Artificial Intelligence
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 27 August
Although Large language models (LLMs) mediate access to knowledge and computational assistance, their capabilities should benefit vulnerable groups in the same way. However, it is unclear whether existing AI systems are inclusive enough for blind and deafblind users to access the same functionality through Braille, whose indicators, contractions, and digital representations introduce distinct requirements for model comprehension. To this end, we introduce BrailleBench, a benchmark for evaluating LLMs in Braille comprehension from different Criteria. BrailleBench aligns 5,570 instances from five datasets, including mathematics, commonsense, and multi-hop question answering across English and Braille Grades 1 and 2. Different configurations are designed to understand whether the systems can comprehend Braille-authored content, express answers in Braille, and complete end-to-end Braille interaction. To ensure the quality and prevent evaluation bias, the benchmark is built through a deterministic, expert-reviewed pipeline via a self-created Braille Toolkit without using any data instances generated by LLMs. We evaluate six representative LLMs from various aspects. The results reveal a persistent gap between print-English capability and Braille accessibility. Braille understanding and expression are asymmetric, where Grade 2 is especially fragile on the input side compared to Grade 1, and fully Braille requests further reduce performance. The experimental observations provide valuable guidance for the development of future Braille AI systems. All related resources in BrailleBench are publicly available for future research.
Lattice Field Theory
Efficient Quantum Simulations of Yang-Mills theory with Maximal-tree Gauge
We develop a quantum algorithmic framework for the efficient simulation of Yang–Mills theories, including the $\mathrm{SU}(3)$ gauge theory in Quantum Chromodynamics (QCD). The framework uses maximal-tree gauge in terms of gauge field variables that removes all local gauge redundancies. In the resulting gauge-fixed formulation and digitization in the field-amplitude basis, we show that Hamiltonian time evolution admits an efficient implementation based on quantum singular value transformation (QSVT). We derive upper bounds on the total number of qubits and gate complexity, finding polynomial scaling with the inverse simulation precision $1/\varepsilon_s$, lattice volume $\mathcal{V}$, gauge coupling $g$, and target energy scale $E$. Our results provide a rigorous complexity-theoretic demonstration that non-Abelian Yang–Mills theories can be simulated efficiently on quantum computers, paving the way toward first-principles quantum simulations of non-perturbative QCD dynamics.
Read the full abstract
Artificial Intelligence
Naive Prompt Optimization: Rethinking the Need for Complex Prompt Search
Efficiently improving autonomous agents across diverse tasks is central to accelerating recursive self-improvement (RSI) in agentic AI, with prompt optimization emerging as a promising approach capable of delivering performance gains comparable to those achieved by fine-tuning model weights, while reducing computational costs in both optimization and serving. However, recent developments increasingly favor unnecessarily complex prompt optimizers. We introduce Naive Prompt Optimization (NPO), a lightweight single-lineage method that iteratively revises prompts using a teacher model with rollout feedback. NPO achieves comparable or better performance than GEPA with fewer rollouts, and its advantage increases with stronger teacher models, suggesting that stronger teacher reasoning can partially substitute for optimizer-side search complexity. In interactive games, NPO remains broadly competitive with GEPA, while GRPO performs better on some tasks less amenable to prompt optimization. We also show that NPO-optimized prompts elicit similar performance improvements when applied verbatim to other student models, especially across models within the same family. Overall, our preliminary results show that simple, linear prompt optimization can rival substantially more sophisticated and complex search procedures.
Read the full abstract
Computation and Language
SCIT: Testing Causal Cache Carriers in Latent Chain-of-Thought Models
Latent chain-of-thought models move intermediate reasoning from emitted text into continuous states, improving compactness but hiding the causal object. We introduce SCIT, the Suffix Cache Interchange Test, a causal protocol that constructs exact source-recipient counterfactuals, patches declared cache segments, and identifies which transformer object carries the counterfactual computation. SCIT combines sufficiency tests with K/V component splits, hidden-state controls, semantic source controls, decoded validation, and matched corruption. On CODI-GPT2 and a Sim-CoT-style GPT-2 reproduction, counterfactual arithmetic transfers primarily through value-cache suffix trajectories rather than hidden states, keys, reusable answer slots, or single-token triggers. Complete sufficiency-and-necessity evidence for the late-value-suffix mechanism holds for the main CODI-GPT2 checkpoint; the Sim-CoT-style checkpoint shows the same sufficiency and decoded-control pattern but insufficient matched-corruption evidence for a necessity call. Beyond these local arithmetic cells, SCIT reveals carrier-regime shifts: arithmetic-like GPT-2/1B cells preserve latent-tail value/KV transfer, whereas competent 8B and repaired non-arithmetic cells route through prompt-prefix or full-cache K/V; boundary cells receive no mechanism call. SCIT therefore contributes a cache-level diagnostic, a checkpoint-specific GPT-2 arithmetic mechanism, and a competence-gated carrier map rather than a universal latent-tail claim.
Read the full abstract
Instrumentation and Detectors
Design and performance of the Fast Beam Condition Monitor for luminosity and background measurement at the CMS Experiment in LHC Run 3
The Fast Beam Condition Monitor (BCM1F) has been used at the CMS Experiment since the first LHC circulating beams in 2008. Originally meant as a beam-induced background monitor for fast beam losses detection, it showed a potential also for luminosity measurements in 2012 running, and has been used for luminosity measurements since the beginning of Run 2 data taking in 2015 as a part of the Beam Radiation, Instrumentation and Luminosity (BRIL) system. Over the years, the system has undergone various upgrades to the sensors, the front-end and back-end electronics, providing improvements in the precision of the measurements, that remain valid in the higher pileup conditions of LHC Run 3 (2022-2026). Based on the experience of all BCM1F Run 2 upgrades, the detector was completely rebuilt prior to LHC Run 3 using AC-coupled silicon-pad diodes and active cooling. This latest detector version exhibits excellent linearity with instantaneous luminosity and achieves nanosecond-level timing precision, enabling improved systematic corrections for luminosity and background measurements. This paper presents a detailed overview of the detector system for LHC Run 3, including the selection and qualification of sensors as well as a summary of the readout system. It also outlines the processing and calibration strategy for luminosity data, discussing operational hurdles and comparing BCM1F measurements to other CMS luminosity measurements to assess the system's performance as a luminometer. Lastly, the implications for the design of a future luminosity detector to be used in the envisioned HL-LHC upgrade are discussed.
Read the full abstract
Optimization and Control
Asymptotic consensus and flocking under decaying persistent excitation on rooted digraphs
In this paper, we investigate first- and second-order alignment models with non-universal interaction, time delays and possible communication failures, extending the results in [17] to interaction digraphs that are only assumed to be rooted and to a weaker Persistence Excitation Condition. In particular, we allow the amount of interaction over time intervals of fixed length to decay polynomially in time. For the first-order Hegselmann-Krause type model, we prove asymptotic convergence to consensus under a suitable condition relating the decay exponent of the communication weights to the maximal distance from the root. For the second-order Cucker-Smale model, we establish asymptotic flocking under an additional assumption on the decay of the influence function. These results show that collective behavior can still emerge under progressively weakening communication and without requiring strong connectivity of the interaction graph.
Read the full abstract
Materials Science
Crystal Structure Prototype Identification via Element-Mapped Rotation-Invariant Descriptors
We present a method for identifying crystal structure prototypes from local atomic environments. A center atom and its neighbors within a distance-normalized cutoff are mapped to anonymous element types (A/B/C/D) by proximity, and a NEP-style descriptor-radial Chebyshev moments plus contracted spherical-harmonic (S-vector) invariants-is computed per type block. The descriptor is rotation-invariant by construction (MLFF-style contraction with analytic normalization constants), so no rotation augmentation is needed. A trainable block-diagonal projection compresses the 316-dimensional raw basis into a 168-dimensional learned descriptor fed to a three-layer MLP. The model classifies each atom into one of 271 AFLOW crystal structure prototypes or the amorphous class (272 classes), reaching 99.38% accuracy on 538,190 per-atom samples. The full pipeline (descriptor extraction + inference) runs on a single CPU core at 3.9 s per 10,000 atoms, making it suitable for website deployment.
Read the full abstract
Theoretical Economics
Strategy-Proof and Minimally Wasteful Random Assignment
We study random assignment of indivisible objects among a set of agents with strict preferences and outside options. When agents may rank some objects as unacceptable, we consider different notions of measuring waste of object(s) from an ex-ante perspective. The most natural one is $q$-agent-object-waste whereby both one agent and one of his acceptable objects are unassigned with at least probability $q$. On the one hand, we show that any mechanism satisfying equal treatement of almost equals (whereby any two agents with identical rankings over objects receive the same probability shares for objects they both regard acceptable), strategy-proofness and ex-post weak non-wastefulness (whereby in any assignment in the support we cannot have that both one agent and one of his acceptable objects are unassigned) must be $q$-agent-object-wasteful with $q\geq \frac{1}{6}$. On the other hand, we show that random serial dictatorship (RSD) attains the minimal bound of $\frac{1}{6}$ in this class for four agents or three objects. In addition, we consider $q$-object-wastefulness where an object remains unassigned with probability $q$, while agents, who consider it acceptable, remain unassigned with aggregate probability $q$. We again show that RSD attains the minimal bound of $\frac{1}{4}$ in this class of mechanisms with respect to $q$-object-wastefulness for three objects. Finally, we show that random deferred acceptance (RDA) may be strictly less agent-object-wasteful than RSD (but at the cost of violating equal treatment of almost equals).
Read the full abstract
Artificial Intelligence
What Makes Good Agentic Data? An ACE Lens on Data Generation for LLM Agents
LLM agents increasingly rely on generated interaction data to learn how to interact with external environments. Agentic data generation must maintain consistency among environments, tasks, interactions, and success signals while producing experience that is useful rather than merely abundant. Existing work spans many agent domains, but domain-centered organization and heterogeneous evaluation often obscure common generation mechanisms and conflate candidate construction with verification and selection. This work develops a two-level framework for the field. First, we represent agentic data as a common factorized object $(E,q,\tau,v)$, comprising an environment specification, task signal, interaction realization, and optional verifier. We organize generation paradigms by their primary anchor and dependency structure. Second, we formulate generation as constrained distribution design through the Accuracy-Complexity-divErsity (ACE) lens. Accuracy establishes the feasible support of grounded and internally consistent data. Within this support, Complexity places learning mass relative to the capability of a declared learner and execution configuration, while divErsity controls coverage and redundancy of data. Using this framework, we explore how prior work verifies generated experience, constructs and calibrates difficulty, and expands behavioral coverage. The literature reveals a shift toward execution-grounded accuracy, learner-relative complexity, and diversity beyond surface variation or dataset size. We further discuss broader directions and emerging trends in agentic data generation through the ACE lens, including their implications for scaling, data sources, training regimes and adaptive learning. Overall, the central challenge is not simply to generate more data, but to continually allocate valid, informative, and non-redundant experience as agents and environments evolve.
Read the full abstract
Machine Learning
LEON: Operator-Structured Latent Evolution for World Modeling
Modeling action-conditioned physical evolution is fundamental to latent world models. World Action Models (WAMs) couple such models with policy learning through predictive supervision or future-conditioned action generation. Yet their latent transitions typically rely on transformer-based predictors, leaving dynamical structure to be learned implicitly from data. We derive a latent-transition formulation from controlled mechanical evolution. A common finite-dimensional projection of observable dynamics yields a shared operator family affine in physical control and an explicit projection residual. Guided by this principle, we introduce the Latent Evolution Operator Network (LEON), which adapts pretrained visual features into learned observables and evolves them through an affine family of shared operator matrices and complementary additive updates. Coefficients inferred from evolving observables and action-related context modulate the operators. We instantiate LEON in two WAM architectures while preserving their distinct original prediction targets and policy interfaces. Across single-arm and bimanual manipulation, LEON improves average policy success over corresponding WAM baselines, including under diverse perturbations. Ablations support its observable adaptation and operator-structured evolution, while controlled-system evaluations demonstrate improved extrapolation beyond training conditions.
Read the full abstract
High Energy Physics (Phenomenology)
An Axial $U_A(1)_{L_\mu-L_\tau}$: UV Completion and Experimental Searches
We propose an anomaly-free and renormalizable axial $U_A(1)_{L_\mu-L_\tau}$ model and study its experimental signatures for $A'$ masses from the MeV scale to the TeV scale. The opposite charges of the left- and right-handed charged leptons forbid the usual muon and tau Yukawa interactions. Their masses are instead generated by a singlet scalar and heavy vector-like leptons through a universal-seesaw mechanism. We focus on heavy vector-like leptons, small light–heavy mixing, and $m_s\gtrsim10~\mathrm{GeV}$. In this limit, the observables considered here depend mainly on $(m_{A'},g_X)$, while the other model parameters are restricted by mixing and perturbativity. We confront this benchmark with current experimental searches. For neutrino trident production, our finite-$m_\mu$ calculation shows that $A'$ modifies the axial weak coefficient, rather than the vector coefficient relevant to the usual $L_\mu-L_\tau$ model. The longitudinal mode enhances muon bremsstrahlung and gives a negative contribution to $(g-2)_\mu$; the latter dominates over the scalar contribution in our benchmark. Combining these results with invisible meson decays and four-muon resonance searches, we summarize the phenomenological constraints in the $(m_{A'},g_X)$ plane. For $m_{A'}\gg m_\mu$, vector and axial final-state-radiation rates become nearly identical, so the corresponding collider limits can be obtained by rate matching. At a future muon collider, the total rate alone does not fully resolve the interaction structure, whereas angular distributions, especially the forward–backward asymmetry in $\mu^+\mu^-\to\tau^+\tau^-$, retain direct sensitivity to chirality.
Read the full abstract
Analysis of PDEs
Rigidity and existence of Busemann profiles for the Allen–Cahn equation on Cartan–Hadamard manifolds
We study entire solutions of the Allen–Cahn equation $\Delta_g u = W'(u)$ on Cartan–Hadamard manifolds of the form $u = U\circ b$, where $b$ is a Busemann function and $U\in C^2(\mathbb{R})$; we call these Busemann profiles. Our main result is a rigidity principle requiring no homogeneity of the manifold: if $\inf \Delta_g b>0$ and $U$ has finite limits $\ell_\pm$, then $W(\ell_-)\le W(\ell_+)$, with equality exactly when $U$ is constant. In particular no nonconstant Busemann layer joins the two wells of a balanced double-well potential, in any dimension $n\ge 2$; under $K\le-\kappa_2<0$ the hypothesis is automatic by Hessian comparison, and if $W'$ is locally Lipschitz it holds under the mere nonvanishing of $\Delta_g b$. The mechanism is a dissipation identity for the Hamiltonian $\tfrac12(U')^2-W(U)$ along the flow lines of $\nabla b$, in which the mean curvature of the horospheres acts as a wave speed. Conversely, when $\Delta_g b\equiv c$ and $W$ is bistable and unbalanced, strictly increasing Busemann fronts joining the wells exist precisely when $c$ is the Fife–McLeod speed; their interfaces are then horospheres of constant mean curvature. On warped-line and rotational surfaces we test sharpness: monotone warpings admit no equal-level profile and rotational surfaces no radial two-well profile, while an even convex warping carries a nonconstant monotone layer whose nodal set is the central geodesic leaf.
Read the full abstract
Numerical Analysis
Enforcing Dirichlet Boundary Conditions in Operator Learning
Operator learning in scientific machine learning is concerned with approximation of maps between infinite-dimensional function spaces; such maps frequently arise as the solution operators of partial differential equations (PDEs). Neural operators have demonstrated broad empirical success at approximating such maps from data. However, most existing neural operator architectures enforce boundary conditions indirectly through training from data even though the boundary condition is often known exactly. Furthermore, existing modifications and approaches that do enforce boundary conditions explicitly suffer from impractical restrictions, including boundary smoothness, uniform grids, and separable, box-like domains. In this work, we propose an architecture which, independently of training, satisfies homogeneous Dirichlet boundary conditions, whilst simultaneously retaining the expressivity of existing kernel-integral neural operator architectures. This is achieved by enforcing the property that the output of each layer is contained in the span of a subset of the homogeneous Dirichlet eigenfunctions of the Laplacian on the output domain. The method requires only that the output domain be bounded with Lipschitz boundary and places no restriction on the choice of discretization, making it applicable to arbitrary mesh data and general geometries. We prove universal approximation for the resulting architecture; furthermore the approach we adopt in the analysis proves universality for a broad class of kernel-integral neural operators thereby uniting existing theory for a variety of operator learning methods. We validate the proposed method on maps defined by the coefficient to solution map in 2D PDEs: Darcy flow on a square domain and the Helmholtz equation on a circular domain. Comparisons are made with alternative methods.
Read the full abstract
Number Theory
Exceptional covers of the projective line with genus-one Galois closure: arithmetic forms over finite fields
Let $k=\mathbb F_q$. We classify, up to two-sided $k$-Möbius equivalence, all separable indecomposable tame exceptional maps $\mathbf P^1_k\to\mathbf P^1_k$ whose geometric Galois closure has genus one. The resulting arithmetic fixed-field forms are encoded by Frobenius-stable affine elliptic quotient data recovered from the cover; we prove a converse and an exact equivalence criterion. The same data determine branch arithmetic, geometric and arithmetic monodromy, the constant field, and behavior over every finite extension. In particular, over each $\mathbb F_{q^r}$, permutation is equivalent to exceptionality. We obtain explicit rank-two support formulas, necessary and sufficient occurrence criteria, and exact two-sided class counts in every tame signature, including the surviving cases in characteristics $2$ and $3$. In characteristic greater than $3$, tameness is automatic. A sharp $3$-adic obstruction arises only for a stronger cubic self-endomorphism realization problem, not for the fixed-field classification itself.
Read the full abstract
Machine Learning
Circuit Condensation: Post-Training that Concentrates a Behavior's Causal Circuit
One approach in mechanistic interpretability explains model behavior through circuits: the components and connections that carry it. Frozen discovery often returns hundreds of edges, making them hard to inspect, compare, or verify exhaustively. We introduce Circuit Condensation, which post-trains models to concentrate behaviors into smaller causal graphs. Each round prunes low-attribution edges and trains a low-rank adapter to match the original through what remains. It keeps the cut only if task accuracy and held-out perplexity remain within tolerance. Across four behaviors and eight models, condensed circuits are smaller than the strongest published frozen-weight baseline in 30 of 32 settings, by 8.1x on average and up to 316x. Repeating the search without weight updates produces larger circuits in 29 of 32 settings, showing that weight updates drive the reduction rather than search alone. For 19 small circuits, exhaustive subset tests show that 11 are exactly minimal and identify removable edges in the rest. Pair ablations reveal dependencies that single-edge tests miss. On indirect object identification, condensation isolates 24 heads, including 17 with documented roles, whereas a matched frozen circuit needs 61 heads, 36 without documented roles. The result is a sufficient subcircuit of the published mechanism rather than a reconstruction of it. The resulting circuit tracks the original model's next-token distribution and, at matched edge count, predicts its errors better than a frozen circuit in 10 of 11 evaluable settings.
Read the full abstract
Combinatorics
Kronecker Products, Polarity Quotients and Large Graph Constructions
In this paper, we establish a structural compatibility between the Kronecker product of bipartite graphs that admit polarity and their polarity quotients, and provide a sharp upper bound on the diameter of these graphs. We use these properties to construct new families of graphs, all of which have orders that asymptotically approach the Moore bound. We show that the bipartite compatibility graphs have a natural polarity, and that the corresponding polarity quotients comprise a new family of graphs of any even diameter $2k$ with degree $(q+1)^k$ for any prime power $q$, much smaller than those of other known families of that diameter and field order $q$. These families generalize the Erdős-Rényi-Brown polarity graphs. Kronecker products of these polarity quotients give another new family of graphs of any even diameter for every pair of prime powers $q,r$. Certain generalized polygons are known to admit polarity with polarity quotients approaching the Moore bound asymptotically. Kronecker products of these polarity quotients give another new family of graphs of diameters $2$, $3$ and $5$, including a diameter-$3$ degree-$18$ graph of order larger than any previously known for that degree and diameter.
Read the full abstract
Other Quantitative Biology
Conservative deterministic Markov models in mathematical biology: uniqueness of steady states, reversibility and computational methods
Ordinary differential equations are commonly used throughout the sciences to build mechanistic models of time-dependent processes. Often, such models are Markov models describing the time-evolution of different interconnected "states". When these models have no "sources" or "sinks", they naturally conserve the total population of the system. When each state is reachable (directly or indirectly) from any other state, these models are called irreducible. For many applications, a deterministic system of ordinary differential equations (ODE) is the most suitable modelling approach. We summarise important mathematical results which show that this irreducibility property guarantees the existence and uniqueness of global stable equilibria, and discuss the computationally-efficient implementation of such ODE-based models. We also discuss the condition of microscopic reversibility and show how it guarantees non-oscillatory behaviour, which enables additional efficiencies in computations. These properties and methods are demonstrated through example models of biological phenomena, where we demonstrate their importance for efficient model fitting and simulation.
Read the full abstract
High Energy Physics (Phenomenology)
Hunting long-lived doubly charged scalars at the HL-LHC
This work studies the collider phenomenology of long-lived doubly charged scalars. After reviewing the existing searches for the doubly charged scalar in both the prompt and the long-lived regimes, we identify an intermediate range of proper decay length, namely, $\mathcal{O}(0.1~\mathrm{mm}) \lesssim c\tau \lesssim \mathcal{O}(100~\mathrm{mm})$, where conventional searches lose sensitivity. We investigate the prospects for probing the doubly charged scalar both in the presence and in the absence of the $\Delta L=2$ Yukawa coupling of the $SU(2)_{L}$ complex triplet. In the presence of this coupling, $H^{\pm\pm}$ can be long-lived only for masses in the range $100$-$150~\mathrm{GeV}$, whereas in the fermiophobic scenario (i.e., in the absence of the $\Delta L=2$ coupling), the range extends to $\mathrm{TeV}$ scale. We propose a displaced-vertex search at the HL-LHC for doubly charged scalars with masses up to $\approx 1~\mathrm{TeV}$ and $c\tau=10$-$100~\mathrm{mm}$ in the fermiophobic scenario, while a benchmark point with a doubly charged scalar of mass $120~\mathrm{GeV}$ and $c\tau = 5~\mathrm{mm}$ is considered when the complex triplet scalar couples to leptons. We show that a cut on the invariant mass of the displaced vertex as reconstructed from the associated tracks can strongly suppress Standard Model backgrounds. We present projected limits on the Drell-Yan pair-production cross section for the doubly charged scalar at $\sqrt{s}=14~\mathrm{TeV}$ with an integrated luminosity of $3000~\mathrm{fb}^{-1}$, considering two illustrative assumptions for the residual background. Displaced-vertex searches thus probe a region complementary to those covered by prompt and heavy stable charged-particle searches.
Read the full abstract
Atomic Physics
Influence of Many-Body Dipole-Dipole Interactions on Excitation Transfer in a Dense Gas
We study non-radiative dipole-dipole induced excitation transfer in dense Rb vapour. We show that density dependence of characteristic time of the excitation diffusion changes from linear to non-linear for high Rb density. It is attributed to the breakdown of binary collisions approach to the dipole-dipole excitation transfer. It is shown that the observed result can be qualitatively described if the mean free path is replaced by mean interatomic distance for the chracteristic diffusion length.
Read the full abstract
Differential Geometry
Torus actions, almost non-negative curvature and fundamental groups
We show that purely topological torus symmetries impose strong restrictions on the fundamental groups of almost non-negatively curved closed manifolds and on their actions on higher homotopy groups, even when the torus action is unrelated to the metric.
Differential Geometry
Optimal Transport and the ABP Method in Higher Codimension
We prove an inverse optimal-transport principle for the ABP contact set of animmersed submanifold. A fixed function on the submanifold determines, for every probability density on a bounded convex target, a source measure and an optimal plan. For a uniform ellipsoidal target, disintegration over the submanifold gives conditional measures whose fiberwise $L^\infty$-norms satisfy a sharp weighted average estimate with constant $\omega_n/\omega_{n+m}$ in codimensions $1\leq m\leq4$. This yields the sharp ellipsoidal Michael–Simon–Sobolev inequality in the same range, including its equality cases and symmetrization consequences. To the best of our knowledge, even in the Euclidean case, the sharp constants in codimensions three and four were previously unknown.
Read the full abstract
Combinatorics
Almost factorial many facets for 0/1-polytopes
A long-standing question posed by Fukuda (1995) and Ziegler (2000) inquires about the asymptotic behavior of $g(n)$, the maximum number of facets that an $n$-dimensional $0/1$-polytope can have. A remarkable result by Bárány and Pór (2001) via probabilistic methods established that $g(n)$ is at least superexponential in $n$. In this paper, we propose a drastic change of perspective, which leads us to show that for each $n\geq 10$ there exists a $0/1$-polytope having at least $(n-\lceil 2\log_2 (n)\rceil - 1)!$ facets. This provides a significant improvement over the currently known lower bounds for $g(n)$. Furthermore, when combined with known upper bounds, our construction establishes the asymptotic behavior of $\log g(n)$ up to an error of $O((\log n)^2)$. The methods employed throughout this paper are elementary and fully deterministic. The underlying ideas in our proof stem from the combinatorics of hypersimplices and permutohedra.
Read the full abstract
Algebraic Geometry
Finite group schemes as fundamental group schemes of smooth projective varieties
In this paper we study the problem of realizing finite group schemes as fundamental group schemes of smooth projective varieties. We establish Bertini-type results and prove the triviality of fundamental group schemes of mildly singular complete intersections in projective space. With this in hand, we are able to go through a Godeaux–Serre construction and prove that, over an infinite perfect field $k$, any finite group scheme $G/k$ can be realized as the $S$-, extended Nori and Nori fundamental group schemes of a connected smooth projective variety of any dimension at least $\dim \text{Lie}(G)+2$.
Read the full abstract
Soft Condensed Matter
Asymmetric Ions in Solution are Similar to Active Brownian Particles
Molecular dynamics simulations of electrolyte ions with charge shifted by the distance $d$ from the ion geometrical center have shown that mechanical equilibrium in the ion's body frame is not established beyond a critical shift distance $d^*$. A sharp crossover to a non-zero body-frame force occurs due to frustration of water molecules failing to compensate the electrostatic pull with a local density augmentation. A solution ion with asymmetric charge experiences a net body-frame force similar to self-propelled motion of active-matter Brownian particles. Thermal equilibrium and zero laboratory-frame force are still maintained, but the diffusion constant drops significantly ($\sim 400$ times) when the charge displacement crosses the threshold value $d^*$. Alternative routes to the diffusion constant become inequivalent and the diffusion constant from the mean-squared displacement is much higher than those from velocity and force correlation functions.
Read the full abstract
Databases
Compositional Online Learning for Semantic Data Processing Systems
An LLM call in a semantic data processing system is expensive enough to dominate query cost, yet slow enough to hide a CPU-side learner's update behind its round-trip. In production, LLM compute accounts for $80-90\%$ of query cost, and each call costs $10^5-10^7\times$ a relational predicate. The latency window inverts a design constraint of classical adaptive query processing, where online learners had to stay lightweight to avoid dominating the predicates they optimize. At LLM latency, per-call gradient steps and per-batch threshold solves fit inside the round-trip. We develop compositional online learning at the LLM call boundary: a framework for combining online-learning components in semantic data processing systems. Each component makes execution-time decisions and refines its learned artifacts online. The design space spans two axes, decision granularity and learner update cadence, and the components share a single learning pattern that hides each trainer step inside the next LLM round-trip. A production case study in Cortex AISQL composes three components: a memoization layer, an online per-call filter-ordering learner, and an online per-batch cascade-routing learner. A conditional cost decomposition assigns each learning component to a distinct factor of per-row LLM cost. Under independence, the two learning components compose multiplicatively to an $11.4\times$ upper bound on a representative conjunction-filter workload. Self-selection at the cascade boundary, sample-budget shrinkage, and selectivity-estimation drift reduce it to a realistic figure near $8\times$.
Read the full abstract
Information Theory
Deterministic Identification over Additive Gaussian Channels
Modern communication systems impose strict demands on data rate, reliability, and power efficiency. In this context, emerging communication paradigms such as identification via channels have become an important topic in post-Shannon information theory, offering the potential for substantially higher identification rates than in conventional channel coding.Deterministic identification is particularly interesting for specialized communication scenarios because it provides a balance between implementation complexity and the communication gains due to higher identification rates. It is therefore a promising communication scheme for future communication systems, including molecular communication systems. Additive Gaussian channels, particularly the additive white Gaussian channel, are among the most important channel models for analyzing the performance of communication systems in information and communication theory. This importance stems from both their mathematical tractability and their ubiquitous appearance in practical applications. To date the deterministic identification capacity for additive Gaussian channels remains unknown even for the simplest case of the additive white Gaussian channel. In this paper, we establish tight bounds on the deterministic identification capacity of additive Gaussian channels by introducing a new perspective on deterministic identification. To this end, we apply results from lattice theory to obtain new capacity results.
Read the full abstract
Symplectic Geometry
A proof of the Arnold-Givental conjecture
We prove the Arnold-Givental conjecture in full generality: given a closed symplectic manifold $(X, \omega)$, an anti-symplectic involution $\tau_X: X \to X$ with fixed point set $L={\rm Fix}(\tau_X)$, and a Hamiltonian diffeomorphism $\phi: X \to X$ such that $\phi(L)$ intersects transversely with $L$, the following inequality holds: \[ \# \big( \phi(L) \cap L \big) \geq {\mathrm dim}_{{\mathbb F}_2} H_*(L; {\mathbb F}_2).\] The proof combines the methods of integral Floer theory of the first and fourth authors, a reduction to Hamiltonian Floer cohomology due to Lu, and a new idea related to localization in a $\mathbb Z/2$-equivariant Floer theory tailored to the problem.
Read the full abstract
Machine Learning
Importance Scoring of Transformer Attention Heads in Learning Tabular Data
Computationally demanding and opaque deep learning models can be better understood and optimized by analyzing how they transform data. While deep transformers have been widely studied in computer vision and natural language processing, their application in tabular data remains relatively underexplored. This paper presents one of the first applications of an importance-scoring metric to interpret multi-head transformer models in learning from tabular data. Experiments conducted on 40 diverse tabular datasets demonstrate robustness to head drops based on the proposed head importance score. In 72.5% of experimental examples, the model remains most resilient to performance drops when heads with the lowest importance scores are gradually removed. In contrast, removing the most important attention head first results in the greatest reduction in classification performance. A closer look at individual head importance scores across six attention layers reveals that important heads are scattered across layers, with no consistent layer-specific trends. In contrast to the image and language domains, the importance of individual attention heads varies considerably across tabular datasets with different schemas and feature spaces. The proposed importance score can improve efficiency and redundancy within transformer architectures. We make the source code for measuring the importance of individual attention heads publicly available.
Read the full abstract
Computer Vision
UniFLM: United Segmentation and Measurement on Fetal Limb Ultrasonic Image
Prenatal ultrasound examination is crucial for assessing fetal limb development and detecting congenital anomalies. However, existing artificial intelligence models often overlook fetal lethal skeletal dysplasias due to the lack of high-quality annotated data and a unified framework for multiple long bones. Moreover, generic segmentation models struggle with the inherent noise and semantic gaps in ultrasound images. To address these challenges, we construct the Fetal Limb Bones (FLB) dataset, comprising high-quality annotations for the humerus, femur, tibia-fibula, and radius-ulna. Furthermore, we propose UniFLM (United Segmentation and Measurement on Fetal Limb Ultrasonic Image), a unified framework for automatic cross-plane segmentation and measurement. UniFLM incorporates a Semantic Alignment Skip Connection (SASC) module to bridge the semantic gap between encoder and decoder features, and a Positive Sampling (PoSamp) strategy to filter noise and extract essential semantic information. Finally, a Point Regression Mapping (PRM) module is introduced to learn clinician annotation patterns for precise bone length measurement. Extensive experiments conducted on the FLB dataset and the public FetalP5 benchmark demonstrate that UniFLM achieves competitive performance with consistent generalization across four bone categories and external multi-center data, supported by comprehensive statistical validation including Bland-Altman agreement analysis and bootstrap confidence intervals. The source code is publicly available at https://github.com/chosen1203/UniFLM.
Read the full abstract
Human-Computer Interaction
An Offline Smartphone Check for Drug Interactions and Past Reactions in Handwritten Prescriptions: A Technical Feasibility Study in Rural Bangladeshi Outpatient Care
Rural Bangladeshi outpatient clinics keep no patient records, prescriptions are handwritten, and physicians have under a minute per patient, so they easily miss a known adverse drug reaction (ADR) or drug-drug interaction (DDI). We built and evaluated an offline Android application that photographs a handwritten prescription, detects and reads each medicine line, resolves brand names to active ingredients, and retrieves, but never predicts, documented interactions and the patient's recorded reactions. This quantitative technical-feasibility study, with qualitative error analysis, involved no patients or clinicians. Reference data combine two Bangladeshi drug registries (27,880 brands) with 190,421 DDI pairs. A YOLOv8n line detector reached mAP@0.5 = 0.687 on held-out prescriptions but recalled only 60.6% of lines at the deployed cutoff. On-device ML Kit read 11.8% of 780 benchmark handwritten words exactly. We trained a CRNN-CTC reader on public handwriting data, decoded it against the brand lexicon, and fine-tuned it on 660 author-labelled prescription crops. On benchmark words, the reader accepted the correct ingredient for 41.5% and a wrong one for 4.2%, versus 21.0% and 5.3% for ML Kit at its preset cutoff (McNemar p<10^-20). On scorable lines from held-out real prescriptions, the reader accepted the correct ingredient for 36.0% [26.8, 46.3] and a wrong one for 2.2% [0.6, 7.8]. In 62 simulated runs on a mid-range phone, median time per patient was 26-31 s and 59 of 62 runs finished within 60 s, but 62% of medicine lines still needed a manual check. A cross-visit medication history added after the evaluation passed seven automated tests and one two-visit functional test but was not evaluated. The pipeline is feasible within the clinic workflow; reading handwriting, not retrieval, is the bottleneck. Nothing here demonstrates a clinical benefit.