak{so}_N,\mathfrak{so}_{N-1})$. In the present paper, we construct a natural bijection between them by means of the representation theory of quantum symmetric pairs corresponding to $(\mathfrak{so}_N,\mathfrak{so}_{N-1})$.
Read the full abstract
Differential Geometry
Some uniform estimates for Kähler–Ricci Shrinkers
In this paper, we establish several uniform estimates for Kähler–Ricci shrinkers without imposing any curvature assumptions. In particular, we prove: 1. a uniform lower bound for the entropy; 2. a uniform lower bound for the asymptotic volume ratio of Kähler–shrinkers with maximal volume growth; 3. a uniform lower bound for the scalar curvature on balls centered at a minimum point of the soliton potential for non-Gaussian shrinkers.
Read the full abstract
Machine Learning
Reciprocity Separates Gradient Flow from Rotation in Conservative Physical Learning
Physical learning lets a trainable material or network use its own physical response to carry error signals, reducing the need for a separately programmed backward computation. We ask what determines whether such a system follows conventional gradient descent or evolves along a genuinely different learning trajectory. Our canonical model is a directed layered transport network in which every node redistributes a fixed amount of flow, so learning preserves positivity and total mass. In this model, conservation constrains only the allowable learning directions. Within the matched response class studied here, adjoint matching gives the physical output response a symmetric form. Non-negative mode-wise feedback then produces a reciprocal closed-loop response and a reweighted gradient flow. Adding an antisymmetric boundary component makes the closed-loop response rotational: the learning path can turn while the error driving that update still decreases at that moment. Turning is not automatically beneficial. Its finite-step effect is set by local curvature, and its accumulated effect also depends on step selection and on the new states visited along the path. Numerical consistency checks reproduce the exact response structure, predict the sign of the local effect across new network families, and show how trajectory drift can negate a local advantage. These results separate the roles of conservation, reciprocity, and nonreciprocity in physical learning.
Read the full abstract
Analysis of PDEs
Inhomogeneous NLS in negative Sobolev spaces via Fourier-Lebesgue and modulation spaces
The purpose of this work is to provide a broader framework for analyzing the 1D mass-subcritical inhomogeneous nonlinear Schrödinger equation (INLS) \[iu_t + u_{xx} \pm |x|^{-b}|u|^{\alpha-1}u=0, \quad 1<\alpha< 5-2b\; \text{and}\; 0< b\leq 1/4.\] in Sobolev spaces of negative order, reaching almost critical Sobolev spaces by working in Fourier-Lebesgue and modulation spaces. More precisely, we establish low-regularity well-posedness in the Fourier-Lebesgue $\widehat{L}^{p}$ spaces for $4/3<p<8$. Primarily, we prove global well-posedness for arbitrarily large initial data via the data decomposition method adapted for the Fourier-Lebesgue spaces, with generalized Strichartz estimates for $p<2$ and classical Strichartz estimates for $p>2$. Furthermore, we obtain analogous results for (INLS) in modulation spaces $M^{p,p'}$ for $4/3<p<2$.
Read the full abstract
Audio and Speech Processing
Likelihood-Constrained Acoustic Reranking for Training-Free Hallucination Mitigation in LLM-Based ASR
Large language model (LLM)-based automatic speech recognition (ASR) systems achieve strong performance on conventional speech data by leveraging powerful linguistic priors and multilingual capabilities. However, under challenging conditions, these priors can override acoustic evidence, resulting in unintended translation, instruction execution, repetition, or catastrophic deletion. We propose Likelihood-Constrained Acoustic Reranking (LCAR), a training-free decoding method that improves acoustic grounding while preserving support from the base model. At each decoding step, LCAR first retains tokens whose base-model likelihood falls within a margin of the greedy token, then reranks them using an acoustic compatibility score computed from attention-pooled audio embeddings and the existing LM head. By restricting acoustic intervention to plausible, model-supported alternatives, LCAR requires no additional training, external detector, reference transcript, or auxiliary model at inference. We evaluate LCAR on four LLM-based ASR systems using human-audited TTS and open-source speech challenge suites. At $\delta=0.60$, LCAR removes 38.8–57.1% of detector-identified hallucination failures while largely maintaining WER/CER on standard open-source test sets.
Read the full abstract
Quantum Physics
Quantum Block Turbo Codes
In the early nineties, the introduction of turbo codes revolutionized classical error correction. The idea was mainly applied on two types of codes: convolutional turbo codes and turbo product codes. The first type of codes was adapted to quantum error correction which initiated the theory of quantum serial turbo codes. In this paper, we present a theory for quantum block turbo codes, the quantum analog of the second type of turbo codes. We describe their iterative decoding algorithm and simulate their performances on a depolarizing channel for different constituent codes.
Read the full abstract
High Energy Physics (Theory)
Anomalies and discrete torsion in Type II and Type I worldsheet theories
We provide a comprehensive study of anomalies and discrete torsion in the worldsheet theories of Type II and Type I string theories on a general ten-dimensional target space. In the modern understanding, the anomaly of a worldsheet theory is characterized by an invertible field theory in three dimensions, and invertible field theories are classified by the Anderson dual of a suitable bordism theory. To study the anomaly cancellation condition, we employ the explicit model of the Anderson dual constructed by Yamashita and Yonekura. We find that, for both Type II and Type I string theories, the anomaly of the worldsheet theory is cancelled if and only if the target space is orientable and admits a spin structure. We also determine the discrete torsion and clarify its physical meaning. It includes the distinctions between Type IIA and Type IIB string theories and between ordinary Type I and Sugimoto string theories.
Read the full abstract
Robotics
Learning to infer and manipulate through distributed whole-arm interaction in a soft robot
In animals such as elephants and octopuses, acquiring non-visual information about an object and physically engaging with it are inseparable processes mediated by rich, large-area interactions between compliant appendages and the environment. Soft robots provide a natural platform for translating this principle into engineered systems. Yet current robotic intelligence makes limited use of physical interaction, treating it primarily as a disturbance to be rejected or, at best, as a means of compensating for object misalignment. Here, we introduce a physical intelligence framework in which distributed compliant interactions jointly reveal task-relevant information and organize manipulation behavior. This results in an intrinsically partially observable problem: key task-relevant information is never measured directly, but must instead be inferred from the history of physical interactions. We propose a reinforcement-learning architecture that addresses this challenge by learning a memory-based control policy end-to-end. The key innovations making this possible are (i) a pretrained exploration policy that provides a reference for broad workspace exploration, (ii) joint optimization that integrates exploration and grasping objectives within a single recurrent policy, and (iii) a two-stage sim-to-real adaptation including observation mapping and policy fine-tuning. We demonstrate this principle through blind whole-arm grasping with a hybrid rigid-soft robotic arm that we equip with IMUs embedded directly within its compliant structure, providing its only source of proprioceptive sensing. The learned policy successfully identifies and grasps various objects by autonomously coordinating workspace exploration, object encounter and localization, inference of grasp-relevant properties, and stable whole-arm wrapping.
Read the full abstract
Medical Physics
Clustering based magnetic assays for SARS-CoV-2 detection with scFv-functionalized magnetic nanoparticles
Magnetic nanoparticles (MNP) can be used in magnetic immunoassays (MIA) by functionalizing them with antibodies. In homogeneous assays that follow a "mix-and-measure" approach, it is then sufficient to add the sample in question and evaluate the response of the MNPs using a magnetic measurement method. Magnetic Particle Spectroscopy (MPS) is a fast and sensitive method for this purpose, capable of determining the binding state by measuring changes in Brownian relaxation. For further analysis, alternating current susceptometry (ACS) is considered, although it is significantly more time-consuming and therefore no option for applications interested in point of care detection. In this study, we aim to further improve the approach of MIA evaluated by MPS. To this end, we use self-synthesized scFv fragments instead of whole IgG antibodies during functionalization in order to keep the size of the BNF-Dextran MNP with 80 nm nominal diameter as small as possible, which allows for greater relative size changes upon binding to an analyte. Virus-like particles (VLP) and the N protein of SARS-CoV-2, which were also produced in-house, are used as analytes. Due to multiple binding sites, these form cluster structures, which, in the case of VLP, were investigated in greater depth ACS to assess their field dependence. In addition, the sensitivity of the assays was analyzed as a function of MNP concentration, and the detection limit for both analytes was estimated. We found that, using scFv-functionalized MNPs, we were able to detect low concentrations of just 490 fM of SARS-CoV-2 VLP and 1.7 nM of the N protein. However, the slightly above-proportional increase in sensitivity as the MNP concentration decreases does not automatically imply a better detection limit.
Read the full abstract
Quantum Physics
PauLie: Fast Classification of Pauli Dynamical Lie Algebras
The dynamical Lie algebra (DLA) governs the controllability, expressibility, and simulation complexity of a quantum system. Explicitly computing it has been a major computational bottleneck: brute-force Lie closure scales exponentially in the number of qubits $n$. Many applications, however, consult only the isomorphism type of the DLA. We introduce PauLie, an open-source framework that decides this isomorphism type for DLAs generated by arbitrary Pauli strings, building on the anticommutation-graph reduction of Aguilar et al. PauLie runs in $O(n|\mathcal{G}|\max(n,|\mathcal{G}|))$ time, where $|\mathcal{G}|$ is the number of generators, turning DLA classification into a routine preprocessing step. We demonstrate its use as a structural oracle for routing Lie-algebraic simulation and Cartan decomposition, diagnosing barren plateaus in variational quantum algorithms, and engineering universal Pauli string generator sets with optimal generation rate.
Read the full abstract
Numerical Analysis
Integral Invariants of Vasudeva Murthy's Relaxation Systems: Analysis and Numerical Validation
Vasudeva Murthy's relaxation approach [A.S. Vasudeva Murthy, J. Comput. Appl. Math., 203(2), pp. 437-443, 2007], originally proposed for the Jin-Xin relaxation model, provides an alternative formulation with invariant properties that is consistent and retains the semilinear structure incomparison to the standard one. In this work, Vasudeva Murthy's relaxation approach for various relaxation systems are proposed such as the shallow water equations, the Broadwell model, the Euler equations with heat transfer and two-dimensional Jin-Xin model. For proposed relaxation models, the associated integral invariants are rigorously established at the theoretical level. The main advantage of the integral invariant is that it provides a conserved quantity for the relaxation system by incorporating the coupled contributions of the solution variables in vector form. To validate the analytical results, numerical simulations are carried out for each model using three second-order numerical schemes: CS-EBT2, a semi-implicit second-order central finite-volume scheme for hyperbolic systems with relaxation source terms [S. Sahu, E. Macca, and R. Samala, J. Comput. Phys., 563, 115100, 2026]; UCS2, a finite-volume central relaxation-type scheme [S. F. Liotta, V. Romano, and G. Russo, SIAM J. Numer. Anal., 38(4), 1337-1356, 2000]; and IMEX-RK2, a second-order Implicit-Explicit Runge-Kutta scheme [Pareschi and Russo, J. Sci. Comput., 25, 129-155, 2005]. Numerical results, compared to exact or finely resolved reference solutions, confirm that the models preserve integral invariants, remain stable under CFL restrictions, and exhibit robust and accurate behavior across all benchmark systems tested.
Read the full abstract
Machine Learning
TrainSDC: Characterizing and Mitigating Silent Data Corruption in Large Language Model Training
LLM training is increasingly vulnerable to silent data corruption (SDC), yet existing protection methods largely treat Transformer computations uniformly because their vulnerability remains poorly understood. We present the first systematic characterization of SDC vulnerability across major computation interfaces in both the forward and backward passes of Transformer training. Our analysis reveals two distinct error propagation mechanisms: forward-pass vulnerability is highly location dependent, with faults on the Q/K path producing persistent training deviations, whereas backward-pass vulnerability is largely governed by gradient exponent distributions rather than computation locations. Motivated by these observations, we propose TrainSDC, a characterization-guided protection framework consisting of Q/K-path recomputation, residual-gain monitoring, and exponent-aware gradient scaling. Experiments on Llama 3.2-1B and Qwen3-0.6B show that TrainSDC maintains training behavior close to fault-free execution under both sparse and dense fault injection while introducing only 1.65%-6.76% runtime overhead.
Read the full abstract
Computer Vision
CORAL: A Benchmark for Structure-aware and Brain-wide Neuron Reconstruction in Light Microscopy
Automatic neuron reconstruction from light microscopy images is a central problem in computational neuroanatomy. While recent methods have achieved encouraging results on local image blocks, it remains unclear whether such progress translates to reconstruction that is both structurally accurate and scalable to the whole-brain scale. We present CORAL, the first benchmark for structure-aware evaluation of automatic neuron reconstruction from light microscopy images at both local and whole-brain scales. Built on a high-quality whole-brain fMOST dataset with carefully curated annotations, CORAL establishes two progressive tasks: block-level reconstruction, which evaluates reconstruction methods under limited spatial context, and brain-wide reconstruction, which assesses complete neuron reconstruction at the whole-brain scale. To account for topological correctness beyond geometric distance similarity, we introduce a structure-aware metric based on fiber prediction. To further achieve complete neuron reconstruction across the entire brain, we develop a brain-wide neuron tracing framework that extends arbitrary local reconstruction methods to the whole-brain scale through an iterative local-to-global process. Using this benchmark, we provide the first structure-aware comparison of mainstream methods for local neuron reconstruction and further evaluate their performance in brain-wide reconstruction. Our results underscore the importance of structure-aware evaluation and the need for more robust methods for complete neuron reconstruction.
Read the full abstract
Group Theory
Fundamental groups of asymptotic cones of Lie groups with the SOL obstruction
We study asymptotic cones of Lie groups, presenting a link between the geometry of the weights and the highly non-simply-connected nature of asymptotic cones. We show that a Lie group that admits a group of SOL-type as a quotient is such that the fundamental group of its asymptotic cone contains the fundamental group of the Hawaiian earring space. We also obtain a strong non-vanishing result for the first homology group of the asymptotic cone. The assumption on the Lie group, introduced by Abels and called the "SOL obstruction", is equivalent to an explicit geometric condition on the weights. Our work builds upon the results of Burillo, who proves the statement on asymptotic cones in the case of SOL, and on those of Cornulier and Tessera who show that the SOL obstruction implies exponential growth of the Dehn function.
Read the full abstract
Cosmology and Nongalactic Astrophysics
A consistent, physical, and analytical model for CMB observables of reionisation I. Derivation, constraining power, and detectability of the auto-spectra
The Epoch of Reionisation imprints its history and morphology on the Cosmic Microwave Background temperature and polarisation anisotropies through two effects: The kinetic Sunyaev Zel'dovich (kSZ) effect and Thomson scattering. To perform joint analyses of these observables and put tight constraints on the reionisation history and morphology, a common reionisation model is necessary. We present a quick, physically-accurate analytical approach to derive consistently the angular power spectra of the three main resulting imprints that are the spatial fluctuations of the Thomson optical depth, the patchy kSZ effect, and the scattering and screening B-modes. The approach differs from existing analytical models, as it is calibrated on high-resolution hydrodynamical simulations and depends on physical parameters of reionisation. Our predicted power spectra are consistent with the ones derived from cosmological simulations, whilst including more physical information – such as the typical size of ionised bubbles and the duration of reionisation. We provide an updated detectability assessment and show that both $C_\ell^{\tau\tau}$ and $C_\ell^{BB}$ could be measured by experiments similar to CMB Stage 4 and CMB-HD, respectively. We show the complementarity of the three observables in constraining reionisation history and morphology, and illustrate the potential of their joint analysis to get a global picture of reionisation. This analytical model will be a powerful tool in the analysis of upcoming CMB data, either alone or in combination with independent datasets such as measurements of the high-redshift 21cm signal. The code to derive these spectra is publicly available online as the preion Python package.
Read the full abstract
Machine Learning
T3S: Improving Multi-Task Reinforcement Learning with Task-Specific Feature Selector and Scheduler
Multi-task reinforcement learning (MTRL) is a technique to train multiple tasks simultaneously, where previous works usually train a single model to solve different tasks by sharing parameters across various tasks. However, these methods are faced with inter-task interference since what parameters should be shared across tasks is not addressed, dramatically reducing learning efficiency. To solve these problems, we propose a novel MTRL framework called Task-Specific feature Selector and Scheduler (T3S), which consists of two components: a feature selector and a task scheduler. Specifically, the feature selectors employ hypernetworks to construct task-specific soft masks, which can be applied by globally shared representation to construct task-specific features. The task scheduler selects tasks for learning through two metrics, where the selection probability is inversely proportional to task progress (e.g., success rate) and task learning speed. Experimental results show that T3S consistently outperforms the state-of-the-art MTRL algorithms on various robotics manipulation tasks.
Read the full abstract
Combinatorics
Cycle-Structure Generating Functions for Special Breakpoint Graphs
Breakpoint graphs originate in comparative genomics, where their alternating cycles encode relationships between genomes. We study a constrained class of three-colored breakpoint graphs associated with permutations and develop cycle-refined generating functions for two extremal families. These families have a natural topological interpretation: their canonical surfaces are, respectively, the sphere and the projective plane. The spherical family is characterized by noncrossing configurations, while the projective-plane family admits a different decomposition involving a distinguished family of Möbius ladders. The resulting generating-function equations retain the full cycle structure but nevertheless admit substantial reductions. This leads to explicit Catalan-weighted evaluations, polynomiality results for refined cycle statistics, and a connection between a natural diagonal specialization and noncrossing trees. The two topological families exhibit markedly different combinatorial mechanisms, providing complementary examples of how local transformations of breakpoint graphs can control refined permutation enumerations. As a further application, the same Catalan-weighted sums arise in asymptotic unitary-Weingarten expansions for entanglement of random Gaussian states in linear optics. The combinatorial results determine the leading and constant-order moment polynomials entering the Rényi entropy expansion, with the projective-plane contribution giving the finite-size constant correction.
Read the full abstract
Analysis of PDEs
Ultracontractivity of heat semigroups with non-local Robin boundary conditions via two-sided bounds for a positive eigenfunction
We study heat semigroups on bounded Lipschitz domains $\Omega \subset \mathbb{R}^{d}$ with dimension $d>2$ under non-local Robin boundary conditions. The boundary operator $B \in \mathcal{L}( \mathrm{L}^{2}(\partial\Omega))$ is allowed to destroy the positivity of the semigroup. We assume that there exists a positive operator $C \in \mathcal{L}(\mathrm{L}^{2}(\partial\Omega))$ such that $$ |Bu| \leq C|u| \quad\text{for every }u\in \mathrm{L}^{2}(\partial\Omega), \qquad C{\bf 1}\in \mathrm{L}^{\infty}(\partial\Omega). $$ Under this assumption, we prove that the semigroup generated by the corresponding uniformly elliptic operator is ultracontractive. More precisely, its norm from $\mathrm{L}^{2}(\Omega)$ to $\mathrm{L}^{\infty}(\Omega)$ has short-time order $t^{-d/4}$. The main step is the construction of a positive eigenfunction $\phi$ of a dominating elliptic operator such that $$ 0 \ < \ \delta \ \leq \ \phi \ \leq \ M $$ almost everywhere in $\Omega$. The upper bound is obtained by power truncations and an iteration of Sobolev exponents. The lower bound follows by comparison with the Neumann semigroup. The two-sided estimate gives an $\mathrm{L}^{\infty}$-bound for the comparison semigroup. Nash's inequality and duality then yield ultracontractivity.
Read the full abstract
Computer Vision
ChessQueries: Toward Better Chess Board Recognition
Chess board recognition is the task of mapping the image of a chess board to the information of which piece is on which square. So far this task has two established benchmarks: ChessCog is synthetic, and ChessReD comes from smartphone pictures of a single chess board setup. We introduce ChessQueries, a new method combining a ViT encoder with a DETR-style decoder, which outperforms existing methods. On the ChessReD benchmark, we improve the state of the art from 15.3% to 99.2%, and demonstrate strong capabilities on out-of-distribution datasets. Our method saturates the task on the two datasets, with an average 0.01 wrong squares per board (vs. SotA: 3.4 / 0.15 respectively). We also share a new, harder public dataset, parsed from broadcasted top-level chess tournaments. Code, model weights and the SLCC data have been released.