Peer Revenue Now accepting positions

Papers

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

All of arXiv

Monday 9 November

Algebraic Geometry

Generalizations of Igusa surfaces

We give a criterion for the contracted product $E \times^G X$ to have a singular Picard scheme, where $E$ is an elliptic curve, $G \subset E$ is a finite subgroup scheme, and $X$ is a projective $G$-normal curve, in the sense of Brion. We show that every smooth projective surface with a singular Picard scheme and a positive-dimensional automorphism group scheme arises in this way. We also illustrate the criterion by constructing two new families of surfaces with singular Picard schemes, arising from $G$-normal curves $X$ for $G = \mathbb{Z}/p\mathbb{Z}$ and $G = \boldsymbol{\alpha}_p$, respectively. As a special case of $G = \mathbb{Z}/2\mathbb{Z}$, we recover Igusa's classical example. Finally, we compute all infinitesimal group schemes which appear as $\operatorname{Pic}^0_S/(\operatorname{Pic}^0_S)_{\mathrm{red}}$ for all (quasi)-bielliptic surfaces $S$. This completes the description of the singular Picard schemes of smooth projective surfaces of Kodaira dimension zero.
Read the full abstract

Artificial Intelligence

Learning to Act with World Models by Valuing What Comes Next

A sparse terminal reward tells an agent little about its steps. Much of the missing information is implicit in the environment's dynamics, because each state transition can change the conditions for success long before the reward arrives. A world model trained on real transitions can learn to predict what each action changes, but a predicted observation does not by itself show whether success has become more likely. With a binary terminal reward, PPO's critic makes this judgment by estimating the current policy's probability of success, but only for states the agent reaches. Our method, WM-PPO, elegantly brings the two together. For each action the policy samples, executed or not, the world model predicts the next observation and the critic values the predicted successor. In the policy channel, comparisons among these values give the policy auxiliary advantages; in the value channel, their mean gives the critic an auxiliary target. A shared gate weights both channels according to how reliably comparisons with similar margins predicted real outcomes earlier in training. The policy, critic and world model all learn from the agent's own interaction, with no teacher or privileged information, and only the policy is used at inference. With a comparable number of real environment steps, WM-PPO raises success from PPO's 63.5% to 87.5% on ALFWorld and from 39.3% to 52.9% on ScienceWorld. It matches the strongest PPO refinements we tested on ALFWorld and exceeds the best of them by 6.5 points on ScienceWorld. The trained policy more often completes prerequisite steps before placing an object or committing to an answer.
Read the full abstract

All authors Chenghao Gao, Zhi Zhang, Dian Jin, Yihe Pan, Doudou Zhou

Combinatorics

Hilton's Conjecture on Uncomplemented Incomparable Families

A family ${\mathcal A}\subseteq 2^{[n]}$ is uncomplemented if it contains no set together with its complement. Families ${\mathcal A}_1,\ldots,{\mathcal A}_t$ are strongly incomparable if every set in one family is incomparable with every set in another family. In 1976, Hilton conjectured that if ${\mathcal A}_1,\ldots,{\mathcal A}_t\subseteq 2^{[n]}$ are uncomplemented and strongly incomparable, then $$\sum_{i=1}^t |{\mathcal A}_i|\le 2^{n-1}.$$ We confirm this conjecture by a matrix rank argument that encodes incomparability and uncomplementedness in the same matrix, which is quite different from the earlier methods on this problem. We also study weak incomparability, in which the same set may occur in different families. Goldwasser, Hilton and Zheng conjectured an upper bound in this setting and claimed that their conjecture is equivalent to Hilton's conjecture. We give counterexamples to their conjecture for every $t\ge3$ and identify the error in the claimed equivalence. We then propose a replacement conjecture and confirm it when $t$ is large. The upper bound $2^{n-1}$ in Hilton's conjecture can be attained if some families are allowed to by empty. It is natural to consider the maximum size of $\sum_{i=1}^t |{\mathcal A}_i|$ for uncomplemented and strongly incomparable families ${\mathcal A}_1,\ldots,{\mathcal A}_t\subseteq 2^{[n]}$ when all families are required to be non-empty. Denote this maximum number by $M(n,t)$. We show that \[ 2^{n-1}-1\le M(n,t)\le 2^{n-1} \] for $n\ge3$ and $3\le t\le n$.
Read the full abstract

Data Structures and Algorithms

Instance-Optimality of Single-Pair Personalized PageRank

We consider discounted random walks in a directed simple graph $G=(V,E)$. Such a walk terminates at each step with a constant probability $\alpha\in (0,1)$ and otherwise continues to a uniform random out-neighbor. Given a pair $(s,t)\in V^2$, we denote by $\pi(s,t)$ the probability of such a discounted random walk starting at $s$ terminates at $t$. This probability is referred to as the Personalized PageRank (PPR) of $t$ with respect to $s$. The problem of estimating $\pi(s,t)$ within a constant multiplicative factor with constant success probability for a given vertex pair $(s,t)$ is called the single-pair PPR estimation problem. We investigate the instance-optimality of single-pair PPR estimation in directed graphs. We present a new algorithm that aims at instance-optimal running time up to polylogarithmic factors. In other words, no correct algorithm can run faster than our algorithm on any valid graph by more than a polylogarithmic factor. We show that our algorithm is instance-optimal in the standard adjacency-list query model if $n$ is not known (as is often the case for most of the large networks to which PPR is applied). When $n$ is known, our algorithm is still instance-optimal for sparse graphs. For denser graphs, when $n$ is known, we show that no algorithm can achieve general instance-optimality. For the multiplicative approximation of PPR, we also improve the worst-case complexity from $O(mn\log n)$ to the near-optimal $O(m\log n)$. This is not only for the single-pair discussed above, but also for the single-source and single-target variants of the PPR estimation problem.
Read the full abstract

All authors Christian Bertram, Mads Vestergaard Jensen, Mikkel Thorup, Hanzhi Wang, Shuyi Yan

Computation and Language

TrajLift: Lifting Memory Recall Trajectories with Latency-Modulated Heat Diffusion

Verbal memory retrieval carries rich diagnostic information for Alzheimer's disease through the order, hesitation, and categorical organization of recalled words, yet computational modeling of this retrieval process remains largely unexplored. A key challenge is that timing and hierarchical structure are diagnostically entangled: the diagnostic meaning of retrieval timing depends on its hierarchical context. We propose TrajLift, a spectral framework that models verbal recall as timed trajectories on a hierarchical graph. Our central mechanism is latency-modulated heat diffusion on the semantic hierarchy, a unified operator that lifts each recalled word into a representation of both multi-scale hierarchical structure and continuous retrieval timing. Formal analysis shows that the resulting representation is first-order separable into structural and temporal components with spectral selectivity across hierarchical levels. On synthetic benchmarks requiring structure-time interaction, TrajLift stays near-perfect on deep hierarchies where all baselines fail. On real-world clinical data, we show that TrajLift outperforms diverse baselines by jointly modeling timing and semantic structure, with gains that grow as retrieval timing becomes more informative.
Read the full abstract

All authors Jiawen Kang, Jinchao Li, Mingyu Cui, Junan Li, Xixin Wu, Helen Meng

Computation and Language

The Deception Archive as a repository of standardized datasets for large-scale verbal deception detection research

Computational advances in natural language processing and machine learning have substantially expanded the use of automated approaches in verbal deception research. However, existing verbal deception datasets remain fragmented across disciplines, as they are often difficult to identify, access, and reuse, limiting cross-study comparisons and the development of robust computational models. We present the Deception Archive, a curated and standardized repository of publicly available verbal deception datasets built upon a large-scale systematic review. Eligible datasets were retrieved, manually inspected, and standardized following a structured data curation workflow. The current repository comprises 41 publicly available datasets containing 387,995 truthful or deceptive statements. The archive represents a heterogeneous repository (e.g., in data size, topic of investigation, ground-truth, etc.) and a living resource for integrating newly available deception datasets over time. By providing centralized access to standardized datasets and accompanying metadata, the Deception Archive facilitates dataset discovery, cross-dataset validation, and reproducible evaluation of computational models.
Read the full abstract

All authors Caterina Borgese, Riccardo Loconte, Lucca Pfründer, Bennett Kleinberg

Mesoscale and Nanoscale Physics

Lattice Commensurability and Majorana sectors of Pfaffian–Anti-Pfaffian interfaces

We study Pfaffian and anti-Pfaffian non-Abelian fractional quantum anomalous Hall states in half-filled moiré lattice systems. Owing to crystal-momentum conservation modulo a reciprocal lattice vector, these lattice systems admit coupled-wire constructions of the two phases without requiring spatially varying magnetic-field-induced momentum shifts. For the resulting Pfaffian–anti-Pfaffian interface, a Umklapp process can gap the counter-propagating charge sector, while the remaining four co-propagating Majorana modes naturally group into two sectors organized by translation symmetries. Our results show that lattice commensurability qualitatively changes the coupled-wire accessibility of non-Abelian states while providing a microscopic momentum structure that directly constrains their edge and interface dynamics.
Read the full abstract

Differential Geometry

Negative normal degree and holomorphic sectional curvature pinching

We obtain an upper bound for the local holomorphic sectional curvature pinching of a Kähler manifold containing a smooth rational curve of negative total normal degree. As an application, we prove that a compact Kähler surface with positive holomorphic sectional curvature and local pinching constant at least $1/9$ is biholomorphic to $\mathbb{CP}^2$, $\mathbb{CP}^1\times\mathbb{CP}^1$, or the first Hirzebruch surface $\mathbb F_1$. For pinching strictly greater than $1/9$, only the first two possibilities occur, confirming a conjecture of Yang and Zheng. We also determine the optimal local pinching constants on $\mathbb F_1$ and $\mathbb F_2$.
Read the full abstract

Statistics Applications

Amortized Transport-Map Likelihoods for Hierarchical Inverse Problems with Application to Wetting

Hierarchical Bayesian models pool measurements, but with the forward model in the likelihood every posterior evaluation costs one forward solve per cell. To reduce this cost, we introduce a simulation-based inference framework in which a triangular transport-map likelihood, trained once per design, serves every cell. We apply it to infer wetting properties from photographs of a stable hole in a thin liquid film. The hole's radius depends, through the nonlinear Young-Laplace equation, on the contact angle and capillary length. One photograph gives rank-one information about them, so we pool many in a hierarchical model whose cells, each a substrate at one volume setting, share the apparatus geometry and the noise scale. A design fixes the number of substrates, the volume settings and the photographs per cell. The map's posteriors agree with direct sampling of the forward model in the wetting properties at a synthetic validation design, and less closely at a larger synthetic scaling design. At the validation design, simulation-based calibration detects no departure for any parameter. There, the map needs several hundred times fewer forward-model evaluations than direct sampling with a gradient-free ensemble sampler for a single posterior, and about five orders of magnitude fewer over a calibration study of 200 posteriors. On real photographs its posterior agrees with direct sampling and is consistent with the measured radii. Trained offline, the amortized map turns new photographs into a posterior with no forward solve. It is thus a major step toward measuring wetting properties in situ without disturbing the liquid.
Read the full abstract

Systems and Control

DC Voltage Control Dominated Oscillations in Electromechanical Frequency Range - Insights from Synchronous Coefficient Reduction Effect of IBR Plants and Network Interactions

Large-scale integration of inverter-based resources (IBRs) is widely known to cause subsynchronous oscillations. This paper investigates a distinct class of oscillations that occur in the electromechanical frequency range (EFR), referred to as EFOs, and are dominated by the DC voltage control (DVC) loops of grid-following IBRs rather than synchronous generator-like controls or phase-locked loops (PLLs). Although open-loop oscillation modes of DVC loops are subsynchronous, the corresponding closed-loop dominant oscillation mode related to DVC can shift downward into the EFR without retuning the control parameters. This frequency shift is caused by network-mediated interactions among the IBRs, which introduce a negative synchronous coefficient into the DVC oscillation loop. This synchronous coefficient reduction effect (SCRE) becomes stronger as the IBR plant scale increases and/or as the minimum eigenvalue of the network admittance matrix decreases. Furthermore, a complex vector-based stability analysis method is proposed, and a stability criterion for DVC-EFOs is derived. Finally, the analytical conclusions are validated using a practical power system configuration implemented in Simulink. The SCRE differs from traditional explanations, in which the open-loop oscillation modes are already located in the EFR, and provides new insight into DVC-dominated oscillations in the EFR for integration of IBR plants.
Read the full abstract

Quantum Algebra

The multiplicity bound for near-group fusion categories of type $\mathbb Z/4\mathbb Z+k$

Let $\mathcal{C}$ be a near-group fusion category of type $\mathbb Z/4\mathbb Z+k$. We show $k\in\{0,3,4\}$, which then gives a complete characterization of near-group fusion categories determined by $\mathbb Z/4\mathbb Z$.

Quantum Physics

High Rate Quantum Block Turbo Codes

Quantum Block Turbo Code is an iterative decoding technique for concatenated quantum error correction codes in product form. In this paper, we study new high rate quantum block turbo codes with different constituent codes. We examine the impact of the change of codes parameters and the order of constituent codes on turbo performance.
Read the full abstract

All authors Khaled Jebari, Luiz Anet Neto, Ramesh Pyndiah, Jean-Louis de Bougrenet de la Tocnaye

Computational Engineering

Computational Structure of Pairwise Housing Connectedness Networks

Directed housing-submarket networks combine common market movement with heterogeneous bilateral relations, and different node-level measures need not yield distinct empirical rankings. We construct a positive pairwise net-connectedness network from the generalized forecast error variance decomposition (GFEVD) of a vector autoregression (VAR) and use it to compare complementary node positions. Net directional connectedness (NET), PageRank, and Viral Centrality summarize net directional balance, recursively weighted incoming position, and normalized multistep reach. When bilateral net differences are nonzero, the network is a weighted tournament. A least-squares edge-flow decomposition separates the pairwise-net matrix into a NET-equivalent gradient component and a NET-orthogonal residual, while a monotone-allocation condition establishes exact PageRank-indegree correspondence in a transitive tournament. In Seoul's 25-district network, the gradient component accounts for 70.9% of the squared Frobenius norm, leaving 29.1% in the residual component. PageRank is strongly aligned with indegree (Spearman $\rho=0.987$), and Viral Centrality with outdegree ($\rho=0.935$); the PageRank alignment exceeds the range of independently weighted simulations with comparable tournament consistency and edge-weight variation. Three common-component treatments reduce total connectedness from 81.98% to 50.60-52.00%, a 36.6-38.3% reduction relative to the raw system, and reverse 97-98 of 300 pairwise net directions. Thus, both the magnitude and the apparent direction of connectedness depend materially on the treatment of citywide common movement. Songpa's positive NET position remains prominent across common-component treatments and dynamic specifications and ranks first in both chronological subsamples.
Read the full abstract

All authors Seo In Jang, Seung Ho Choi, Ji-Eun Shin, Hyunwoo Lee

Numerical Analysis

Involution-Preserving schemes through constrained mesh motion: a novel Arbitrary-Lagrangian-Eulerian framework

We introduce a novel structure-preserving finite volume method that exploits mesh motion to enforce differential involutions arising in hyperbolic systems of conservation laws. The key idea is to leverage the geometric degrees of freedom of an Arbitrary-Lagrangian-Eulerian (ALE) framework to construct a constrained mesh motion that is intrinsically compatible with the involution constraints embedded in the governing equations. The method is developed within a cell-centered formulation on moving quadrilateral meshes, where discrete divergence and curl operators are constructed from vertex-based differential operators. Particular care is devoted to the consistent evolution of the mesh geometry and the satisfaction of the Geometric Conservation Law (GCL). The proposed framework is applied to two representative hyperbolic systems featuring distinct involution constraints: the ideal magnetohydrodynamics equations, endowed with the divergence-free condition on the magnetic field, and the hyperbolic Euler-Heat equations, subject to a curl-free constraint on the thermal impulse. Second-order accuracy in time is achieved through a Runge-Kutta discretization, with the geometric constraints consistently enforced at each stage. A comprehensive set of classical benchmark problems is presented to assess the accuracy and effectiveness of the method. The numerical results demonstrate that appropriately constrained mesh motion provides an effective mechanism for preserving differential involutions, establishing a novel connection between mesh geometry and the structure-preserving discretization of hyperbolic systems.
Read the full abstract

Machine Learning

Harnessing Vision-Language Models for Federated Unsupervised Visual Recognition

Federated learning (FL) is a promising research paradigm that enables collaborative model training across distributed clients without data sharing. To handle unlabeled data in local clients, existing unsupervised FL research has primarily focused on tasks like clustering and self-supervised representation learning, leaving the critical problem of unsupervised visual recognition largely unexplored. To bridge this gap, we introduce FedUR, the first framework designed for unsupervised recognition by leveraging the zero-shot capabilities of pre-trained vision-language models like CLIP. Specifically, FedUR addresses data heterogeneity through a pseudo-labeling mechanism with a class-aware data selection strategy. This strategy first estimates a global pseudo-label distribution, then allocates pseudo-labeled data to clients accordingly. To handle the uncertainty of conventional positive pseudo-labels, FedUR further incorporates class weights and a negative pseudo-labeling strategy. Furthermore, FedUR proposes a personalized prompt aggregation protocol where general visual prompts are aggregated on the server, and personalized text prompts encoding local knowledge are retained on clients. Extensive experiments demonstrate that FedUR consistently outperforms baseline methods under both label-level and feature-level distribution shifts. We believe this work paves the way for further advances in federated unsupervised recognition. Code will be available at {github.com/tim-learn/FedUR/}}.
Read the full abstract

All authors Jian Liang, Kuangpu Guo, Ran He, Tieniu Tan

Machine Learning

NearFormer: Linear-Size Sparse Attention via Randomized Near Neighbors

Graph transformers can capture long-range interactions, but their quadratic attention cost limits their use on large graphs. We introduce NearFormer, a graph transformer whose global attention is restricted to a randomized near-neighbor (NN) graph. NearFormer uses an efficient randomized construction to connect each node to a small number of near neighbors from a broader neighborhood defined by a label-independent feature or spectral geometry. For a fixed retained budget \(k\), the resulting attention support has linear size \(\mathcal{O}(|E|+nk)\). The NN graph is built once by within-bin search and then held fixed. We use these edges as sparse global-attention supports, with optional learned attention biases, while preserving the original graph for local message passing. Experiments on long-range graph benchmarks and synthetic structural tasks show that NearFormer provides a favorable accuracy–scalability trade-off relative to local message passing, dense and sparse attention.
Read the full abstract

Mesoscale and Nanoscale Physics

Phonon emission due to dc-driven electron heating in a two-dimensional electron gas containing artificial scatterers

We report the observation of phonon-induced resistance oscillations (PIROs) up to order $\varepsilon_\mathrm{{ph}} = 8$ in a two-dimensional electron gas containing artificial scatterers, activated by electron heating due to an applied dc current which enables phonon emission. Under equilibrium conditions, only two weak PIRO maxima are detectable. The dc current simultaneously generates Hall field induced resistance oscillations (HIROs). A precise analysis of their amplitude, which is modulated by the double resonance with PIROs, dramatically enhances the visibility of the latter compared to the equilibrium case. We identify transverse acoustic phonons with a sound velocity of $s = 3.4$\,km/s as the relevant mode and analyze the oscillatory amplitude modulation in the subsonic regime ($\varepsilon_{\mathrm{dc}} < \varepsilon_{\mathrm{ph}}$) within the theoretical framework of Dmitriev et al. At the sound barrier ($\varepsilon_{\mathrm{dc}} = \varepsilon_{\mathrm{ph}}$), where the electron drift velocity reaches the speed of sound, we observe a phase shift in the oscillatory electron-phonon interaction that independently confirms recent experimental findings.
Read the full abstract

All authors Frederik Bartels, Mihai Cerchez, Jan Lange, Dominique Mailly, Klaus Pierz, Hans W. Schumacher, Thomas Heinzel

Computation and Language

Blind Variation, Selective Retention: Improving Creative Idea Generation in AI Scientists

Scientific discovery requires ideas that are original, useful, and non-obvious. Yet AI scientists are typically given as much task information as possible during idea generation, which can anchor generation toward familiar approaches. Inspired by the cognitive framework of Blind Variation and Selective Retention (BVSR), we propose a discovery framework that separates these two roles of information: it generates proposals under deliberately limited task information to broaden exploration, then learns from proposals previously verified on the same task and their outcomes to retain promising candidates for costly empirical evaluation. We evaluate BVSR across five proposer models and 12 scientific tasks. Limiting task information substantially increases proposal originality and non-obviousness, and produces a larger share of highly creative and high-utility proposals than existing state-of-the-art prompting-based baselines. Learning from prior proposal outcomes on the task further improves candidate selection, discarding half of the candidates while retaining most top-performing proposals. When incorporated into existing AI-scientist workflows, BVSR achieves stronger discoveries across gene expression analysis, drug response prediction, anomaly detection, scaling law discovery, and brain activity prediction. These results suggest that effective AI scientists should use information differently across stages of discovery: constrain task-specific information during exploration while exploiting accumulated empirical experience on the task.
Read the full abstract

All authors Guowei Xu, Jiaru Zou, Mete Erdogan, Haotian Ye, Pan Lu, James Zou

Instrumentation and Methods for Astrophysics

Upgrade of the MITHiC testbed for phase and amplitude control in the focal plane

High-contrast imaging for the direct detection of exoplanets is limited by stellar brightness and by wavefront errors. Coronagraphy and wavefront control reduce these effects and enable contrast levels compatible with current testbeds and future missions such as the Roman Space Telescope. Within this framework, the MITHiC testbed at the Laboratoire d'Astrophysique de Marseille is being upgraded with a new deformable mirror used to control the wavefront. It complements the existing setup, which includes a classic Lyot coronagraph and a spatial light modulator (SLM) enabling phase and amplitude control to optimize the dark-hole region. We also implement the Zernike wavefront sensor (ZELDA) to correct low-order aberrations. This contribution presents recent progress on the MITHiC upgrade and initial results, highlighting its role as a versatile platform for developing advanced high-contrast imaging techniques.
Read the full abstract

All authors Anton Neisius, Élodie Choquet, Vincent Chambouleyron, Arthur Vigan

Optimization and Control

On Relative Smoothness in Dually Flat Geometry

We study the geometric meaning of relative smoothness on dually flat manifolds and its implications for the convergence analysis of mirror descent. Our main result yields one-sided counterparts of the classical Levi–Civita characterizations of Lipschitz Riemannian gradients, each equivalent to relative smoothness. We further investigate the stationarity of mirror descent iterates for possibly nonconvex objectives and show that relative smoothness alone does not guarantee convergence of the Riemannian gradient norm to zero. We identify a one-sided condition on the cubic tensor that ensures such convergence and yields an $O(1/N)$ stationarity bound. Examples based on the negative entropy potential illustrate the role of metric variation.
Read the full abstract

Functional Analysis

Whitney-Type Real-Analytic Approximation on Banach Spaces I: The $C^0$ and $C^1$ Cases

The classical Whitney theorem on real-analytic approximation is a fundamental result in finite-dimensional analysis and geometry. Extending this theory to general infinite-dimensional Banach spaces involves longstanding open problems. This paper is the first part of our study of Whitney-type real-analytic approximation on Banach spaces, focusing on the $C^0$ and $C^1$ cases. We prove that a real Banach space admits a separating polynomial if and only if it admits a polynomial topological embedding into a real Hilbert space. We also introduce classes of mappings that locally factor through polynomial maps into finite-dimensional spaces and develop methods of real-analytic approximation for these classes. Using these tools, we establish fine $C^0$ and $C^1$ real-analytic approximation theorems for broad classes of Banach spaces, including nonseparable ones. As applications, we resolve several open problems, including those concerning Whitney-type extensions with real analyticity off closed sets, real-analytic approximation preserving Lipschitz constants, and the real-analytic equivalence of diffeomorphic open subsets of Hilbert spaces.
Read the full abstract

Information Retrieval

TS-GEAR: Taobao Shangou GEnerative Ad Retrieval

On-demand delivery platforms such as DoorDash and Uber Eats impose a constraint absent in traditional e-commerce: merchants beyond the user's delivery range are invalid candidates rather than merely low-ranked. Generative retrieval has recently shown strong potential in large-scale retrieval systems. Adapting it to this setting, however, poses three challenges. First, existing semantic IDs encode geography only as a soft signal, so reachability can only rely on a non-differentiable post-generation hard truncation. Second, user preferences vary substantially across both regions and queries, so aggregating historical behaviors without conditioning on geography and the current query yields biased interest estimates. Third, the retrieved candidate set is evaluated by its collective value, yet autoregressive training scores each sequence independently, with no set-level signal to discourage redundant near-duplicates. To address the challenges, we propose TS-GEAR, a unified generative retrieval framework for on-demand delivery: (1) Geo-Semantic ID splits merchant identifiers into geographic prefixes and semantic suffixes, making reachability a prediction target rather than a post-hoc filter; (2) Query-Geo Target Attention aggregates historical behaviors using the real-time query and location as dual targets, while encoding each behavior's own location into its attention key; (3) GMPO (Group Marginal Policy Optimization) rewrites advantages using a deduplicated group reward and marginal advantage within GRPO, avoiding auxiliary diversity objectives. Deployed online with a lightweight 0.1B backbone, TS-GEAR serves over 150 million queries per day in the search advertising pipeline of Taobao Shangou, consistently outperforming state-of-the-art generative retrieval methods on both HitRate and geographic validity, and achieving 3.8% gains in ad impressions and 4.2% in ad revenue.
Read the full abstract

All authors Guofeng Tong, Xu Zhang, Guangliang Yu, Haiqiang Wang, Guannan Cong, Xing Guo, Yang Zhou, Jingjing Duan, Shuang Li, Feng Shi, Hualei Liu

Cosmology and Nongalactic Astrophysics

Investigating the Viability of Dedicated Microhertz Gravitational Wave Pulsar Timing

The microhertz gravitational wave (GW) frequency regime remains one of the least explored regions of the GW spectrum. This is due to the physical limitations of our current and upcoming GW observatories such as LIGO and LISA, which are tuned to dHz and mHz respectively. Proposed experiments targeting the $\mu$Hz regime are ambitious, expensive and unlikely to be operational within the next 20 years. Pulsar Timing Arrays (PTAs), with their Galaxy-sized baselines, occupy the other end of the spectrum; the frequencies they are sensitive to are a function of the observing cadence and total dataset timespan. Existing PTA campaigns have cadences and timespans corresponding to the nanohertz regime, however the possibility exists that they can be 'pushed' to detect higher frequencies. This paper investigates the viability of a custom microhertz-sensitive PTA, observing a small number of millisecond pulsars with a daily cadence, over the course of 1 and 2 years respectively. We assume the use of both 100-m and 500-m class radio telescopes for these observations, and an idealised noise model with negligible systematics to determine the most optimistic detection threshold. We compute the sensitivity of our proposed high-cadence PTA configurations to power-law signals and individual GW sources. We find that a 500-m class telescope would likely be able to set new upper limits on individual sources and GW background signals in the $0.1-1~\mu$Hz range; while a 100-m class telescope would, in the best case, only match that of the performance of the EPTA second data release in the same range. Finally we consider a possible $\mu$Hz PTA consisting of SKA-Mid and DSA observing 24 pulsars; we find this performs on par with the 500-m class while using far less telescope time, presenting a more feasible solution to microhertz-GW pulsar timing.
Read the full abstract

Machine Learning

Replication of Thermal Comfort Cohort Models: No Support for Improved Thermal Preference Prediction

Cohort comfort models (CCMs) aim to overcome the extensive individual data requirements of personal comfort models (PCMs) by grouping occupants based on demographic and psychometric characteristics (cold start) or similar thermal preferences (warm start). We evaluated the validity of the reported CCM performance gains through (i) a methodological assessment of the original study, (ii) a replication using one of the two original datasets (DORN: 20 participants, 6 months) under a revised experimental protocol, and (iii) a replication using an independently collected dataset (PersonAI: 72 participants, 4 weeks). Classification performance was evaluated using F1-micro, as in the original study, and F1-macro to account for class imbalance. Under F1-micro, a trivial baseline that always outputs the majority class performed on par with cohort models. Under F1-macro, all cohort models achieved lower median scores than a simple baseline that makes random predictions based on the estimated distribution of thermal preferences. In our methodological assessment we identified the following key issues: Limited data per participant and per minority class, reliance on a metric insensitive to class imbalance, missing trivial baselines, inadequate cross validation strategy, and multiple occurrences of information leakage. Our findings challenge the original study's conclusion that CCMs improve thermal preference prediction. CCMs did not outperform simple class-distribution baselines. Future studies should improve experimental designs, including controlled thermal conditions, to obtain sufficient observations across thermal preference classes.
Read the full abstract

All authors Thomas Hirsch, Katja Ceplak, Veronika Windisch, Samina Kadkhoda Masoumali, Matias Quintana, Clayton Miller, Stefano Schiavon, Gerald Schweiger

Robotics

Embodied First Exam

Early successes on agentic robotics point to a potential paradigm shift toward using coding agents to control robots. Existing robotics benchmarks are built around visuomotor policy learning, leaving it open whether coding agents can reliably control diverse robots under physical-world constraints. We therefore introduce the Embodied First Exam, a unified, comprehensive, and agent-native evaluation framework where coding agents can choose their own control approach: direct control, code to control, or train a policy to control. We hope it serves as a first exam for coding agents before they enter the physical world, with 1,156 tasks in 35 suites on 47 robots, including 9 new suites that go beyond pick-and-place to spatial puzzles, dexterous manipulation, multi-robot interaction, dynamic whole-body control, and sim-to-sim engineering. Each task can be run in up to three evaluation settings. The Standard Setting mimics the real world: the agent receives only sensor observations and, by default, cannot reset the environment. Its Realtime variant keeps the environment running while the agent reasons. The Privileged Setting gives the agent access to the full simulator state and unlimited resets, allowing it to search for an oracle solution for training purposes. Success rates in this setting serve as an approximate upper bound. On the 456 core tasks, the strongest of eight agents, GPT-6 Astra in Codex, achieves a success rate of 34% in the Standard Setting, leaving much room for improvement. To see where agents fail, we label the stage at which failed attempts go wrong and compare failures across agents, settings, and stage categories.
Read the full abstract

All authors Kangrui Wang, Qineng Wang, Changhe Chen, Pingyue Zhang, Chi Gui, Feishi Wang, Canyu Chen, Wenli Xiao, Xiaoyu Huang, Tonghe Zhang, Linjie Li, Zihan Wang, Minyeong Kim, Yiyou Sun, Mingkai Deng, Jinyu Hou, Qi Zhu, Han Liu, Eric P. Xing, Huaxiu Yao, Zhiwen Fan, Yuejiang Liu, Dawn Song, Pieter Abbeel, Li Fei-Fei, Jiajun Wu, Yuke Zhu, Guanya Shi, Ruohan Zhang, Manling Li

Representation Theory

From Frobenius seeds to self-orthogonal modules over symmetric algebras

We construct finite-dimensional symmetric algebras admitting nonprojective self-orthogonal modules from Frobenius algebras equipped with a module satisfying explicit stable and homological conditions. The resulting modules have complexity three, and their stable endomorphism algebras are isomorphic to the algebra of dual numbers over the ground field. The construction combines a variable input with a fixed auxiliary factor carrying the required twisting, and preserves the input through specified quotient and corner constructions. The proof combines derived induction, an obstruction to perfect factorization, and a cone construction over a Gorenstein triangular matrix algebra. For a symmetric algebra and a self-orthogonal generator, we describe the derived category of the opposite endomorphism algebra through a recollement with a two-term $2$-symmetric dg algebra. For our counterexamples, it is formal and quasi-isomorphic to an explicit four-dimensional dg algebra. Applications to cyclic coverings of a quantum exterior algebra yield connected examples with explicitly determined numbers of simple modules. The construction works over every field containing an element of infinite multiplicative order.
Read the full abstract

Robotics

GeoMind: A Benchmark for Spatial Understanding in Robotic Manipulation

Robotic manipulation has advanced rapidly in mapping visual observations and language instructions to actions. However, reliable manipulation in open-world environments requires robust spatial understanding: the ability to identify targets in complex layouts (spatial reasoning) and adapt to scenes that continuously evolve through interaction (spatial memory). Yet existing embodied benchmarks fall short in evaluating spatial understanding: they often introduce strong visual cues that bypass spatial reasoning, and focus on static scenes that offer little insight into an agent's capacity for spatial memory. To address this gap, we introduce GeoMind, a systematic benchmark for structured spatial understanding through tabletop manipulation. GeoMind comprises 54 evaluated task variants organized around spatial reasoning and spatial memory. It also incorporates a scalable pipeline that automates the generation and annotation of task instances with verified targets and executable interactions. Extensive evaluations across diverse embodied agents reveal their consistent weaknesses in spatial understanding. Our findings highlight that these agents struggle to track identities after spatial changes and convert relational or geometric constraints into executable actions. We release the GeoMind benchmark at https://github.com/Notfound-JH/GEOMIND-benchmark.git.
Read the full abstract

All authors Jinghe Wang, Xinrui Cao, Duo Wu, Chenghao Gu, Yong Zhong, Linjia Kang, Tianyi Xiong, Zhi Wang

Dynamical Systems

Functions with finite energy on the Sierpinski gasket

This paper develops a systematic framework for constructing finite-energy functions on the Sierpiński gasket using Lipschitz maps and various classes of contractions. Several properties of the constructed functions are investigated, including the min–max property and bounds for their energy. Several examples of finite-energy functions, along with their graphical representations, are provided to illustrate the functions constructed using the proposed framework. It is further shown that finite energy does not necessarily imply membership in the domain of the Laplacian. The framework is then applied to a class of fractal interpolation functions on the Sierpiński gasket, for which necessary and sufficient conditions for finite energy are established under suitable contractivity assumptions.
Read the full abstract

Differential Geometry

Scalar curvature deformation of isoparametric tubes

We resolve Ge's local scalar curvature rigidity conjecture for spherical isoparametric tubes. The conjecture fails throughout its stated radius range for families with at least two distinct principal curvatures, via arbitrarily small smooth deformations supported in any regular annulus. Geodesic balls of radius less than $\frac{\pi}{2}$ are locally rigid under $C^2$ smallness. We also prescribe sufficiently small smooth nonradial scalar-curvature increments with fixed compact interior support, under antipodal symmetry for families with 2, 4, or 6 distinct principal curvatures and without symmetry for families with 3 distinct principal curvatures.
Read the full abstract

Combinatorics

A Rooted Structure Theorem for $5$-Connected Graphs and $K_6^-$-Minors

Jorgensen conjectured that every $6$-connected graph with no $K_6$-minor is apex. For a $6$-connected graph containing a $K_4$ on vertices $a_1,\dots,a_4$, being apex is equivalent to $G-a_i$ having a planar embedding with the other three roots on a common face for some $i$, so that for such graphs the conjecture becomes a rooted planarity statement. Robertson, Seymour, and Thomas showed that the failure of the corresponding three-rooted planarity is witnessed by a legless tripod. We introduce a four-root structure, called a frame, that extends this certificate, and prove a structural theorem: every $5$-connected graph containing a frame rooted at four prescribed vertices contains a subgraph of one of five explicit types, denoted $A$–$E$. This theorem does not depend on the edges among the roots. As an application, we show that when the four roots span a $K_4^-$, each of the five configurations yields a $K_6^-$-minor. Together with a planar argument for the case without a frame and an induction based on a theorem of Kawarabayashi, this gives a proof that every non-planar $5$-connected graph contains a $K_6^-$-minor. This statement also follows from a recent theorem of Lo; our proof is independent and is driven by the rooted structure theorem, whose method we expect to apply to Jorgensen's conjecture with other target configurations.
Read the full abstract