Peer Revenue Play credits. Real papers.

Papers

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

All of arXiv

Wednesday 9 September

Computation and Language

Two-Token Features and Small-Large Ensembles for VLM Hallucination Detection

We present our system for the SHROOM-Visions 2026 shared task on character-level VLM hallucination detection. A small ($4$B-parameter) VLM is fine-tuned as a per-token classifier reading a two-token feature from its own hidden states, and is ensembled with a $\sim$400B zero-shot VLM judge at prediction time. Both components see off-the-shelf OCR of any visible in-image text. We use synthetic hallucination data generated by the large model as a source of ensemble diversity, and use validation to select feature layer, training data and OCR grounding. Our official entry reaches mean Cor $0.487$ / Cor-lbl $0.387$ on the hidden test set, placing $6$th/$28$ (EN), $6$th/$21$ (FR), $8$th/$21$ (IT) and $7$th/$22$ (ZH) on the task's primary Cor-lbl metric.
Read the full abstract

Robotics

RoboFolDeX: A Physical-World Benchmark for Long-Horizon Robotic Manipulation of Deformable Objects

Embodied AI, including vision-language-action and world-action models, must operate reliably in the physical world. Yet methods that perform well in simulation can degrade substantially on real robots, especially in long-horizon deformable-object manipulation, where policies must track changing states and execute reliable multi-stage bimanual interactions. Existing real-robot benchmarks mainly focus on short-horizon rigid-object tasks and offer limited coverage of long-horizon deformable manipulation. We introduce RoboFolDeX, a physical-world benchmark built entirely from real-robot data, with garment folding as its primary task. Since real-robot data collection is costly, RoboFolDeX studies how heterogeneous physical experience can be reused efficiently. The benchmark is organized around four research axes: leveraging human intervention and recovery data collected during deployment; transferring data across tasks, including across garment categories and from rigid to deformable-object manipulation; reusing data across scenes with changes in lighting, background, and layout; and transferring data across robotic embodiments. RoboFolDeX provides 2,000+ hours of real-robot data spanning 20+ tasks and 10+ embodiments. We also establish a fair real-robot evaluation platform for externally submitted policies, with standardized tasks, held-out physical objects, controlled initializations, and a unified execution protocol. The platform is publicly accessible at https://ai.midea.com/#/fold-challenge. We hope RoboFolDeX will serve as a unified testbed for heterogeneous real-robot data reuse and reliable long-horizon deformable manipulation.
Read the full abstract

All authors Chenhuan Liu, Yi Xu, Feng Wu, Hanyang Wang, Wenxiao Kuai, Weihao Ding, Shan Wang, Yang Liu, Shuyong Gao, Wenqiang Zhang

Dynamical Systems

Platonic constellations of periodic motions in the $(n + 1)$-body problem

We study the spatial $(n+1)$-body problem formed by one heavy central mass together with $n$ equal masses placed on a single orbit of a polyhedral rotation group $H\in\{T,O,I\}$, so that $n=|H|\in\{12,24,60\}$. Imposing the symmetry $q_{L}=L\,q_{I}$ for $L\in H$ reduces the problem to a single $2\pi$-periodic reference curve, with reduced action $A_{H}=A_{0}+\varepsilon A_{1}$, in which $\varepsilon$ is the inverse central mass and $A_{0}$ is the Kepler action. At $\varepsilon=0$ the critical set contains, as one connected component, the five-dimensional manifold of Kepler ellipses of minimal period $2\pi$, which we prove to be a nondegenerate critical manifold. A Lyapunov–Schmidt reduction along this manifold turns the continuation problem into the search for nondegenerate critical points of an explicit function $\Phi(e,\psi)$ of the eccentricity $e$ and the spatial orientation $\psi$, a nondegeneracy we verify by a computer-assisted proof. We thereby obtain, for each of the three groups, families of periodic solutions of the $(n+1)$-body problem with $n+1\in\{13,25,61\}$ bodies, bifurcating from Kepler ellipses and carrying the full tetrahedral, octahedral, or icosahedral symmetry.
Read the full abstract

Signal Processing

Spatial sparse sampling-based iterative optimization framework for GNSS Direct Position Estimation

Direct position estimation (DPE), a promising technique in Global Navigation Satellite Systems (GNSS) receivers, enables estimation of position, velocity, and time (PVT) solutions directly from correlator outputs. The conventional grid search (GS)-based DPE is computationally intensive, as it relies solely on locating the peak of the cross ambiguity function (CAF), and it does not fully leverage the PVT information present in the correlation values. This paper proposes an iterative optimization DPE framework that capitalizes on spatial coherence and spatial gradient via spatial sparse sampling (SS). In SS-DPE framework, the correlation outputs of spatial sampled PVT points all serve as measurements, where the spatial gradient and spatial coherence are derived to capture the information density and diversity of different correlation values. The analytical Cramér-Rao Bound (CRB) is derived and indicates that both spatial coherence and gradient determine the theoretical performance limit via the noise covariance and Jacobian matrices analysis. The proposed theoretical framework not only validates the feasibility of sparse sampling but also guides weight optimization to distinct correlation values, effectively integrating these insights with a general gradient-based optimization. Theoretical derivations are validated via Monte Carlo simulations. Field experiments further demonstrate the practical feasibility of the proposed SS-DPE optimization framework. Comparative analysis shows that the proposed SS-DPE achieves comparable PVT estimation accuracy to the conventional GS-DPE while consumes only sparsely sampled correlation values, improving the information utilization efficiency and reducing the computation load.
Read the full abstract

All authors Wei Gao, Rong Yang, Jihong Huang, Xingqun Zhan, Yonggang Zhang

Numerical Analysis

Order elevation of directly self-starting sub-step implicit integrators for transient dynamics

Directly self-starting implicit methods are attractive for transient analysis because they avoid auxiliary starting procedures while retaining the original first- or second-order governing equations. However, most existing formulations usually fix the last sub-step at the end of each time interval, which restricts the attainable order. This study develops a generalized $s$-sub-step implicit framework by releasing this constraint and treating all sub-step locations as design variables. The resulting methods admit a unified Runge–Kutta representation for both first- and second-order transient systems and preserve identical effective matrices over all sub-steps. Accuracy conditions are derived by simultaneously matching the numerical amplification factor and load operator, thereby accounting for both homogeneous and forced responses. For $s=1,~\cdots,~6$, two complementary families are obtained: $s$th-order members with user-controllable high-frequency numerical dissipation and adjustable sub-step locations, and $(s+1)$th-order members obtained by selecting the sub-step locations, with fixed dissipation. The latter reach up to seventh-order accuracy without increasing the number of sub-steps, although some high-order members are $A(\alpha)$-stable with stability angles extremely close to $90^\circ$. Analytical amplitude and phase errors further reveal parity-dependent superconvergence in undamped systems, and appropriate parameter selections can substantially increase either phase or amplitude accuracy beyond the formal order. Numerical benchmarks confirm the predicted convergence orders and the controllable suppression of spurious high-frequency responses.
Read the full abstract

All authors Jinze Li, Yaokun Liu, Kewei Chen, Hua Li, Kaiping Yu

Information Retrieval

LiteRAG: Cost-Efficient Graph-Based Retrieval-Augmented Generation

Graph-based retrieval can improve multi-hop question answering, but existing approaches often incur high query-time costs and produce diffuse, oversized contexts that reduce generation efficiency. We present LiteRAG, a graph-based retrieval method that replaces expensive retrieval-time LLM control with query-conditioned algorithmic exploration and reasoning-chain context construction. On DistComp, a benchmark for multi-hop retrieval over distributed-systems papers, LiteRAG attains the highest overall quality among the evaluated methods (0.798) while reducing per-query latency by over 100$\times$ and cost by over 99% relative to GraphRAG Global and DRIFT. On UltraDomain, it matches LinearRAG on overall quality while using about 14$\times$ fewer tokens. An ablation study indicates that LiteRAG's query-adaptive thresholding and community-aware hub penalization are the main drivers of its token-efficiency gains.
Read the full abstract

Differential Geometry

On a question by Firey concerning uniqueness

We prove that the curvature equation $\sum_{j=1}^n \alpha_j E_j(\tau_{\mathcal{M}})=\sum_{j=1}^n \alpha_j E_j(\tau_{\mathcal{N}})$ yields uniqueness up to translation for any two closed $C^2_+$ hypersurfaces $\mathcal{M}, \mathcal{N}\hookrightarrow\mathbb{R}^{n+1}$ whenever $(\alpha_1,\ldots,\alpha_n)\in\mathbb{R}_{\geq 0}^n\setminus\{0\}$ is log-concave and has no internal zeros. This gives an affirmative answer to a uniqueness question posed by Firey in the $C^2_+$ class.
Read the full abstract

Computation and Language

The Answer Path and the Grounding Instruction in LLM Question Answering over Knowledge Graphs

A graph retrieval-augmented generation pipeline chooses which triples to put in the prompt, a syntax to write them in, an order to write them in, and a sentence telling the model what to do with them. We vary all four over six large language models and two knowledge-graph question answering benchmarks. Two of the four choices move the answer and the other two are flat. The first is whether the answer path, the triples needed to reach the answer, is in the prompt at all. Holding the number of triples fixed and replacing every triple that is not on the chain with material from an unrelated entity changes answer accuracy by +0.003 F1, while removing the chain costs most of what the graph was worth. Retrieval budget belongs on recall, and precision in the range we can test buys nothing. There is no retriever here: subgraphs come from gold SPARQL, so precision describes the context we build, not a system setting. The second is the grounding instruction. With no facts in the prompt, telling a model to answer using only the provided facts drops F1 from 0.299 to 0.035, a factor of 8.63. That figure describes an evaluation with an empty context arm rather than a working pipeline, and an experiment that applies the instruction to its context arm but not to its no-context baseline manufactures a spurious finding that graph context hurts at depth. We found one in our own results and retract it. Syntax, triple order and subgraph size produce no effect we can measure at multi-hop depth. The comparison that would price the grounding instruction against correct context is not measurable with a format-sensitive scorer, because the instruction determines the response format; we report it as an open contrast rather than a number.
Read the full abstract

Fluid Dynamics

A stochastic modeling framework to generate 2-D rough-wall high-Reynolds-number turbulent boundary layers

Atmospheric surface layer flows are computationally challenging, predominantly due to surface roughness and the high Reynolds number, both of which demand exceptionally high spatial resolution in the near-surface region. We have developed a 2-D stochastic-based model for the generation and the streamwise concatenation of instantaneous, step-like velocity profiles featuring key elements of wall turbulent flows, i.e., uniform momentum zones (UMZ) and shear layers (Ehsani et al. 2024a,b), and vortices (Ehsani et al. 2026). The model is extended herein to the top of the logarithmic layer to reproduce the high-Reynolds-number, rough-wall, turbulent boundary layer measured by Saddoughi and Veeravalli(1994), without the support of a UMZ dataset, using only a handful of critical flow parameters: the Taylor microscale {\lambda}T, the boundary layer height {\delta}, the friction velocity u{\tau}, and the aerodynamic roughness length z0. The primary challenge lies in the integration of the stochastic model with the scaled distributions of the UMZ and vortex characteristics. The resulting statistical moments, energy spectra, and structure functions are compared against the experimental results. The validated code is made available in a GitHub repository.
Read the full abstract

Optics

Optical design for implementing non-collinear circularly polarized high harmonic generation in an enhancement cavity

We introduce an original optical cavity design intended to efficiently output-couple extreme ultraviolet (XUV) light produced via cavity-enhanced high harmonic generation (CE-HHG). It supports the amplification of oppositely circularly polarized modes, crossing with a small angle in the high finesse cavity, where circularly polarized HHG scheme (NCP-HHG) will be implemented. We present an analytical model, numerical simulations, and experimental results obtained with a low-power continuous (CW) laser, demonstrating the strong potential of the cavity as an efficient XUV output coupling method. In particular, it is well suited for producing the 7th harmonic (H7) from an Ytterbium frequency comb at 8.4 eV, which is of particular interest for precision nuclear spectroscopy of Thorium.
Read the full abstract

All authors Pranshu Dave, Thierry Auguste, Olivier Girard, Thierry Ruchon, David Bresteau

Combinatorics

On Two Conjectures Related to the Boros-Moll Sequences

The Boros-Moll sequences $\{d_i(m)\}_{0\leq i\leq m}$ are defined as $$d_i(m)=2^{-2m}\sum_{k=i}^m 2^k \binom{2m-2k}{m-k}\binom{m+k}{k}\binom{k}{i}.$$ Consider the ratio sequence $$u_i(m)=\frac{d_{i-1}(m)d_{i+1}(m)}{d_i(m)^2}.$$ Chen and Gu conjectured that $\{u_i(m)\}_{2\leq i \leq m-2}$ is both reverse ultra log-concave and log-concave. In this paper, we prove the reverse ultra log-concavity conjecture using bounds of Chen-Gu and Zhao, and prove the log-concavity conjecture asymptotically by showing that $\{u_i(m)\}_{2\leq i \leq m-2}$ is strictly log-concave for all sufficiently large $m$. The key ingredient in the latter result is a recurrence of Kauers and Paule, which we interpret as a nonlinear discrete dynamical system through a backward map. We construct an approximation to the ratio sequence using its stable limiting fixed point and combine localization and contraction arguments with finite-difference estimates and separate interior and edge analyses to obtain the desired strict log-concavity.
Read the full abstract

Computational Complexity

On the Tightness of Standard Relaxations for Mixed-Integer Bilevel Linear Programs

Exact algorithms for solving mixed-integer bilevel linear programs (MIBLPs) typically rely on sequences of lower and upper bounds that converge to the optimal value. These procedures are commonly initialized using the single-level relaxation (SLR), obtained by omitting the follower's optimality condition and solving the resulting single-level optimization problem. In this paper, we investigate whether, for broad classes of MIBLPs, the resulting standard bounds admit uniform improvements that can be computed within the same computational complexity regime. For pure continuous bilevel linear programs, we show that, unless $P = NP$, neither the SLR-based lower bound nor its associated upper bound can be uniformly improved in polynomial time, even for the class of min-max problems. We then extend this analysis to the class of pure integer min-max bilevel linear programs under the assumption that the polynomial hierarchy does not collapse. First, we show that the continuous relaxation of the SLR admits no uniform polynomial-time computable improvement. We then prove that neither the SLR itself nor its associated upper bound admits a uniform improvement by a polynomial-time algorithm with access to a mixed-integer linear programming (MILP) oracle. Importantly, this rules out uniform improvements by iterative MILP-based approaches, including cutting-plane-based and decomposition algorithms. Overall, our results demonstrate that the SLR-based bounds are, in a complexity-theoretic sense, unimprovable systematically within their natural computational regimes.
Read the full abstract

Systems and Control

Mitigating Degradation Attacks in Cooperative Autonomous Driving via Intention Sharing: A Vehicle-in-the-Loop Study

Communication delays induced by cyber attacks present a critical challenge to the safe operation of connected autonomous driving. This study investigates the use of intention sharing communication strategy to enhance the resilience of model predictive controllers under Denial-of-Service attacks. We employ a vehicle-in-the-loop testbed integrating a real drive-by-wire vehicle with a microscopic traffic simulator and vehicle-to-X communication infrastructure. We emulate Denial-of-service attacks that induce communication delays of up to two seconds. We evaluate three controller variants: baseline status-sharing, intention-sharing, and delay-aware intention-sharing control. Experimental results reveal that while baseline control suffers significant performance degradation and frequent collisions under adversarial delay, intention sharing eliminates collisions and maintains behavior near nominal levels for the tested scenarios. These findings demonstrate the practical potential of intention-sharing architectures for safeguarding connected vehicles against network-layer degradation.
Read the full abstract

All authors Prakhar Gupta, Tyler Ard, Rongyao Wang, Jagruti Sahoo, Judith Mwakalonge, Ardalan Vahidi, Yunyi Jia

Algebraic Geometry

Very general quartic sixfolds are not stably rational

We prove that a very general quartic sixfold admits no decomposition of the diagonal, and in particular is not stably rational. The proof uses specialization to an explicit quartic sixfold birational to a quadric bundle of the type considered by Colliot-Thélène and Ojanguren. We also prove stable irrationality for very general six-dimensional intersections of four quadrics.
Read the full abstract

Robotics

CougarTail & CUB: A General-Purpose Mast and Central Utility Board for Cylindrical Underwater Enclosures

Cylindrical watertight enclosures are widely used across various underwater systems, from unmanned underwater vehicles (UUVs), to remotely operated vehicles (ROVs), to various sensor platforms. However, electronics are typically built on rectangular PCBs arranged in horizontal stacks, which inefficiently occupy the circular cross-section volume that is critical for both payload capacity and buoyancy management. This paper presents CUB (Central Utility Board) and CougarTail, a general-purpose system designed to address this gap. CUB is a circular PCB sized for 4-inch-diameter enclosures that consolidates a Raspberry Pi Compute Module 5 (CM5) and an STM32 microcontroller, while also providing power management features and auxiliary connections. Mounted coaxially, CUB reduces the electronics stack of our CougUV from 200 mm of tube length and 709 g to 25 mm and 156 g, returning that length and mass budget to payload and buoyancy trim. CougarTail is an open-source companion sensor mast that houses a GPS antenna and two dual-band (2.4 and 5 GHz) omnidirectional PCB antennas. Both components are validated through bench testing and integration on a CougUV platform, our small open-sourced torpedo UUVs.
Read the full abstract

All authors Ben Washburn, Clayton Smith, Eli Gaskin, Brighton Anderson, Braden Meyers, Brady Moon, Joshua Mangelson

Optimization and Control

Circuit-Embedded Feeder-State Reconstruction for Low-Voltage Distribution Networks under Extreme Measurement Sparsity

Low-voltage (LV) distribution feeders increasingly require internal state information, yet real-time telemetry is often limited to feeder-head active and reactive power, while downstream demand and production are described only by static contract metadata. This paper proposes a physics-based AC circuit-embedding reconstruction framework for this extreme-sparsity regime. Feeder-head measurements, matched metadata, and class-level profiles first determine bus-level load and production priors. These priors parameterize auxiliary constant-power elements coupled to the physical feeder through impedance branches, and the feeder state is reconstructed by solving the resulting nonlinear AC circuit equilibrium. The resulting AC equilibrium enforces the assigned virtual-element powers, while weighted least squares (WLS) reconciles measurements and pseudo-measurements through weighted residual fitting. The method is evaluated on a real 459-bus Portuguese LV feeder model using an in-sample-calibrated synthetic benchmark of 672 fifteen-minute snapshots. The proposed method converged for all 672 snapshots and achieved voltage RMSE, MAE, and maximum absolute error of 0.0033, 0.0014, and 0.0337 p.u., respectively, outperforming both static-prior WLS baselines. Under the same constructed priors, matched-prior WLS achieved a lower RMSE of 0.0029 p.u., while the proposed circuit embedding reconstructs the feeder state by solving the nonlinear AC equilibrium, bypassing WLS residual fitting entirely and requiring no specification or tuning of pseudo-measurement variances.
Read the full abstract

Rings and Algebras

A note on generalized distributive sets of a finite Dickson nearfield

Let $\mathbb{F}_{q^n}$ be a finite field where $q=p^l$ ($p$ is a prime) and a Dickson nearfield $R$ associated to the field. In this note we investigate the basic properties of the generalized distributive set $D(a,b)$ where $a,b\in R$. The set is well-known when $n=1, 2$. For $n>2$, we characterize when $D(a,b)$ is a subfield of $\mathbb{F}_{q^n}$. Furthermore we find the unique largest subfield of $\mathbb{F}_{q^n}$ contained in $D(a,b)$. We also provide a condition to guarantee that $D(a,b)$ is not a subfield of $\mathbb{F}_{q^n}$. We end the note by posing the question of when $D(a,b)$, as a space over $\mathbb{F}_p$, has a primitive normal basis.
Read the full abstract

Statistical Mechanics

Italian Business-to-Business Invoicing Data: A Network Analysis

We present a comprehensive description of the network of Italian Business-to-Business commercial relationships, based on the universe of electronic invoices collected by the Italian Tax Office. The firm-to-firm detail of the data is exploited to describe the distribution of the numbers of buyers (customers) and sellers (suppliers) per firm, the centrality of each firm in the production network, and the average distance between firms. We characterize for the first time the firm-to-firm network describing the Italian production system, and present its geographic and sectoral breakdown. The analysis reveals heavy tailed distributions for the numbers of buyers and sellers per firm, implying a scale-free structure of the network. The distributions of centrality values display heavy tails as well, indicating a strong concentration of importance in a relatively small number of firms. We estimate the tail exponents for all these distributions, finding in all cases lower exponents for downstream (buyer-side) distributions than for the upstream (seller-side) ones.
Read the full abstract

Computation and Language

$\Phi$-Bench: Can Large Language Models Engineer the Infrastructure That Powers Them?

Large language models (LLMs) now show strong reasoning and coding abilities, raising the prospect that they could assist in developing and optimizing the very infrastructure that powers them. Existing benchmarks, however, mainly focus on isolated kernels, predefined operators, or pre-specified optimization targets, and cannot evaluate open-ended, long-horizon LLM infrastructure engineering. To address this gap, we present $\Phi$-Bench, a benchmark of 85 tasks grounded in real LLM training and inference codebases. The tasks cover nine categories of LLM infrastructure and come in three formats, ranging from kernel function completion to long-horizon implementation and repository-wide end-to-end optimization. They are built by a construction pipeline that turns real pull requests, issues, and repository code into executable tasks with validated tests, and every reward is gated by correctness and anti-hacking checks. Across eight frontier models, the strongest model, Claude Opus 5, scores only 36.53%, and every model performs worse on long-horizon, repository-level implementation than on kernel-level tasks. Trajectory analysis further suggests that stronger models treat optimization as controlled experimentation, validating ideas with cheap local runs and ruling out confounders before drawing conclusions. We will open-source $\Phi$-Bench to facilitate future research on LLM agents' ability to develop and optimize the LLM infrastructure that powers themselves.
Read the full abstract

All authors Leilei Ding, Shumin Wang, Yuting Huang, Fanqi Wan, Yinmin Zhang, Qi Han, Yiming Xu, Feiyuan Zhang, Xiaomeng Chu, Guoliang You, Wuyang Zhang, Daxin Jiang, Yanyong Zhang

Machine Learning

Hierarchical and Permutation-Invariant Feature Transformation Learning via Policy-Guided Embedding Search

Feature transformation improves predictive performance on tabular data by constructing informative abstractions from raw features. Recent generative approaches encode transformation knowledge into continuous embedding spaces for efficient exploration of candidate strategies, but face three key limitations: (1) overlooking hierarchical relationships between low-level features, operations, and high-level abstractions; (2) enforcing order-sensitive embeddings on inherently permutation-invariant transformation sequences, thereby introducing systematic bias; and (3) relying on gradient-based search, which is ill-suited to non-convex transformation spaces. We propose a framework with two complementary components. First, a permutation-invariant hierarchical module captures interactions across features, operations, and abstraction levels, with a self-attention pooling mechanism that maps semantically equivalent structures to consistent embeddings aligned with downstream performance. Second, a policy-guided multi-objective reinforcement learning strategy initializes the search from empirically strong seeds and jointly optimizes predictive accuracy and transformation efficiency. Extensive experiments on diverse tabular benchmarks demonstrate the effectiveness and robustness of our framework against strong baselines. Our code and data are publicly available at: https://github.com/RayLiu1103/PHER.
Read the full abstract

All authors Rui Liu, Tao Zhe, Yanyong Huang, Sankha Narayan Guria, Xiao Luo, Wei Fan, Yanjie Fu, Dongjie Wang

Computer Vision

UOT-Gap: A Variational Principle for the Modality Gap in Vision-Language Models via Unbalanced Optimal Transport

Vision-language models such as CLIP embed images and text in a shared space, where modality-specific distributions often remain separated. Existing accounts connect this modality gap to initialization, contrastive dynamics, and information imbalance, while its distributional and pairwise contributions to retrieval remain unresolved. We introduce UOT-Gap, a training-free variational diagnostic that models frozen image and text embeddings with unbalanced entropic optimal transport (UOT). The UOT optimum separates transport, coupling complexity, and marginal mass variation; a complementary pair-aware residual compares observed image-caption pairs with the UOT soft matching. On Flickr8K and COCO-1K with frozen CLIP, OpenCLIP, and SigLIP encoders, caption degradation reduces Flickr8K Recall@1 from 0.559 to 0.003. Across six dataset-model conditions, the pair-aware residual tracks retrieval degradation with mean absolute Spearman 0.973, compared with 0.392 for the mean gap. The association remains stable across five random COCO-1K subsets at $0.954\pm0.026$, with a minimum of 0.943. UOT barycentric updates reduce the transport objective while degrading retrieval, distinguishing geometric objective descent from task improvement. These results establish UOT-Gap as a diagnostic for caption quality, modality alignment, and retrieval robustness.
Read the full abstract

All authors Zonglin Yang, Huilan Ma, Xudan Zheng, Yuejun Xie

Analysis of PDEs

Convergence Rate toward Rarefaction Wave under Periodic Perturbation for Generalized Korteweg-de Vries-Burgers equation

In this paper, a rarefaction wave under space-periodic perturbation for the generalized Korteweg-de Vries-Burgers equation is considered. It is shown that if the initial perturbation around the rarefaction wave is suitably small, then the solution of the system tends to the rarefaction wave. The stability of solution under periodic perturbation is an interesting and important problem since the perturbation keeps oscillating at the far fields. The key of proof is to construct a suitable ansatz carrying the same oscillation as the solution. Then we can find cancellations between solutions and ansatz such that the perturbation belongs to some Sobolev space. The nonlinear stability can be obtained by the energy method.
Read the full abstract

Analysis of PDEs

Identifying unknown time delay and spatially varying coefficients in a reaction-diffusion equation from boundary measurements

This work investigates an inverse problem for a general class of linear reaction-diffusion systems incorporating multiple delayed contributions, namely retarded diffusion, retarded time derivatives, and retarded source terms. The objective is to simultaneously recover the unknown time lag $\tau>0$ and spatially heterogeneous coefficients from boundary flux measurements alone. The recovery strategy exploits the singular temporal behavior generated by an incompatibility between the prescribed initial history and the boundary data. In contrast to prior inverse problems for delay equations, which assume $\tau$ known and are confined to ODE or abstract settings, our approach operates in a parabolic PDE framework with spatially varying coefficients. Once $\tau$ is identified, we establish a Lipschitz stability estimate via a Carleman inequality, in the case of time-independent coefficients $p=p(x)$, $q=q(x)$. This is the first result to simultaneously recover an unknown delay and spatially dependent coefficients in a delayed parabolic PDE from boundary data.
Read the full abstract

Artificial Intelligence

Why Sample What You Can Enumerate? Exact Policy Optimization for Genomic Tool Selection

Reinforcement learning over a frozen reasoner has become a common recipe for teaching a policy which external tools to invoke. We show that this recipe becomes structurally mismatched in specialist scientific settings where the complete tool-subset space is enumerable. There, a small set of recurring computational capabilities covers the domain, so the space of tool subsets is combinatorial yet small enough to enumerate, and GRPO still estimates an action expectation from a handful of sampled rollouts. Worse, the approximation degrades as training succeeds: as the policy concentrates on preferred subsets it resamples them, sampled rewards collide, and the group-normalized advantage vanishes. On genomic reasoning the fraction of questions yielding no reward signal rises from 0.2% under a uniform reference policy to 20.8% after GRPO training. As a remedy, we introduce FGPO (Full-Group Policy Optimization), which (1) scores every tool subset and optimizes the exact action expectation, so each update sees the complete action space, and (2) precomputes the reward of each question–subset pair into an exhaustive table, removing frozen-reasoner calls from the training loop entirely. Across five frozen reasoners and three genomic benchmarks, FGPO outperforms GRPO in all 15 settings by 6.75 points on average and up to 14.20, while a standard on-demand GRPO schedule would require 2.4 times as many frozen-reasoner reward evaluations and, on GenomeQA, FGPO cuts invoked tools per question from 2.36 to 1.40.
Read the full abstract

All authors Haoyue Liu, Xiaoyu Ma, Ye Chen, Zhichao Wang, Xiaoying Tang

General Relativity and Quantum Cosmology

Testing Source Dependence in Gravitational-Wave Tests of General Relativity with Mutual Information

Phenomenological deviation parameters used in gravitational-wave tests of general relativity need not represent universal fundamental properties of the sources. Instead, a deviation parameter can be a derived quantity that depends on source properties. Apparent violations of general relativity arising from waveform systematics or data-quality issues can also depend on source properties. In addition to diagnosing individual events, it is helpful to identify systematic trends at the population level. We introduce a nonparametric framework that uses mutual information to test whether the deviations depend on source properties. We validate the framework using simulations with massive graviton waveforms. Assuming we do not have prior knowledge of the underlying theory, massive graviton effects can be captured by a parameterized post-Einstein coefficient that is expected to depend on the luminosity distance and source mass. The results show that the mass and distance dependencies become clear for graviton masses of $\geq 1\times10^{-22}\,\mathrm{eV}$ at detector sensitivities representative of the current fourth observing run, while the controlled general relativity simulations yield null results. We then analyze public real-data products for tests of general relativity from GWTC-4.0 and 4-OGC, including phenomenological parameterized tests, modified dispersion tests, and parity violation tests. The phenomenological deviation parameters have tentative correlations with effective spin, and correlations of modified-dispersion or parity-violation parameters with source mass are visible. These results demonstrate that catalog-level dependence can identify patterns and thus be used to diagnose the origin of deviations in testing general relativity. The framework therefore complements single-event analyses and parameterized hierarchical population tests.
Read the full abstract

Complex Variables

Meromorphic solutions of first-order differential equations with rational exponential coefficients

We study first-order differential equations $f'=R(e^z,f)$, where $R\in\C(t,w)$. We prove that a meromorphic solution on the whole complex plane is algebraic over $\C(e^z)$ unless $R$ is a polynomial of degree at most two in its second variable. Such an algebraic solution necessarily has the form $S(e^{z/q})$, with $S$ rational and $q$ a positive integer, giving an affirmative answer to Question~6.2 in Gundersen's collection. The geometric input is Guillot's theorem on single-valued trajectories of meromorphic vector fields on surfaces. The additional argument compares the fibration provided by that theorem with the original exponential coordinate. A ramification calculation forces every finite nonzero branch value $a$ to satisfy $Da=a$, which excludes such a value and reduces the comparison to a two-point cover. Divisor divisibility and relative algebraic closedness then recover a Riccati equation over the original coefficient field. No growth hypothesis is imposed. We also identify the algebraic degree with the least integer deck period and describe the growth and value fibres of the resulting solutions.
Read the full abstract

Operator Algebras

Every $\mathrm{AW}^*$-Algebra is Normal

Using set-theoretic methods, we prove that every $\mathrm{AW}^*$-algebra is normal, resolving a question of Wright that has stood open for 46 years. This was previously known in the case of $\mathrm{AW}^*$-factors, by work of Saitô and Wright. We also show that if there is a model $V$ of $\mathrm{ZFC}$ with an $\mathrm{AW}^*$-algebra that fails to be monotone complete, then in some forcing extension of $V$ there is an $\mathrm{AW}^*$-factor that fails to be monotone complete, implying that any $\mathrm{ZFC}$ proof that all $\mathrm{AW}^*$-factors are monotone complete yields a $\mathrm{ZFC}$ proof that all $\mathrm{AW}^*$-algebras are monotone complete. Both results build on transfer principles for Boolean-valued factor representations originally developed by Ozawa. This work was assisted by the Danus LLM orchestration system.
Read the full abstract

Representation Theory

Finiteness and growth of brick chain filtrations

We study Ringel's brick chain filtrations over finite-dimensional algebras: filtrations whose factors are filtered by copies of individual bricks, ordered so that morphisms from earlier bricks to later ones vanish. First, using submodule varieties, we prove that the largest submodule of a fixed module belonging to a torsion class takes only finitely many values as the class varies, answering Pavón's question. We deduce finiteness of brick chain filtrations and the bound $2^{d^2}$ for modules of dimension $d$. For $\tau$-tilting finite algebras, we bound their number by the multinomial coefficient determined by simple composition multiplicities. Finally, we construct families of bricks over the three-arrow Kronecker algebra whose filtration counts grow exponentially in the square of composition length. We prove this growth using a Littlewood–Richardson formula for submodule counts and the hook-length formula. In particular, the counts eventually exceed the factorial of composition length. These results answer Ringel's two questions.
Read the full abstract

Quantum Physics

Information capacity of quantum statistics: Fock-state tests of a discrete binary-sequence model on cloud photonic quantum processors

Our central premise is that quantum mechanics may be the statistical limit of a more fundamental discrete theory: any such theory equips a physical system with a finite information capacity, and its departure from quantum statistics is controlled by how much of that capacity the system uses. We show that commercial cloud photonic quantum processors have reached the precision required to bound this capacity from below, using the binary-sequence model of Powers et al. as the concrete test theory: outcome probabilities arise from counting discrete sequences of length $n$, quantum mechanics is recovered as $n \to \infty$, and $n$ measures the information capacity of the register behind a prepared state. Photon Fock states $|1\rangle$, $|1,1\rangle$, heralded $|2\rangle$, and cascaded beam-splitter pairs are measured on programmable interferometers with dominant systematics determined in situ. The model's composition-consistent parametrization, singled out by requiring that rotations compose, recovers quantum mechanics with deviations $1.24/n$; a random-effects likelihood analysis calibrated by parametric bootstrap excludes all $n \le 100$: the information capacity of the register carrying the two-photon state, if finite, exceeds $10^2$. Cascaded beam splitters test the composition law directly: the data are split-invariant, excluding naive count composition at $8\sigma$ and confirming the interference-sign rule. Model-independently, curve-averaged deviations from the quantum partition law larger than $2.3\times10^{-2}$ are excluded at 95% CL, and the originally published linear parametrization is excluded outright. Because the compilation offset is frozen per circuit it is calibratable, opening the $10^{-3}$ floor ($n \sim 10^3$) to current hardware: cloud photonic processors are quantitative instruments for quantum foundations, and information capacity an experimentally boundable quantity.
Read the full abstract

All authors Chiran Wijesundara, Octavia T. Volpe, Dejan Stojkovic, Herbert Fotso, Tim Thomay

Robotics

Adaptive Shared Control with Online Bounded-Rational Human Behavior Estimation

This work considers adaptive shared human-robot control for nonlinear control-affine systems, where the assumption of a fully rational human is relaxed and the robot adapts its assistance to observed boundedly rational human behavior. We use a level-k bounded-rationality model of the two-player game to construct a finite bank of candidate human and robot policies through alternating best-response computations, with the associated value functions and policies approximated using adaptive dynamic programming. During the shared-control interaction, state-transition residuals compare the measured system evolution with the trajectories predicted by the candidate human policies. The residuals are accumulated using a forgetting factor and mapped to a probabilistic human-behavior model over the finite candidate bank. Rather than selecting a single candidate or averaging stored robot policies, the robot computes a distribution-aware one-step best response by minimizing an expected cooperative cost over the complete estimated human behavior distribution. For a quadratic terminal-value approximation and Euler state propagation, this response admits a closed-form solution expressed in terms of the expected human input. The proposed methods are evaluated in simulations of a benchmark nonlinear system stabilization task, and of a planar manipulator shared control setup. The reported results show decreasing Kullback-Leibler divergence between the estimated and simulated human behavior distributions, and a lower accumulated running cost for the robot agent over the shared control interaction period, than the maximum-probability and probability-weighted alternative policies baseline.
Read the full abstract