Peer Revenue Past citations are no guarantee of future returns

Papers

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

All of arXiv

Friday 9 October

Graphics

MeshRepair: Autoregressive Topology Repair for Generated Artisan Meshes

Xiyu Wang, Yufei Wang, Ruocheng Wu et al. (+3 more)

Recent artisan mesh generation methods produce meshes whose topology comes close to the preferences of human artists. But gaps to artist quality remain because generated meshes commonly contain holes, stray faces, and duplicated faces. Regenerating a mesh from scratch is costly, and a new draw may change the preferred topology entirely or introduce errors elsewhere. In this work we study artisan topology correction. The task is to correct the topology of a generated mesh (the draft) in place, toward the layout an artist would make. We propose a method for this task. It is a face-level correction network based on a prefix-LM. It reads the mesh bidirectionally and emits an edit sequence autoregressively. The sequence consists of a deletion phase followed by an addition phase. We propose the novel formulation of the deletion head which picks faces to delete autoregressively and the addition head that selects vertices to form a face to add. The addition head also adds vertices when needed. Untouched faces pass through unchanged, so the corrected mesh keeps the vertices and the layout of the draft. The draft enters as a prefix and only the edits are generated. A correction therefore takes far fewer decoding steps than generating every face again with an autoregressive mesh generator. We benchmark our method against repair tools on scripted defects and on held-out drafts from two mesh generators. Across both scripted defects and generated drafts, our method improves the quad-equivalent face F1, removes non-manifold faces, and cleans up the topology.
Read the full abstract

All authors Xiyu Wang, Yufei Wang, Ruocheng Wu, Zhihao Li, Lanqing Guo, Bihan Wen

Computation and Language

Does System One Transfer? Calibrated Typed-Decision Models, Label Budgets, and Translation for Indonesian Regional Languages

Romi Fadli, Rusdah, Wahyu Ramadani Putri Utami et al. (+1 more)

Calibrated typed-decision ("System One") models return a label and a probability in a single forward pass at a small fraction of the cost of large language models (LLMs), yet their evaluation has been almost entirely English. We report a pre-registered evaluation of the commercial Jev model and the open Laya-multilingual model on Indonesian regional languages, using the parallel NusaX sentiment (12 languages) and NusaTranslation sentiment and emotion (7 languages) test sets. Jev is more often correct on sentences that are lexically closer to Indonesian in all three datasets, and presenting the human Indonesian translation instead raises its accuracy by up to 37.2 points, most in languages distant from Indonesian. Its expected calibration error does not grow on regional languages, but its ability to rank its own errors degrades in all three datasets. Retrained classifiers need more than 500 labels per language to match zero-shot Jev in 9 of 10 NusaX regional languages. A confidence-gated cascade that sends Jev's 10-30% least confident decisions to a local IndoBERT classifier matches or exceeds the better of the two systems, and routing to an LLM instead matches that LLM at 24-77% lower API cost. Of six further commercial LLMs (three Claude and three GPT models), five are 4-14 points more accurate than Jev, at several hundred times the API cost where it is reported, while the smallest (Claude Haiku) is less accurate. We release the pre-registration, code and all predictions.
Read the full abstract

All authors Romi Fadli, Rusdah, Wahyu Ramadani Putri Utami, Zulfa Malika Amana

Combinatorics

Free products, principal sums, bipartition complete matroids and the linking operation $(\mathcal{M}_{SP} \vee \mathcal{M}_{PQ}) \times (S \uplus Q)$

H. Narayanan

In this paper we examine the special class of bipartition complete matroids, which includes free products and principal sums of matroids (Crapo and Schmitt, 2005; Bonin and Kung, 2015). A matroid $\mathcal{M}_{SQ}$ on $S \uplus Q$ is said to be $\{S,Q\}$–complete iff whenever $b^1_S \uplus b^1_Q, \, b^1_S \uplus b^2_Q,$ and $b^2_S \uplus b^1_Q$ are bases of $\mathcal{M}_{SQ},$ we also have $b^2_S \uplus b^2_Q$ as a base of $\mathcal{M}_{SQ}.$ We use the linking operation $\mathcal{M}_{SP} \leftrightarrow_{\mathcal{M}} \mathcal{M}_{PQ} :\equiv (\mathcal{M}_{SP} \vee \mathcal{M}_{PQ}) \times (S \uplus Q)$ (Theja and Narayanan, 2014) for this study. We construct the $\{S, Q\}$–augmentation of an arbitrary matroid and show that this augmentation is again a matroid, that the dual of the augmentation is the augmentation of the dual matroid and that the $\{S, Q\}$–augmentation of an $\{S, Q\}$–complete matroid leaves it unchanged. We also show that bipartition completeness is stable under minors in the appropriate bipartitioned sense. Finally, we show that free products and principal sums of matroids $\mathcal{M}_{S},\mathcal{M}_{Q}$ are $\{S, Q\}$–complete and give explicit minimal decompositions for them.
Read the full abstract

Combinatorics

New mixed cages, bounds and computational results

Claudia Marlene De la Cruz and Miguel Ángel Pizaña

An $(r,z,g)$-mixed cage is a mixed graph that is $r$-regular in edges, $z$-regular in arcs, of girth $g$ and of minimum order $n=n(r,z,g)$. We give a general lower bound for $n(r,z,g)$. In the case $g=3$, we determine the exact value $n(r,z,3)=r+2z+1$. For $g=4$ we give two improved lower bounds and we determine the exact value $n(r,z,4)=2r+4z$ for $r\geq 2z-2$. We also present computational results, obtaining mixed cages and computational lower bounds by means of exhaustive backtracking searches with symmetry reductions. Moreover, we give an upper bound for $n(1,z,g)$ that matches all our computed values.
Read the full abstract

Atomic Physics

Additional ionization during nuclear $\beta^{\pm}$ decay in atomic systems

Alexei M. Frolov

An original method is developed for numerical calculations of probabilities of the bound-free atomic excitations during the nuclear $\beta^{\pm}$ decays in actual atoms and ions. To achieve this goal we have derived and tested the closed analytical formulas for the final state probabilities of different excitations in the few-electron atomic systems. This study is the first non-approximate and essentially complete investigation of the atomic bound-free excitations during the nuclear $\beta^{\pm}$ decays in atoms.
Read the full abstract

Artificial Intelligence

Can Language Models Play the Wikipedia Game?

Alexander Hickey, Joanna Knight and James Bishop

Language models increasingly support or automate information retrieval from document collections too large to index. This task is difficult when it requires specialist linguistic or scientific knowledge. We used the Wikipedia Game to test language models on an analogous information retrieval task. In the game, players must reach a target article using only article links. We compared sentence embedding models and LLMs across five Wikipedia languages and five academic-literature datasets. We construct very large local graphs of Wikipedia pages and of paper citations using OpenAlex and provide code to build and navigate them. This allows the study of language models' navigation capabilities without relying on third-party APIs. We compared the models with human and TF-IDF benchmarks. The Wikipedia Game format enables robust comparisons across techniques, languages, and domains. No model significantly outperformed humans, although the best models (GPT-o3 mini, GPT-4o, QwQ:32B, and DeepSeek-R1:70B) achieved comparable success. Performance and rankings varied across languages and topics, with implications for practitioners utilising these techniques.
Read the full abstract

Computer Vision

Towards Physically Grounded Articulated 3D Assets: A Multi-Agent Pipeline and Benchmark

Xudong Cai, Ruijie Zhu, Haoliang Shang et al. (+5 more)

In this paper, we study physically grounded articulated 3D asset generation—moving the goal from reproducing how objects look to capturing how they move and function. Inspired by the literature on procedural and agentic generation, we present Artisan, a system that extends the latest method Articraft by introducing a multi-agent pipeline with physical simulation feedback, enabling comprehensive attributes beyond geometry, appearance, and physically plausible articulation, such as collision-free motion, stability under gravity, and accurate mass and material properties. Furthermore, we construct a unified benchmark spanning four key aspects: spatial fidelity, physical plausibility, downstream task performance, and overall quality. The benchmark encompasses 100 reference assets from ArtVIP and LightWheel, 50 stability prompts, and 50 ABO objects with reference masses, with protocols evaluating various metrics. Through extensive evaluation of more than 10 representative models, Artisan achieves competitive spatial fidelity while attaining the lowest collision rates and highest stability rates among the evaluated methods. Together, our system and benchmark establish insights for advancing 3D generation towards physically plausible, functionally usable assets.
Read the full abstract

All authors Xudong Cai, Ruijie Zhu, Haoliang Shang, Ruining Li, Marc Pollefeys, Shangzhe Wu, Andrea Vedaldi, Chuanxia Zheng

Quantum Physics

An entangling Hubbard quasiparticle collider on a digital quantum processor

Chan Bin Bark, Hanbyul Kim, Yoonil Koo et al. (+6 more)

Scattering of quasiparticles provides direct access to the interactions that govern strongly correlated systems. In this work, we realize the quasiparticle scattering experiment of the Fermi-Hubbard model on a digital quantum processor by colliding a doublon and a holon in the half-filled chain of 21 sites encoded on 42 qubits of the ibm_boston processor. For the first time, we measure the transmission probability of quasiparticles of the Hubbard model on quantum hardware, which agrees with the Bethe-ansatz scattering matrix without free parameters. Beyond the scattering probabilities, the collision generates entanglement by superposing the transmitted and reflected components, which we infer from the particle-number fluctuation. In sharp contrast, a control collision between two doublons shows neither the interaction dependence nor the number fluctuation, as expected for indistinguishable particles. Our work demonstrates that the ability to create quasiparticles of a chosen flavor on demand and program their motion and interactions makes a digital quantum processor a new experimental toolset for probing the entanglement structure generated by quasiparticle scattering.
Read the full abstract

All authors Chan Bin Bark, Hanbyul Kim, Yoonil Koo, SeungBeom Hong, Seik Pak, Kwang Jun Ahn, Sangmo Cheon, Youngseok Kim, Moon Jip Park

Robotics

Splat2Act: From Captured Scenes to Personalized VLA Manipulation

Xiaomeng Ye, Krishna Jaganathan, Gyubeum Lim et al. (+3 more)

A user who brings a generalist manipulation policy into their home or office wants it to pick up their objects from their shelves, counters, and tables. Yet even a strong generalist like $\pi_{0.5}$ achieves 0% pickup success in our captured deployment scenes without adaptation, and collecting real demonstrations in every user's environment is prohibitively expensive. We present Splat2Act, an automated real-to-sim-to-real pipeline that personalizes a pretrained vision-language-action (VLA) policy to a specific scene and object set using only a handheld phone scan and images of the user's objects. Unlike prior real-to-sim work that reconstructs individual tabletops, Splat2Act operates on building-scale indoor environments: it builds a photorealistic digital twin with 3D Gaussian Splatting and collision geometry, automatically discovers tabletop and non-tabletop receptacles, constructs feasible pickup tasks, and generates successful demonstrations with a privileged simulation oracle, with only scene and object capture required from a human. Applied to 101 scanned campus environments with a median navigable area of 104.87 m$^2$ and 7 receptacles, the pipeline yields 852 receptacles and over 68,000 pickup demonstrations. We fine-tune $\pi_{0.5}$ in two stages: Stage 1 adapts to hundreds of public object assets, drawn either from the deployment scene alone or from all 101 scenes; Stage 2 personalizes to the user's own objects on the deployment scene's receptacles. Our strongest personalized policy reaches 62.5% aggregate pickup success in the target scene. Deployed on a real Stretch 3, the same policy achieves 63.9% success on a 12-task subset, closely tracking its 66.7% simulated success on those tasks. These results suggest that automated real-to-sim-to-real training can personalize VLA manipulation policies using only simulation-oracle data.
Read the full abstract

All authors Xiaomeng Ye, Krishna Jaganathan, Gyubeum Lim, Gunjan Chhablani, Muhammad Zubair Irshad, Zsolt Kira

Computer Vision

InstantFusion: A Shared Latent Interface for Heterogeneous Image Generators

Xuanlang Dai, Jiazi Bu, Yujie Zhou et al. (+5 more)

Diffusion/flow models have emerged as the de facto paradigms for image generation, giving rise to a diverse family of modern generators with increasingly heterogeneous architectures and representations. Despite their remarkable generation capabilities, the cross-model alignment of latent representations and the consistency of their underlying denoising dynamics remain underexplored. In this work, we uncover an intriguing cross-model compatibility: independently trained generators exhibit similar latent spatial organization, while directly handing off intermediate states between unaligned models can still preserve coarse visual content. Motivated by these observations, we propose InstantFusion, a lightweight shared latent interface that connects the denoising trajectories of heterogeneous image generators. Specifically, to enable flexible cross-model state translation, model-specific Latent AutoEncoders (LAEs) are introduced to project intermediate states from different generators into a shared latent space and decode them into the latent space of a target generator. This shared space is constrained by two complementary objectives: aligning intermediate states derived from the same image at matched noise levels and synchronizing their corresponding local denoising velocities. Once trained, InstantFusion allows denoising to seamlessly transition across frozen generators without requiring a dedicated mapping for every model pair. The versatility of InstantFusion is demonstrated across acceleration, multi-preference composition, cross-architecture on-policy distillation, and cross-model composition, achieving competitive performance across these diverse settings.
Read the full abstract

All authors Xuanlang Dai, Jiazi Bu, Yujie Zhou, Yuhong Liu, Beichen Zhang, Penghui Yang, Ziyu Liu, Yuhang Zang

Robotics

Noise Space Learning for VLA Adaptation

Junjie Lu, Chuheng Zhang, Yucen Wang et al. (+4 more)

Pretrained vision-language-action (VLA) models encode broad manipulation priors, yet adapting them to specific tasks requires balancing limited robot interaction, computational resources, and safety constraints. Noise space learning (NSL) adapts flow-matching-based VLAs by learning a context-conditioned distribution over noise while keeping the pretrained VLA fixed. Building on prior noise-steering methods, we formulate NSL as an interface for VLA adaptation that accommodates different learning algorithms and instantiate it with PPO, SAC, and AWR. Theoretically, we provide an information-theoretic analysis of how NSL enables efficient adaptation. Empirically, we study these NSL algorithms on LIBERO, RoboTwin, and four real-robot manipulation tasks under interaction, computation, and execution constraints. Across these settings, we develop a suite of effective NSL variants and derive practical recommendations for choosing among them under different resource and deployment regimes.
Read the full abstract

All authors Junjie Lu, Chuheng Zhang, Yucen Wang, Yuhua Jiang, Xinyao Qin, Min Xu, Li Zhao

Earth and Planetary Astrophysics

OGLE-2016-BLG-0523: A Low-mass Brown Dwarf or super-Jupiter in a $>100$ au Orbit

Hongjing Yang, Weicheng Zang, Yoon-Hyun Ryu et al. (+37 more)

To search for very wide-orbit binary-lens events, we cross-match microlensing events detected by OGLE-IV and KMTNet between 2011 and 2024. After removing contaminants from variable stars, we identify OGLE-2016-BLG-0523 as a microlensing event showing two short-lived bumps in its light curve, with the two separated by 8.36 yr. We investigate three scenarios to explain the two bumps, including a binary-lens single-source (2L1S), a single-lens binary-source (1L2S), and two unrelated single-lens single-source events (2$\times$1L1S). The 1L2S interpretation is excluded due to its poorer fit to the data, while the 2$\times$1L1S scenario is disfavored by the low event rate, unphysical total source flux, and unphysical color for one of the two putative sources. For the surviving 2L1S interpretation, the Einstein timescales of the two bumps are $8.343 \pm 0.076$ days and $3.231 \pm 0.047$ days, respectively, implying a binary mass ratio of $q = 0.15$. A Bayesian analysis indicates that the primary lens is most likely a low-mass M dwarf, while the companion is probably a low-mass brown dwarf or super-Jupiter. The inferred projected separation is about 450 au, making the companion the widest low-mass object discovered via microlensing to date. Our discovery highlights both the promise and challenges of identifying very wide, low-mass companions through microlensing. Although upcoming facilities such as Roman and Earth 2.0 space telescopes will improve sensitivity to such events, overcoming degeneracies and distinguishing them from variable phenomena will remain essential for building a robust statistical sample.
Read the full abstract

All authors Hongjing Yang, Weicheng Zang, Yoon-Hyun Ryu, Andrzej Udalski, Yuchen Tang, Michael D. Albrow, Sun-Ju Chung, Andrew Gould, Cheongho Han, Kyu-Ha Hwang, Youn Kil Jung, In-Gu Shin, Yossi Shvartzvald, Jennifer C. Yee, Sang-Mok Cha, Dong-Jin Kim, Hyoun-Woo Kim, Seung-Lee Kim, Chung-Uk Lee, Dong-Joo Lee, Yongseok Lee, Byeong-Gon Park, Richard W. Pogge, Przemek Mróz, Michał K. Szymański, Jan Skowron, Radoslaw Poleski, Igor Soszyński, Paweł Pietrukowicz, Szymon Kozłowski, Krzysztof A. Rybicki, Patryk Iwanek, Krzysztof Ulaczyk, Marcin Wrona, Mariusz Gromadzki, Mateusz J. Mróz, Jiyuan Zhang, Shude Mao, Hongyu Li, Qiyue Qian

Computer Vision

Images Talk Back: Guiding MMDiTs by Suppressing Image-to-Text Feedback

Yucheng Yang, Xiaoxiao Sun, Shusheng Yang et al. (+2 more)

Multimodal diffusion transformers (MMDiTs) allow the evolving image to update internal text representations through joint attention. These representations then guide subsequent image updates, raising a question: does this feedback help correct deviations from the prompt, or preserve them? We find that image-to-text (I2T) feedback can preserve the objects and attributes already represented in the evolving image, including errors. Transplanting I2T messages can rescue failures or disrupt successes. At matched intervention strengths, suppressing the influence of I2T feedback improves generation, whereas opposite controls degrade it. These findings motivate Semantic Feedback Guidance (SFG), which suppresses the influence of I2T feedback by guiding away from an auxiliary branch that strengthens this feedback and weakens text-to-image conditioning. The in-place variant, iSFG, directly suppresses I2T feedback and strengthens text-to-image conditioning, requiring no extra model evaluations. Temporal interventions motivate running the SFG branch during the first 20% of image sampling steps, requiring about $1.2\times$ the base NFE without classifier-free guidance (CFG). Experiments across three image-generation backbones show gains in alignment and preference-model scores. On FLUX, adding SFG to CFG raises GenEval from $0.6661$ to $0.7479$; on SD3.5M, iSFG adds $+1.33$ HPSv2.1 over CFG at zero extra NFE. The control also transfers to text-to-video and image-to-video generation, improving VBench scores. Code is available at https://github.com/yyc1215lll-creator/SFG.
Read the full abstract

All authors Yucheng Yang, Xiaoxiao Sun, Shusheng Yang, Jiakang Chen, Serena Yeung-Levy

Probability

Sandpile avalanches in two dimensions are finite

Ahmed Bou-Rabee

In the abelian sandpile model on $\mathbb{Z}^d$, a site with at least $2d$ grains topples by sending one grain to each neighbor. In a box $[-N,N]^d$ whose boundary absorbs grains, adding grains at random sites and toppling defines a Markov chain. Athreya and Járai (2004) constructed the stationary state on $\mathbb{Z}^d$ as the limit of its stationary distributions as $N\to\infty$. They asked whether, when a grain is added at the origin of a sandpile in the stationary state, the resulting avalanche of topplings is finite. Járai and Redig (2008) proved that the answer is yes for $d\geq 3$, using the transience of random walk. We prove, using a different method, that avalanches are also finite for $d=2$.
Read the full abstract

High Energy Physics (Theory)

A universal sector of $\mathcal{N}=1$ class $\mathcal{S}$ spectra

Ritabrata Bhattacharya, Abhay Katyal and Oscar Varela

We determine a universal sector of the spectrum of light operators in four-dimensional $\mathcal{N}=1$ class $\mathcal{S}$ superconformal field theories arising from M5-branes wrapped on a compact hyperbolic Riemann surface $\Sigma_2$. We work at large $N$, analysing Kaluza–Klein fluctuations about the dual Bah–Beem–Bobev–Wecht anti-de Sitter backgrounds for general values of the twist parameter $z$. To obtain these spectra, we specialise recently derived mass operators to modes that are constant on $\Sigma_2$ by imposing a suitable $z$-dependent generalised $G$-structure. We then organise the resulting eigenstates into long and protected $\mathcal{N}=1$ supermultiplets with definite flavour quantum numbers. As $z$ varies, rich patterns emerge in the spectrum, including both isolated multiplets and infinite towers of multiplets, which we characterise analytically.
Read the full abstract

Data Structures and Algorithms

Constructive Arboricity in Almost-Linear Time for Dense Graphs

Iman Gholami, Mohammad T. Hajiaghayi, Mahdi JafariRaviz et al. (+3 more)

The arboricity of a graph is the minimum number of forests needed to partition its edges. We present a randomized algorithm that computes the exact arboricity of a simple graph and constructs an optimal forest partition in $\min\{n^{2+o(1)},m^{4/3+o(1)}\}$ time both in expectation and with high probability, where $n$ and $m$ denote the numbers of vertices and edges. In particular, it runs in almost-linear time, $O(m^{1+o(1)})$, on dense graphs with $m=\Theta(n^2)$. We also give the first known almost-linear-time algorithm that constructs a partition into at most one more forest than optimal.
Read the full abstract

All authors Iman Gholami, Mohammad T. Hajiaghayi, Mahdi JafariRaviz, Alireza Kaviani, Soheil Mohammadkhani, AmirMohammad Shahrezaei

Quantum Physics

A Warped Born Rule Turns an Observer into a Maxwell's Demon

Olumakinde Ogunnaike

Quantum mechanics seems to be an incredibly delicate theory. Slight alterations to nearly any part of its structure can lead to huge qualitative differences and wildly unphysical predictions. Here, we explore the many-body consequences of replacing the Born rule $p(m)\propto\|P_m|\psi \rangle|^2$ by a warped rule $p_b(m)\propto\|P_m|\psi\rangle\|^{2b}$. Specifically, in systems with conserved particle number, evolution punctuated by repeated local particle observations acquires a probabilistic bias over measurement histories with an effect controlled by the sign of $b-1$. For $b>1$, this bias generically drives the gas to collect into nearly frozen configurations. Similar to the action of Maxwell's demon, the simple act of observation seems to extract entropy from a closed system. For $b<1$, the bias is reversed, generically scouring out long-wavelength fluctuations to produce a hyperuniform gas. We study this effect in a one-dimensional quantum gas, where these two phases are derived analytically for a particular measurement scheme and corroborated numerically for more general procedures with discrete-time projective-measurement circuit models using matrix product state simulations. We additionally discuss how this bias manifests in a common technique that often employs this warped rule, the “replica trick,” implying qualitative differences between finite-replica and replica-limit calculations.
Read the full abstract

Computer Vision

Hide and Seek: HIDE Concepts to SEEK explanations for black box models

Bhanu Tokas, Rahul Nair, Rockwell Jackson et al. (+1 more)

Interpreting machine learning model predictions is a central challenge, particularly for black-box vision models. Concept-based attribution methods interpret models using human-understandable concepts. Existing concept-based methods can only interpret classification and regression models. They require access to model weights or large concept-annotated datasets, and most methods cannot provide image-level attribution. We propose Hide and Seek (HnS), a new method for concept-level attribution that does not require access to model weights or labeled concept datasets. HnS can interpret a wide range of vision models. Given a concept and image set, HnS retrieves concept-aligned images using a vision-language model, localizes the concept within each image, masks the localized region, and measures attribution from the change in the black-box model's output between the original and masked images. This enables faithful attribution by directly measuring the model's behavior and supporting local attributions. We validated HnS on CelebA and COCO-Stuff. Compared to existing methods, HnS showed better agreement with ground-truth attributions. We demonstrated the applicability of HnS on diverse tasks, including image captioning and satellite image segmentation, where no existing concept attribution method can be used. We provided our anonymous code in the supplementary material.
Read the full abstract

All authors Bhanu Tokas, Rahul Nair, Rockwell Jackson, Hannah Kerner

Analysis of PDEs

A Riesz-Map Preconditioned Extragradient Method for Price Formation in Mean Field Games

Yeva Gevorgyan and Diogo Gomes

We develop and analyze a projected extragradient method, preconditioned by a Riesz map, for a first-order price-formation mean field game in one space dimension. The model couples Hamilton–Jacobi–Bellman and transport equations through a global market-clearing constraint. Finite propagation justifies truncating the first-order transport equation to a bounded interval. A monotone weak formulation incorporates the mixed initial–terminal data. Its Riesz map uses $L^2$ norms for density and price directions and a space-time $H^1$ norm for value-function directions, preconditioning the residual without adding higher-order regularization terms to the MFG equations. A density-only argument rules out cocoercivity on the localized admissible set in every equivalent Hilbert metric, motivating extragradient updates. For $H(x,p)=p^2/2$, we use a monotone upwind discretization with adjoint-consistent transport and a state-constraint boundary closure. We establish a mesh-independent Lipschitz bound on bounded convex admissible sets with uniform density, slope, and price bounds. With exact metric projection onto these sets, the iteration converges to a solution of the discrete variational inequality on each fixed grid, with an admissible step-size threshold independent of both mesh sizes. Convergence under mesh refinement is not proved. On a quadratic benchmark, all reported runs reach the stopping tolerance with one fixed step size across a factor-of-16 reduction in both mesh sizes. Price errors and value-function errors measured away from low-density regions decay approximately at first order under refinement, while density errors decrease more slowly.
Read the full abstract

Quantum Algebra

The Haagerup VOA at $c=8$

Andrew Riesen

We announce an explicit construction of a unitary, strongly rational vertex operator algebra $U$ of central charge $8$ inside $V_{E_6\oplus A_2}$, with a ribbon equivalence $\mathrm{Rep}(U)\simeq Z(\mathcal H_3)$ realizing the Haagerup double. The vertex algebra is generated by six affine currents and three specified weight-two fields; its weight-one Lie algebra is $\mathfrak{sl}_2\oplus\mathfrak{sl}_2$, at levels $1$ and $39$. We outline the argument, which combines exact lattice calculations, an all-weight strong-generation induction and $C_2$-cofiniteness, semisimple descent for a full corner of Zhu's algebra, modular Galois symmetry, and holomorphic $E_8$ extensions.
Read the full abstract

Algebraic Topology

Constructive Bounds for the Topological Expressivity of Neural Networks

Kosio Beshkov

The expressivity of a neural network $\Phi$ describes the class of functions that a network can express. The term topological expressivity refers to a family of measures particularly suited to understand binary classification and is often formulated in terms of the Betti numbers of a decision boundary $\beta (\Phi^{-1}(0))$. Despite its popularity, this measure is not as useful for regression and other problems that do not involve a decision boundary. Instead, we propose to study topological expressivity in terms of the Betti numbers of the output of a neural network $\beta(\Phi(X))$. We formally prove that a ReLU network with architecture $(d,\sum\limits_{k=1}^{d } \beta_k(3k+2),\sum\limits_{k=1}^{d}\beta_k(k+1))$ can express spaces homotopy equivalent to wedge sums of spheres with arbitrary Betti numbers. This gives upper bounds for any arbitrary connected ($\beta_0=1$) sequence of Betti numbers, but by constructing spaces homotopy equivalent to products of $N$ spheres of dimension $d_1,...,d_N$ we achieve much better bounds $(\sum\limits_{n=1} d_n, 3\sum\limits_{n=1}\limits^N d_n + 2N, \sum\limits_{n=1}\limits^N d_n + N) $ for Betti numbers given by $\prod_n (1+x^{d_n})$. Finally we conjecture a much stronger tight bound which gives the architectural bound $(d,d+1,d+1)$ for a space homeomorphic to the sphere.
Read the full abstract

High Energy Physics (Phenomenology)

The LZ inelastic splitting as a probe of anomaly mediation

Seung J. Lee and Taewook Youn

The LZ 248 keV recoil candidate admits an interpretation in terms of pseudo-Dirac dark matter with a mixing-suppressed inelastic $Z$ interaction. In the MSSM, the Higgsino limit conflicts with the adopted solar-neutrino bound, while a gaugino realization requires an unprotected mass degeneracy. We realize such spectra in the MRSSM, where an $R$-symmetric Dirac bino produces the required $Z$ transition and wino coannihilation supplies the observed relic abundance. We impose the Higgs and gauge coupling relations of $N=2$ supersymmetry at a high scale and evolve them to low energy. With the remaining model parameters held fixed, we obtain spectra spanning roughly 300-850 GeV that satisfy the adopted constraints and have a small diagonal axial current. If anomaly mediation dominates the Majorana splitting, its local scale is tens of MeV for splittings of a few hundred keV. A five-dimensional warped construction generates this scale through Goldberger-Wise stabilization and a loop correction to the no-scale Kähler potential. The spectra remain testable through elastic scattering, solar neutrinos and long-lived charged states.
Read the full abstract

Robotics

Chronos Policy: Learning Robot Policies on Physical Time

Changhao Wang, Jaden Clark and Shuran Song

Robot policies often convert sensing and control into fixed-rate observation-action sequences, leaving physical time implicit in sequence position. This assumption breaks when observation rates or delays vary, or when inference latency changes. We introduce Chronos Policy, which makes physical time explicit by conditioning on observation capture times and desired action execution times. We study two ways to incorporate time into a diffusion transformer: adding timestamp embeddings to token features, or encoding physical time in positional encoding through time-aware rotary positional encodings (time-aware RoPE). We find that time-aware RoPE is more robust, allowing attention to depend directly on physical time while keeping token content unchanged. During training, we temporally randomize which observations are sampled and when the action chunk begins, exposing the policy to a wider range of sensing and execution schedules that may occur during deployments. This also enables learning from asynchronous, multi-rate data without forcing all measurements onto a fixed temporal grid. Across simulation and real-world experiments, Chronos Policy improves performance on dynamic tasks, where observation-to-action timing is critical, and on multisensory tasks, where the relative timing between observations matters. On standard quasi-static tasks, it preserves nominal performance while improving robustness to changes in sensing and inference timing.
Read the full abstract

Machine Learning

AI Models Can Provably Hide Arbitrary Capabilities

Andis Draguns, Sumeet Ramesh Motwani, Raymond Douglas et al. (+1 more)

Capability evaluations aim to surface dangerous model behaviors before deployment, but their reliability depends on the assumption that hidden capabilities can be elicited. We show this assumption does not hold under adversarial conditions by constructing backdoor attacks that embed encrypted neural networks within host model weights, decrypting and executing them only upon receiving a secret trigger. Using digital lockers, these hidden circuits remain provably hard to elicit and interpret under cryptographic assumptions, even given full access to model weights, extending prior work from hiding fixed strings to arbitrary computations. We implement our proof of concept constructions in PyTorch and evaluate elicitation through fine-tuning, weight noising, gradient-based trigger search, and latent adversarial training, using a small chemical reaction prediction task as a proxy for CBRN-relevant capabilities. Our results demonstrate resistance to these methods and, together with our cryptographic guarantees, reveal a limit of elicitation-based capability audits. We release our models to support the improvement of defenses.
Read the full abstract

All authors Andis Draguns, Sumeet Ramesh Motwani, Raymond Douglas, Christian Schroeder de Witt

Symplectic Geometry

On some rigidity results for high-dimensional confoliations

Fabio Gironella and Seungook Yu

The main result is an obstruction to the presence of contact regions for c-symplectic confoliations in any odd ambient dimension. As a first application, we prove a 5-dimensional analogue (under stronger assumptions) of a Reeb stability theorem for confoliations in dimension 3 due to Eliashberg and Thurston. Refining work of the first author with Cardona, we further give two new notions of contact approximations of c-symplectic confoliations in any odd ambient dimension generalizing the 3-dimensional notion of Eliashberg-Thurston. As a second application of our main theorem, we then give an example, in any ambient odd dimension, of a c-symplectic (and in fact strong symplectic) foliation that cannot be contact approximated, in either of these two senses. This relies on the property, of independent interest, that contact structures approximating (in these senses) a weakly c-symplectic confoliation are also weakly fillable. The proofs of all these statements rely on pseudo-holomorphic curve theory. Lastly, we describe several natural classes of examples of these types of contact approximations of c-symplectic confoliations.
Read the full abstract

High Energy Astrophysical Phenomena

Adiabatic evolution of hadronic neutron stars

Domenico Logoteta, Francesco Zeno Constanzo, Ignazio Bombaci et al. (+3 more)

In this paper we compute several structural properties of static and rotating neutron stars, considering both rigid and differential rotation. To this end, we use a finite-temperature beta-stable equation of state (EOS) extending the zero-temperature hadronic EOS derived by Fortin et al. (2017). We account for the formation of hyperons and the presence of a neutrino-trapped component. Our goal is to provide insights into the long-term evolution of a protoneutron star (PNS) formed after a core-collapse supernova (CCSN) or a remnant resulting from a binary neutron star merger (BNSM). The EOS is derived using a relativistic mean-field quantum many-body theory within the Hartree approximation. Coupling constants are determined using experimental data on hypernuclei and hypernuclear matter to better account for uncertainties in hyperon single-particle potentials. Our EOS models reproduce the typical features of cold neutron stars and PNSs for nucleonic and hyperonic compositions. In particular, they are consistent with measured masses (2.01 M_sun and 2.14 M_sun for PSR J0348+0432 and PSR J0740+6620) and dimensionless tidal deformabilities from GW170817. Thermal contributions act differently depending on composition and rotation state. Finally, our evolutionary scheme provides a simple way to interpret the long-term evolution of BNSM remnants or PNSs from CCSNe, complementing dynamical simulations unable to cover full evolutionary timescales.
Read the full abstract

All authors Domenico Logoteta, Francesco Zeno Constanzo, Ignazio Bombaci, Sebastiano Bernuzzi, Albino Perego, David Radice

Computer Vision

Can machines reason non-verbally?

Shaheer U. Saeed, Zion Tse, Gregory Slabaugh et al. (+2 more)

Modern machines can solve equations, follow logic and reason through language, yet they still struggle with problems that require visual non-verbal reasoning. This is evident even in advanced systems such as DeepSeek, ChatGPT and Gemini, which remain limited on visual tasks such as completing abstract patterns or delineating ambiguous regions-of-interest. Unlike explicit intermediate steps in verbal reasoning which can be verified through logic or automated tests, visual reasoning involves implicit intermediate steps with no clear methods for verification. Non-verbal reasoning is therefore challenging to model and, as a result, remains a fundamental limitation of current machine intelligence. Here we introduce GenVerOpt, a general framework that unifies generation, verification and optimisation into a reasoning pipeline for machines. By generating candidate solutions, verifying their correctness, and iteratively refining them, GenVerOpt enables per-sample optimisation rather than relying solely on dataset-level trends captured during model training. This process mirrors human problem-solving where multiple possibilities are imagined, evaluated, and refined until a correct solution emerges. Across diverse domains, GenVerOpt demonstrates capabilities beyond existing systems: it surpasses humans on abstract IQ puzzles, matches clinicians in prostate cancer segmentation, exceeds experts in liver vessel reconstruction and outperforms state-of-the-art approaches in tumour boundary alignment. These results show that machine reasoning can extend beyond language, establishing visual non-verbal reasoning as a foundation for more flexible and generalisable intelligence.
Read the full abstract

All authors Shaheer U. Saeed, Zion Tse, Gregory Slabaugh, Daniel C. Alexander, Yipeng Hu

Machine Learning

KL Before and After Mixing: A Round-Complexity Separation in Parallel Sampling

Tianyi Wu

A parallel sampler can choose at random which coordinates to generate together. For a target $p$ and the output law $q_U$ of a schedule $U$ drawn afresh for each output, we compare the schedule-averaged divergence $\mathbb{E}_U D_{\mathrm{KL}}(p\|q_U)$ with the output-law divergence $D_{\mathrm{KL}}(p\|\mathbb{E}_U q_U)$, in nats. Rounds commit independent draws from exact conditional marginals frozen at the round's start, each output's schedule is drawn before sampling from a finite mixture of schedules, and every scheduled round is charged. For the uniform distribution on binary strings of even length $n$ with $n/2$ ones, the least expected number of rounds at a fixed tolerance $\varepsilon>0$ is $\Theta_\varepsilon((\log n)^2)$ under the first criterion but $(e^{-\varepsilon}+o(1))\log_2\ln n$ under the second, and the double-logarithmic order is attainable with the same round count on every run. For this target, a count-state recursion gives the exact least output divergence at every round cap over all causal randomized policies. The separation extends to weighted fixed-count targets with bounded weight ratios and quota fractions bounded away from zero and one, without permutation symmetry, and to the full-support Beta$(1,1)$–Bernoulli target, where the sharp expected cost under the output-law criterion is $(e^{-\varepsilon}/2+o(1))\log_2\log_2n$, attained with rare $n$-round runs. Prior work implies the order of the schedule-averaged cost on the balanced target and its $O_\varepsilon((\log n)^2)$ upper bounds on both extensions.
Read the full abstract

Networking

Goal-Oriented Bundle Management and Flow Control for Deep-Space Communications via Reinforcement Learning

Sajjad Baghaee, Aimin Li, Shanmugarajan Balasubramanian et al. (+1 more)

Deep-space communications operate under intermittent connectivity, long propagation delays, and relay storage constraints. In such networks, the Bundle Protocol enables store-carry-forward communication, requiring relay nodes to manage limited storage and scarce contact opportunities. Relay-side scheduling and dropping decisions are inherently goal-oriented: transmission resources should be allocated to bundles most valuable to downstream mission objectives. Conventional schedulers, such as FIFO, do not explicitly optimize task utility and may provide unfair service across heterogeneous planetary sources. This paper presents a closed-loop modeling and learning framework for relay-side bundle and flow management in a contact-driven deep-space network, where lunar and Martian updates are forwarded through a shared-buffer relay to Earth. We quantify destination-side information freshness using excess age $\Delta_i^e(t)=[\Delta_i(t)-d_i(t)]^+$, where $\Delta_i(t)$ denotes the Age of Information of source $i$ and $d_i(t)$ is a source-dependent delay baseline. Based on this metric, we define a goal-oriented loss that supports source-dependent nonlinear penalties, allowing prioritization according to mission relevance. We formulate relay control as a constrained semi-Markov decision process (CSMDP) and solve it using masked Lagrangian Proximal Policy Optimization (PPO). Feasibility-aware action masking restricts transmissions to bundles that can complete within the remaining contact window. DTNSim simulations indicate improved age-based utility and inter-flow fairness compared with baseline schedulers. The proposed framework jointly optimizes relay-side bundle scheduling, dropping, and flow control under contact-driven dynamics with explicit fairness considerations across heterogeneous sources.
Read the full abstract

All authors Sajjad Baghaee, Aimin Li, Shanmugarajan Balasubramanian, Elif Uysal