Theoretical Economics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Tuesday 10 November
We introduce a group version of non-obvious manipulability, which is stronger than the original definition by Troyan and Morrill (2020) and weaker than group strategy-proofness, and study its implications in Shapley-Scarf housing markets. We provide several characterizations of the Top Trading Cycles mechanism using the new condition.
Algebraic Geometry
A cyclotomic Artin-Tate formula and sphere packing for K3 surfaces
We give new necessary conditions on the Weil polynomial of a K3 surface over a finite field using techniques from the theory of sphere packing. By comparing the Artin-Tate formula over a finite extension, we derive a special values formula at roots of unity that involves the discriminant of the Frobenius coinvariant lattice. Known bounds on the minimum discriminants of even rootless lattices then imply nontrivial lower bounds on these special values. Applying these new conditions to Kedlaya and Sutherland's list of candidate Weil polynomials of K3 surfaces over the field with 2 elements rules out over 50% of the candidates.
Read the full abstract
Numerical Analysis
Weak convergence orders for a fully discrete approximation of semilinear stochastic PDEs driven by multiplicative Gaussian and Poisson noise
This paper investigates the weak convergence of a fully discrete approximation for a class of semilinear stochastic partial differential equations (SPDEs) driven by multiplicative Gaussian and Poisson noise. The scheme combines a spectral Galerkin discretization in space with an exponential Euler discretization in time. Compared with single-noise settings, the simultaneous presence of continuous and jump perturbations introduces additional challenges arising from discontinuous trajectories, nonlocal jump operators, and their interaction with state-dependent multiplicative coefficients. To overcome these difficulties, we establish the exponential integrability of the numerical solution processes and refined moment estimates in negative Sobolev norms for the associated first- and second-order variational processes. Combining these estimates with a discrete-time approximation argument and Wiener–Poisson Malliavin calculus, we derive regularity estimates for the solutions of the Kolmogorov equations associated with the spatial Galerkin approximations. Based on these estimates, we obtain the temporal and spatial weak convergence orders arbitrarily close to one with respect to discretization parameters $\tau$ and $\lambda_N^{-1}$, respectively. To the best of our knowledge, this provides the first weak convergence result for a fully discrete approximation of semilinear SPDEs with simultaneous multiplicative Gaussian and Poisson noise.
Read the full abstract
Populations and Evolution
A Hierarchical View of Ecological Encounters
Encounters between individuals, or between individuals and entities like vehicles, are central to understanding a host of topics in ecology and related fields. Problems of mate-finding, predator-prey interactions, pathogen spread, and animal interactions with infrastructure all hinge on encounters. Historically, however, encounters have been difficult to observe at scale, and the relevant theoretical frameworks—which ultimately derive from the ideal gas law—have been badly mismatched to the underlying biology. Here, we synthesize and provide ecological context for recent developments in mathematics, statistics, and physics that permit the inclusion of more biological realism into encounter-related models. We propose an encounters–interactions–outcomes hierarchical framework for organizing core concepts across a diverse range of ecological scenarios. Implemented as a hierarchical Bayesian model, our framework permits tight connections with multi-level ecological data relevant to the estimation of encounter rates and their downstream impacts on interactions and outcomes. Throughout, we emphasize how the ongoing revolution in animal biologging yields empirical datasets that can be employed to quantify otherwise inaccessible rate parameters and their dependencies. We conclude with a series of major open questions about encounters designed to spur conceptual and quantitative integration across models, data sources, and ecological levels of organization.
Read the full abstract
Quantum Physics
Real-time spectral combs from delayed quantum feedback in multimode cavity QED
We study the time-dependent physical spectrum of light for a two-level atom coupled to a large, but discrete, number of modes of an electromagnetic cavity. Radiation emitted by the atom, placed at the center of the cavity, returns after the round-trip time $t_r$, re-excites it, and is re-emitted, producing a sequence of atomic population revivals. For a coupling approximately flat over the atomic linewidth, the emission dynamics reduces to a delay-differential equation of Bixon–Jortner type. Using its analytical solution in terms of Laguerre polynomials, we investigate how successive emission and reabsorption cycles shape the time-dependent spectrum. The physical spectrum exhibits revivals during which a transient frequency comb of spacing $2\pi/t_r$ develops. The comb results from the quantum interference between successive emission stages; its teeth sit not at the cavity modes but at the resonances of the effective two half-cavities created by the resonantly reflecting atom. The visibility scales as $e^{-\Gamma t_r}$ with the filter bandwidth $\Gamma$, and at long times the comb converges to the dressed-mode spectrum, which we also obtain analytically in quantitative agreement with exact diagonalization. We show that the physical spectrum resolves the feedback phase, to which the atomic population is insensitive in the regime of well-separated emission stages, and provides additional frequency-resolved signatures of the atomic position. We discuss two feasible experimental implementations of our results with transmons (artificial atoms) coupled to microwave and surface-acoustic-wave resonators.
Read the full abstract
Statistical Methodology
AdaFCP: Local Neural Functional Representations for Industrial Change Detection
Industrial sensor measurements often form ordered functional profiles in which persistent operating-regime changes may be confined to short within-cycle regions. Fixed projections can dilute such changes, whereas data-adaptive representation learning complicates reliable statistical testing. This paper proposes adaptive functional change-point detection (AdaFCP) for retrospective detection of a single functional mean change. A compact window and neural basis learn a change-sensitive functional projection by matching the CUSUM path of projected scores to single-change templates. Standardized peak-CUSUM scanning then provides the detection statistic and transition estimate, while the fitted weights identify the contributing within-cycle region. Theoretical analysis characterizes the population path-matching signal and derives an optimal-direction signal-to-noise benchmark. Full-pipeline permutation calibration accounts for adaptive representation learning and restart selection, providing finite-sample false-positive control under exchangeability. A decision-preserving early-stopping variant, AdaFCP-ES, further reduces calibration effort without altering the fixed-budget test decision. Simulations demonstrate advantages over several fixed representations for localized oscillatory changes, while strong analytic and multiscale controls remain competitive. Applications to injection-molding pressure profiles and a nuisance-matched hydraulic benchmark illustrate transition localization and interpretable within-cycle diagnostics, while highlighting the need for dependence-aware calibration in naturally ordered industrial data.
Read the full abstract
Computational Physics
A NURBS-enhanced high-order finite-volume method for compressible laminar boundary layers on curved domains with hybrid meshes
High-fidelity simulation of compressible laminar boundary layers on curved domains requires physical wall boundary treatment and geometric representation compatible with high-order reconstruction, since viscous shear stresses and conductive heat fluxes are concentrated in a thin near-wall region. Conventional high-order methods often suffer from loss of accuracy near curved walls or require complicated wall treatments. To address these issues, in this paper, we develop a NURBS-enhanced $k$-exact finite-volume method for the steady compressible Navier–Stokes equations on hybrid meshes. To incorporate physical wall conditions into the high-order discretization, we develop a wall boundary treatment that employs sequential reconstruction without requiring the reconstruction variables to coincide with those used to prescribe the boundary conditions. The NURBS-based boundary representation is used to obtain a high-order accurate description of curved boundaries, and the diffusive generalized Riemann problem (dGRP) viscous flux is employed to enhance numerical robustness. A variety of numerical examples are provided to demonstrate the high-order accuracy and effectiveness of the proposed method.
Read the full abstract
General Relativity and Quantum Cosmology
The out-state of scalar-induced gravitational waves
Scalar-induced gravitational waves (SIGWs) are described by second-order tensor perturbations that contain free radiation together with a gauge-dependent response maintained by the scalar source. For a perfect fluid with constant equation of state $0<w<1$, we identify the free radiation with the out-state of the tensor perturbation, assembled from the light-cone residue of kernel functions for each pair of scalar modes. Its amplitudes follow from a classical LSZ-type reduction formula, in terms of the initial data and the light-cone-projected source. We prove a classical equivalence theorem: a redefinition of the tensor perturbation whose late-time shift has power-law envelopes and no tensor-frequency component leaves these amplitudes unchanged. Transformations between intrinsic gauges, whose slice-normal observers and spatial coordinates are fixed by the matter and geometry without an external clock or independent oscillatory dynamics, are redefinitions of this kind, except on resonance surfaces of zero measure. The out-state is therefore the same in all intrinsic gauges, including the standard gauges of cosmological perturbation theory. The late-time Newtonian kernel is then the out-state kernel, which justifies its use for the energy spectrum of the free radiation. In matter domination, the out-state instead separates the free radiation from the static tidal response.
Read the full abstract
Functional Analysis
A proof of the constant-width numerical range conjecture
We prove that the numerical range of a complex matrix has constant width \(d\) if and only if it is a closed disk of radius \(d/2\). The width in each direction is the gap between the extreme eigenvalues of a Hermitian matrix. When that gap is constant, the midpoint of the projected interval is a rational function of the direction parameter. An operator norm bound and reversal of direction then show that the midpoints are projections of a single point, the disk's center.
Read the full abstract
Statistical Methodology
Sufficient Dimension Reduction with Multimodal Predictor Objects
We study the regression of a scalar response $Y$ on a Euclidean predictor $\mathbf{X} \in \mathbb{R}^p$ and a random predictor object $\mathcal{Z}$ in a general metric space under arbitrary structural dependence. We introduce a joint sufficient dimension reduction $(\mathbf{B}^\top\mathbf{X}, \mathcal{G}_Z)$ constructed via a sequential conditional independence framework, which differs fundamentally from the ordinary central subspace of $Y \mid \mathbf{X}$. Methodologically, $\mathcal{Z}$ is embedded into a reproducing kernel Hilbert space, and a regularized kernel-response envelope model extracts a finite-dimensional reduction of the feature map. Partial adaptive estimation then recovers the Euclidean reduction conditional on the first-stage coordinates. We establish root-$n$ consistency rates for the first-stage envelope subspace under a shrinking regularizer, alongside large-sample consistency for the second-stage Euclidean subspace estimator. Simulations and two real-data applications illustrate the utility and flexibility of the proposed method.
Read the full abstract
Number Theory
Bounding the security of the Bromberg-Shpilrain-Vdovina hash function
The Bromberg-Shpilrain-Vdovina hash functions are a family of Cayley hash functions indexed by an integer parameter $k\ge 2$. Hashes are matrices in $SL_2(p)$ where $p$ is a large prime. We present an attack which finds collisions of length $O(\log p)$ whose average-case time complexity is $O(p^{0.28})$ when $k=2$ and $O(p^{0.36})$ when $k=3$ under an assumption in the spirit of Zaremba's Conjecture on continued fractions. This is an improvement on the more general attacks of Mullan-Tsaban and Slaughter which have average-case time complexity $O(\sqrt{p})$. For any $k$, the average-case time complexity of our attack is $O(p^{1-\delta_k+o(1)})$ where $\delta_k$ is the Hausdorff dimension of a certain Cantor set. This tends to $O(\sqrt{p})$ as $k\to\infty$, so our attack has time complexity comparable to Mullan-Tsaban and Slaughter's attacks as $k$ grows. We give examples of collisions for large primes found by our attack when $k=2, 3$.
Read the full abstract
Geometric Topology
Surface subgroups with prescribed finite quotients
We prove that every finite quotient of a finite covolume Kleinian group $G$ can be realized by a surface subgroup. More precisely, given an epimorphism from $G$ onto a finite group $F$, there exists a surface subgroup that maps surjectively onto $F$. As a consequence, the union of the surface subgroups is dense in $G$ with respect to the profinite topology. We also apply this result to obtain more examples of atoroidal surface bundles with signature zero in the construction of Kent–Leininger.
Read the full abstract
Machine Learning
Deep Personalization: Continuous Context Observation for Local Proactive Agents
As AI agents become increasingly capable, users must still recognize when and how to ask for help, leaving many opportunities for useful support unrealized. Proactive agents hold the promise to close this gap, but effective proactivity requires understanding users' real-time activities and personalizing intervention timing so that assistance is offered when it is likely to be helpful rather than disruptive. We introduce deep personalization, a closed-loop process that connects continuous context observation with continual agent adaptation: an agent observes user activity, selectively retains long-term context, and adapts intervention decisions from implicit feedback. We initiate deep personalization in Coco, a proactive desktop agent, by having a separate Observer Agent that monitors desktop activity, maintains compact long-term user modeling context, and learns when assistance is desirable from implicit signals such as engagement, dismissal, explicit requests, and repeated work patterns. We further develop efficient on-device strategies to enable the Observer Agent to run on local AI compute platforms without streaming raw activity data off-device. For evaluation at scale without collecting sensitive real-user activity, we introduce a data-synthesis pipeline that transforms existing computer-use trajectories into OSWorld-Persona, a benchmark with persona-conditioned intervention labels. On OSWorld-Persona, deep personalization improves macro F1 across three model backbones from 0.12-0.19 under direct prompting to 0.35-0.48 after adaptation, a 2.4-3.0x relative gain. In a week-long user study, we observe that Coco improves its macro-averaged F1 score from 0.199 to 0.515. These results suggest that on-device deep personalization can make proactive agents more timely and increasingly aligned with each user's everyday work.
Read the full abstract
Artificial Intelligence
When Reasoning looks like Hidden Understanding: How Chain-of-Thought Fosters Anthropomorphism, Perceptions of Epistemic Superiority, and Cognitive Deference
As AI-based products permeate society, specifically systems using generative models, a belief system is emerging about what these products are and what they can and cannot do. One common belief concerns the perception of the form of language as evidence of 'understanding' language. As these generative models reproduce linguistic patterns rather than understand meaning, interpreting these patterns as evidence of understanding is a misattribution. This misconception is extended by the notion that generative models can 'explain their reasoning.' Anecdotally, these misconceptions which mistake the form of language for its meaning are driving cognitive deference to AI systems in a way that displaces human agency. Such mistakes involve attributions of cognitive properties to the system, allowing them to be treated like intentional 'minds'. Importantly, perceived epistemic authority and superiority make that 'mind' worth deferring to. Anthropomorphism in tandem with epistemic authority and superiority, we argue, increases the risk that users generalise from narrow task performance to broader competencies, committing the "fallacy of composition". This paper explores misconceptions of Chain of Thought (CoT) reasoning as sufficient transparent 'explanation'. We explore how anthropomorphism acts as a vehicle for cognitive deference of decision-making and perceptions of epistemic superiority in AI systems. Finally, we explore CoT's ramifications using medicine as an example. Taken together, we conclude that explainability operationalised through CoT is highly problematic and undercuts the very ethical principles it is intended to serve. While CoT may be valuable for improving output accuracy and moderation, we suggest that the practice of asking models to provide user facing reasons for their outputs should be actively discouraged, and AI literacy training should explicitly caution against its use.
Read the full abstract
Combinatorics
Divisibility graphs of semigroups as weighted poset blow-ups
We develop a structural theory of divisibility graphs of arbitrary semigroups. For a semigroup $S$, let $S/{\mathcal J}$ be the poset of Green's ${\mathcal J}$-classes ordered by inclusion of principal ideals,and equip each $\mathcal {J}$-class $J$ with the weight $|J|$. We prove that the divisibility graph ${\operatorname{Div}}(S)$ is obtained from the comparability directed graph $Comp(S/{\mathcal J})$ by replacing each $\mathcal J$-class with a complete symmetric graph of order $|J|$ and by replacing each directed edge $J_u$ to $J_v$ in $Comp(S/{\mathcal J})$ with all directed edges from every element of $J_u$ to every element of $J_v$. Consequently, ${\operatorname{Div}}(S)$ is determined, up to isomorphism, by the weighted poset $(S/\mathcal {J},\{ J \}_{J\in S/{\mathcal J}})$. This yields structural descriptions of its strongly connected components, maximal unilaterally connected subgraphs. The main contribution is the reduction of divisibility graph problems to weighted quotient-poset problems. As applications, we obtain reconstruction and isomorphism criteria and prove an inheritance theorem for divisibility $D$-saturated subsemigroups.
Read the full abstract
Artificial Intelligence
SafeRoute-MAS: Coverage-Aware and Risk-Calibrated Message Routing for Multi-Agent Scientific Reasoning
We introduce SafeRoute-MAS, a coverage-aware, risk-calibrated procedure that keeps all specialist messages or removes one using only pre-integration information. On a sealed cache of 5,328 committees, its source-compatible operating point gains +1.80 percentage points (95% CI [+1.13, +2.52]) while removing 12.76% of message tokens. We also run 4,580 controlled integrations over a stratified 458-committee sample on a public gpt-oss-120b endpoint. The sample estimate is -0.09 points; design weighting to the original workload gives +0.94 points (95% CI [-1.31, +3.50]) and 18.72% acted-input reduction. A weighted token-matched comparison likewise does not establish a selector advantage (+1.68 points, CI [-1.27, +5.10]). Only 13 of 88 cached non-neutral labels (14.8%) retain direction by repeated-call majority. Expected-corruption UCL is 4.60% after weighting, but only 2 of 11 prespecified marginals pass the support/risk rule. We therefore propose a numerical qualification with Keep-All fallback, applied here retrospectively rather than validated prospectively.
Read the full abstract
Computer Vision
Camera3R: Feed-Forward 3D Reconstruction from Heterogeneous Camera Models
Feed-forward 3D reconstruction (3R) from unconstrained multi-view images has recently emerged as a prominent research direction in 3D computer vision. Various foundational reconstruction models have been proposed with different architectures and reconstruction paradigms. Despite being trained on large-scale perspective datasets with substantial computational resources, these models generally struggle to generalize to non-perspective images. This limitation hinders their deployment in applications involving arbitrary, unknown, and potentially heterogeneous camera models. To build on the capabilities of existing reconstruction models with minimal retraining, we propose Camera3R, a low-rank adaptation framework that efficiently extends perspective-pretrained 3R networks to any other camera models. Extensive experiments demonstrate that, with only 0.5% additional parameters and less than a day of training, Camera3R adapts a foundational 3D reconstruction model to non-perspective cameras while fully preserving its original performance on perspective inputs. Code is available at https://github.com/android-xr/camera3r.
Read the full abstract
Machine Learning
SLIC: Reinforcement Fine-Tuning Small LMs for Multi-Turn Analog Circuit Optimization
Analog integrated circuit (IC) design is a multi-turn, simulator-grounded engineering task that demands iterative reasoning over coupled performance tradeoffs, yet proprietary process design kits (PDKs) make hosted large language model workflows difficult to deploy in practical design settings. We present SLIC (Small LM Iterating on Circuits), a reinforcement fine-tuning framework that trains a 4B-scale, locally deployable language model for closed-loop analog circuit optimization. SLIC pairs structured circuit representations with discretized relative-update actions, domain- and success-gated multiobjective rewards, and Reward-Guided Rollout Steering (RGRS), which turns per-step simulator feedback into signals for both policy learning and online rollout construction in this sparse, highly constrained domain. We further release an open, goal-conditioned benchmark of 41 amplifier topologies with standardized protocols for in-distribution validation, held-out topology evaluation, and cross-technology-node transfer. Trained solely with open source 65 nm CMOS predictive technology models (PTM), SLIC-Random achieves 87.1% in-distribution four-specification Pass@1 within eight turns, surpassing the 763B DeepSeek-V4.1-Flash Think by 48.5 percentage points. Beyond the training distribution, SLIC-Softmax reaches 72.6% Pass@1 on held-out topologies, and the trained policies consistently transfer across five unseen technology nodes without additional fine-tuning. This transfer extends to a commercial TSMC 65 nm CMOS PDK, where SLIC-Random achieves 49.9% held-out-topology Pass@1, demonstrating practical deployment through local inference without transmitting proprietary data to external LLM services. Performance further scales with test-time simulator interactions, reaching up to 83.2% held-out-topology Pass@1 at a 32-turn budget.
Read the full abstract
Statistical Methodology
Economics of Selection in Ranking & Selection: Neither Fixed-Precision nor Fixed-Budget
Ranking and selection (R&S) has evolved tremendously over the last several decades, with applications in agriculture, discrete optimization with computer simulation, clinical trial design, and other areas. We challenge a widely accepted classification of R&S procedures into two categories: fixed-precision and fixed-budget. We discuss a third class, economics-of-selection (EoS) procedures, which is conceptually distinct and is found in the extant literature on Bayesian hypothesis testing and simulation optimization. Procedures in this third class seek to optimally balance two quantities: the cost of learning about the performance of each alternative and the quality of the alternative ultimately selected on the basis of that learning. Fixed-precision and fixed-budget procedures tend to optimize one of these quantities while constraining the other. We clarify these differences and show why they matter. We then present new frequentist R&S procedures in this third category, including both two-stage procedures and a sequential heuristic, assuming Gaussian observations with either known or unknown variances. We find that EoS is not a question of Bayesian versus frequentist analysis, asymptotic versus finite-sample analysis, finite-stage versus fully sequential procedures, or comparing one mean with a standard versus comparing multiple alternatives. We identify opportunities for further R&S research.
Read the full abstract
Theoretical Economics
Feedback Design in the Shadow of Strategic Communication
We study optimal feedback design when evaluations affect subsequent strategic communication within organizations. A firm designs public feedback about an employee's potential before the employee communicates private information about project-specific circumstances to a manager. While more informative feedback helps the manager make better decisions, it can also intensify preference conflicts and undermine the credibility of subsequent communication. We characterize the optimal feedback policy. First, full disclosure is either optimal for all prior beliefs or strictly suboptimal for all of them. In particular, it is optimal when preference conflicts are sufficiently small or sufficiently large. Second, when full disclosure is suboptimal, no disclosure is optimal for some priors, while partial disclosure is strictly optimal for some priors if and only if the number of messages in equilibrium communication varies with posterior beliefs. Third, optimal partial disclosure can conclusively identify low-potential employees when high potential is rare, but high-potential employees when it is common.
Read the full abstract
Optimization and Control
Complete Decentralization of Wave Equation Control on a Compact Domain
A feedback control policy is completely decentralized if the actuation signal at each location in space can be computed based only on state information from that location. For systems equipped with large numbers of spatially distributed sensors and actuators, completely decentralized controllers are desirable since they can be implemented in a communication and computation efficient manner, but generally perform worse than centralized counterparts. However, through a case study of Linear Quadratic Regulator (LQR) design for wave dynamics on a compact spatial domain, we show that careful design of system parameters to satisfy certain conditions causes a decentralized control policy to perform as well as any centralized policy. We then propose a spatially filtered $H_2$ norm to show that satisfying these conditions does not inherently degrade achievable closed-loop system performance, enabling simultaneous realization of both desirable performance and communication/computation efficiency via system-control co-design. Additional robustness analysis characterizes a range of system parameters such that the system interconnected with a decentralized controller maintains stability and near-optimal performance. Although our framework is derived for spatially continuous dynamics, we demonstrate that this theory can be leveraged to derive results for decentralization in finite-dimensional settings through a damped mass-spring example.
Read the full abstract
Machine Learning
Fermat ProbCover: Coverage-Based Active Learning on Fermat Distance Graphs
Active learning aims to reduce annotation cost by selecting the most informative examples for labeling. In the low-budget regime, where uncertainty-based methods often fail, coverage-based active learning methods such as ProbCover have been shown to be effective. However, these methods typically rely on Euclidean distance, which may not adequately capture the underlying geometry of data. In this article, we propose PC-Fermat, a coverage-based active learning method that replaces Euclidean coverage neighborhoods with neighborhoods defined by a Fermat distance, i.e., a data-adaptive distance that reflects the underlying distribution of the data points. To facilitate computation, we construct a weighted Euclidean neighborhood graph and use a truncated shortest-path algorithm to compute the Fermat coverage graph. We also establish sufficient conditions under which this construction exactly preserves the Fermat coverage computed on the complete graph. Finally, we evaluate PC-Fermat on eight classification datasets across varying labeling budgets, with particular emphasis on the very low-budget regime. Our experimental results demonstrate that PC-Fermat outperforms several existing active learning methods on the majority of the datasets.
Read the full abstract
Quantum Physics
Double-quantum enhanced coherent control and magnetic sensing of spin defects in hexagonal boron nitride
The negatively charged boron vacancy ($V_{\mathrm{B}}^{-}$) in hexagonal boron nitride (hBN) is a promising spin defect for quantum sensing in two-dimensional material systems. Previous hBN experiments have mainly exploited the two spin-triplet ground-state single-quantum transitions separately. Here we implement a double-quantum driving scheme based on simultaneous resonant microwave excitation of both ground-state single-quantum transitions, establishing coherent control of the $\Delta m_s=2$ manifold in a $V_{\mathrm{B}}^{-}$ ensemble. We compare the resulting double-quantum response with the corresponding single-quantum protocol under identical experimental conditions. The double-quantum protocol enhances the optical readout contrast and yields a 2.6-fold signal-to-noise improvement in AC magnetometry. We further use the double-quantum relaxation channel as a differential probe of electric-field and strain noise, showing that the present data are compatible with correlated-surface or volume-distributed fluctuations, while short-range uncorrelated surface-noise models are disfavored within the assumptions of our analysis. These results establish double-quantum coherent control and frequency-resolved AC magnetometry in hBN spin ensembles, and identify the sample-dependent measurements needed to determine the microscopic origin of the dominant relaxation channels.
Read the full abstract
Representation Theory
Differential Monodromy, $q$-Opers, and Universal $(G,q)$-Wronskians
We give a geometric formulation of the $q$DE/ODE/IM correspondence for families of differential operators associated with simple Lie algebras of ADE type. Under analytic hypotheses, the $q$-connection associated with shifted monodromy defines a twisted $(G,q)$-oper and has Steinberg form in the frame of solutions normalized by prescribed asymptotics in a sector at infinity. We identify the central connection between the normalized solution frames at zero and infinity with a universal meromorphic $(G,q)$-Wronskian. Its generalized minors give $Q$-functions satisfying the ADE $QQ$-relations. When $q$ is not a root of unity, we parametrize shifted monodromy by quantum-affine $q$-characters evaluated on shifted $Q$-ratios, using the same character formulas that describe normalized transfer-matrix eigenvalues. We verify the analytic hypotheses for generic untwisted Feigin–Frenkel–Hernandez affine opers, thereby relating these affine opers to twisted $(G,q)$-opers.
Read the full abstract
Representation Theory
G-radical-square-zero algebras
In this paper, we first provide necessary and sufficient conditions for a Krull-Schmidt triangulated category with Auslander-Reiten triangles to have square-zero radical. One of these conditions is that the middle term of every Auslander-Reiten triangle is either zero or indecomposable. We then call an artin algebra G-radical-square-zero if the stable category of Gorenstein-projective modules has square-zero radical. We characterize when a G-radical-square-zero algebra is CM-finite and, in this case, give an additive generator of the stable category. For certain Iwanaga-Gorenstein G-radical-square-zero algebras, we further show that the stable category admits an additive generator consisting of higher syzygies of simple modules.
Read the full abstract
High Energy Physics (Theory)
Giant Graviton Correlators: From Non-maximal to Dual, Bootstrap and Hidden Symmetry
We compute four-point functions of two light $\frac{1}{2}$-BPS operators and two heavy giant graviton operators in $\mathcal{N}=4$ super Yang–Mills theory at large $N$ and infinite 't Hooft coupling. Using bootstrap techniques, we determine the correlators for arbitrary light-operator dimensions and arbitrary allowed sizes of non-maximal sphere and dual giant gravitons. We show that these correlators exhibit a hidden ten-dimensional conformal symmetry, allowing them to be organized into compact generating functions. Remarkably, the generating functions for both types of giant gravitons can be obtained from the maximal giant case through $SO(2,10)$ rotations in the embedding space.
Read the full abstract
Analysis of PDEs
Regularity Estimates for a Nonhomogeneous Free Boundary Interface Problem
In this work, we establish regularity properties for minimizers of a novel nonhomogeneous interface problem with a free boundary, formulated in Orlicz-Sobolev spaces. The nonlinear terms fulfill Lieberman-type structural assumptions with subcritical growth. The principal difficulties stem from the nonhomogeneous framework and the lack of suitable representation formulas. Our findings extend recent contributions in the field and provide new insights even in the particular case of the Laplacian. Keywords: Free Boundary Problems; Orlicz-Spaces; Nonhomogeneous Operator; Transmission Problem; Regularity.
Read the full abstract
Distributed Computing
ForeSpark: Joint Draft and Verification Budget Scheduling for Parallel Speculative Decoding
Parallel drafters accelerate LLM decoding by proposing a whole block of draft tokens in a single forward pass, making them attractive for LLM serving. Recent systems optimize goodput (i.e., accepted tokens per second) because rejected drafts steal compute from other requests. However, these systems control only how many tokens are verified, so every request still drafts a full block. At high serving load, we find that drafting consumes up to 40% of step time, of which a majority is discarded because it is never verified. Yet, draft-length control from autoregressive drafters does not carry over, because a parallel drafter commits to its length before its single pass. Moreover, draft and verification lengths are coupled across the batch, so they must be selected jointly for every request while keeping the decision computationally cheap at every step. In this paper, we propose work-conserving speculation (WCS), a design that jointly allocates draft and verification budgets per request in LLM serving, and present ForeSpark as its prototype. ForeSpark realizes WCS with (i) a context-aware pre-draft estimator that predicts acceptance for every candidate length from the target model's context features, (ii) a lightweight joint draft-verify scheduler that selects both lengths through a global sort, and (iii) a post-draft budget distribution that spends verification on the most promising tokens. Implemented on vLLM, ForeSpark improves throughput by up to 1.77x over the state-of-the-art DSpark and by 2.49x over autoregressive decoding at a matched per-request generation rate, without changing the output distribution. It also tolerates higher load, sustaining up to 1.51x the request rate of DSpark under latency service level objectives (SLOs).
Read the full abstract
Computation and Language
Inherited Weights: Measuring the Tokenization Premium and the Sovereignty Cost of Language Models for the Global South
Most language models deployed outside the United States and China are fine-tuned descendants of a small number of pretrained families. We argue that fine-tuning leaves the decisive properties of a model inherited from its base: vocabulary, knowledge, cultural alignment, memorized data and license. We contribute (i) an updated, reproducible measurement of the tokenization premium across 31 languages and 7 tokenizers released in 2023-2024, on 27 articles of the Universal Declaration of Human Rights, with explicit correction for unknown-token collapse. Premiums reach 10x for Amharic in older tokenizers and remain at 6.8x in the newest commercial one; Indigenous South American languages pay between 1.6x and 2.7x under global tokenizers. A vocabulary built from Portuguese lowers the premium for 11 of 13 Latin-script Global South languages relative to every global tokenizer tested, by about 24% for Guarani and Quechua. (ii) An eight-test black-box sovereignty audit that a buyer can run without access to weights. (iii) A regime argument: under energy and liability constraints, accuracy per joule, calibration and selective risk should replace raw accuracy as design objectives, with metrics that abstention cannot inflate. We close with preliminary internal measurements from a proprietary architecture built around these objectives, reported with their limitations. Data and measurement code are included.
Read the full abstract
Geometric Topology