Probability
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 14 September
We establish a stochastic control representation for $L^p$-norms on Wiener space. For every $p\ge1$ and every non-negative universally measurable functional $\varphi$, we show that $$\|\varphi(W)\|_p = \sup_a \mathbb{E}\Big[ e^{-\frac12\int_0^T\|a_t\|^2\,dt} \varphi\Big( W+\sqrt{p-1}\int_0^\cdot a_t\,dt \Big) \Big],$$ where $W$ is a $d$-dimensional Brownian motion and the supremum is taken over all progressively measurable controls $a$ satisfying $\int_0^T\|a_t\|^2\,dt<\infty$ almost surely. This identity can be viewed as a multiplicative $L^p$-analogue of the Boué–Dupuis variational formula. We also provide an extension to strong solutions of stochastic differential equations with path-dependent coefficients. Building on the variational formula, we discuss finite-$p$ estimates for small-noise limits of stochastic differential equations with path-dependent coefficients. These include quantitative Freidlin-Wentzell bounds, a functional Gaussian approximation and concentration estimates, and moderate-deviation bounds. Moreover, we discuss consequences for functionals determined up to a stopping time, obtaining in particular probability bounds that identify the sharp exponential decay rate after optimization over controls. The proof of the variational formula combines probabilistic arguments with convex duality methods, viscosity theory for Hamilton-Jacobi-Bellman equations, and a dynamic programming principle.
General Relativity and Quantum Cosmology
Analysis of GWTC-3 with multiple quasicircular models
The interpretation of gravitational wave sources depends on the specific choice of model used to interpret signals. Previous analyses of GWTC-3 event candidates using SEOBNRv4PHM, IMRPhenomXPHM, and NRSur7dq4 have sometimes arrived at notably different conclusions about event properties. Subtle analysis settings and code differences can also produce notable differences. We revisit 60 publicly available events from the first three observing runs, released across GWTC-1, GWTC-2, GWTC-2.1, and GWTC-3, using a consistent analysis framework and three waveform models: two state-of-the-art time-domain models with higher-order modes (SEOBNRv5PHM and IMRPhenomTPHM) and one older frequency-domain model without higher-order modes (IMRPhenomPv2). The two state-of-the-art models agree well overall, although about 20% of events have a visible difference in at least one one-dimensional marginal posterior. By contrast, the older model often arrives at qualitatively different conclusions for key events across the mass spectrum. We also identify a bug in published LVK GWTC-2.1 SEOBNRv4PHM results and in a later reprocessing of GW200105. At the adopted sampling rates, the model's Nyquist-frequency restriction excluded part of the allowed parameter space and artificially truncated the mass-ratio posterior away from equal mass for seven low-mass events. These results emphasize the need to use multiple state-of-the-art waveform models and carefully validated analysis settings to characterize uncertainty in gravitational-wave source properties.
Read the full abstract
Functional Analysis
Function spaces for weighted subcoercive operators
We develop a theory of Sobolev, Besov and Triebel–Lizorkin spaces associated with weighted subcoercive operators on any real connected Lie group.
Machine Learning
Sharp Regret Bounds and a Task-Covariance Correction for Spectral Representation Learning
Spectral contrastive learning keeps the $k$ directions along which two views of the data depend most strongly, while downstream tasks may need other directions. Existing guarantees for spectral features assume that the distribution of tasks has no preferred direction. We show that this isotropy condition is the wrong criterion. With linear probing, the task distribution affects the average probing gain only through the matrix $B\Lambda B^\top$, where $B$ records the dependence between views and $\Lambda$ is the task covariance: optimal features span its top-$k$ eigenspace, whereas spectral features span that of $BB^\top$. Transfer therefore depends on alignment, and strongly uneven task preferences cost nothing when they favour the spectral directions. Otherwise the regret of spectral features is at most linear in the anisotropy of $\Lambda$. No representation chosen without task information improves this rate when the leading spectrum is flat, and a spectral gap makes it quadratic. With labelled tasks, the same matrix gives a correction: Fisher discriminant directions for a frozen encoder, or a one-term change to the spectral contrastive loss. We bound the regret of the correction by the error in estimating $B\Lambda B^\top$, and compare this error with the misalignment of spectral features to decide when to correct. The analysis extends to finite samples, continuous variables and noisy views. Correction reduces synthetic held-out regret from $0.858$ to $0.003$ with $2000$ labelled tasks, and the decision rule gives the lowest regret among the tested rules at every bank size. On CIFAR-100, spectral features of frozen encoders keep most of the task signal on unseen classes, and where a bank of other classes would lower accuracy, the rule keeps them in every case.
Read the full abstract
Cryptography and Security
First Galileo SAS Authenticated Time Solution
Spoofing attacks against civilian GNSS receivers have grown more common, especially near conflict zones where they now disrupt civil aviation, maritime operations, and critical infrastructure on a daily basis. Spoofing is possible because legacy civil GNSS signals are largely predictable in both their navigation data and ranging codes, allowing an attacker to forge a signal that imposes a false position and time on an unsuspecting receiver. Cryptographic authentication schemes such as Galileo's Open Service Navigation Message Authentication (OSNMA) mitigate this threat by verifying the authenticity of the navigation data. The ranging code itself, however, remains unprotected. To close this gap, Galileo is introducing a Signal Authentication Service (SAS) in the E6-C signal, which directly authenticates ranging measurements. SAS is currently transmitted by only two satellites in an elliptical orbital plane, of which at most one is visible at a time, meaning a full position solution is not yet possible; however, a georeferenced receiver can still obtain an authenticated time solution. This paper presents, to the authors' knowledge, for the first time, a timing solution computed from an authenticated civil GNSS signal. We develop a snapshot software receiver implementing a simplified version of the Galileo SAS protocol to compute the receiver clock bias from an authenticated pseudorange, using radio-frequency data recorded with an engineering prototype software-defined radio receiver from Septentrio. We evaluate the resulting timing solution using recordings from both SAS-capable satellites collected at different locations, demonstrating the feasibility of authenticated timing ahead of full SAS operational deployment.
Read the full abstract
Functional Analysis
Fourier rearrangements on discrete groups: $L^p$ and operator-norm subsequence convergence
Let $\Gamma$ be an infinite finitely generated group that is hyperbolic or of polynomial growth. For $f$ belonging to the reduced group $C^*$ algebra $C_r^*(\Gamma)$, we construct a rearrangement of its individual Fourier terms with a subsequence converging to $f$ in the operator norm. For every $2\le p<\infty$ and $f$ belonging to the noncommutative $L^p$ space $ L^p(\hat \Gamma)$, we obtain convergence in the $L^p$ norm. The proof combines uniformly bounded finite-support Fourier cutoffs with an elementary Bernoulli even-moment estimate. We give a self-contained free-group argument, then establish a general criterion and verify it for the two classes above. The operator-norm conclusion establishes the Révész property for these groups. In particular, it confirms the polynomial-growth conjecture of Hamm, Hayes, and Petrosyan and disproves their conjecture that the free group $\F$ does not have this property.
Read the full abstract
Numerical Analysis
Regularization and compression of integral operator kernels with tensor paraproducts
We present a new perspective on integral operator kernels given by a nonlinear function of the distance. This perspective is useful since it permits one to quasilinearize the nonlinear transformations acting on the tensor Haar expansion of the distance, d(x,y). This leads to a new representation comprised of a principal and residual term with desirable features; namely, enhanced local regularity of the principal component (and in certain situations rapid off-diagonal decay leading to O(N) storage) and a residual expansion of improved approximation as a consequence of enhanced global regularity. Numerical experiments on the potential kernel are included
Read the full abstract
Analysis of PDEs
Sharp gradient bounds for uniformly elliptic PDE, eigenvalue asymptotics, and the Landis conjecture
We study three classical problems in the theory of linear second-order uniformly elliptic equations: (i) gradient estimates for the Dirichlet problem, (ii) estimates and asymptotics for the first eigenvalue of an elliptic operator, and (iii) the Landis conjecture on exponential decay. Our main results give versions of (i) in which the constant is sharply specified in terms of the norms of the coefficients of the operator and the size of the domain; and use these to strongly improve on known results for (ii) and (iii) by allowing both more general operators and weaker regularity assumptions on the coefficients, and by giving quantitative estimates. The proofs use a unified approach, relying on three ingredients: interior and boundary Harnack inequalities with sharp constants, duality arguments, and a $C^1$ estimate based on a rescaling procedure. The sharpness of the gradient and spectral estimates is demonstrated through various counterexamples.
Read the full abstract
Artificial Intelligence
AlgoEvo: Self-Evolving Agentic Search for Automated Algorithm Discovery
Large language models have advanced automated algorithm discovery by synthesizing executable code, but existing frameworks trap them in rigid search pipelines with pre-defined control flows. This limitation restricts adaptive reasoning, blocks cross-paradigm transfer, and overlooks richer execution feedback. To bridge this gap, we introduce an end-to-end framework, AlgoEvo, a unified agentic architecture that transforms automated algorithm discovery into an interactive, knowledge-accumulating process. An autonomous agent dynamically inspects, diagnoses, and edits code based on runtime feedback. A design skill hub decouples paradigm-specific knowledge from the core discovery engine, allowing a unified workflow to seamlessly handle single-heuristic, multi-objective, and multi-component design. Meanwhile, a hierarchical experience bank organizes search trajectories into a task-level tree to guide exploration and consolidates cross-task patterns into reusable skills. Across six representative benchmark tasks, AlgoEvo reaches state-of-the-art performance with as little as 7% of the evaluation budget and reduced token consumption, demonstrating strong intra-task accumulation, cross-task transfer, and the ability to reproduce or exceed the strongest existing methods through flexible skill activation.
Read the full abstract
Systems and Control
Hydrogen-Diesel Dual-Fuel Engine Operation Using Real-Time GRU-Based Nonlinear Model Predictive Control
Hydrogen-diesel dual-fuel (H2DF) combustion reduces combustion-out CO2 emissions but exhibits nonlinear cycle-to-cycle dynamics at high hydrogen energy shares (HES). This work evaluates nonlinear model predictive control (NMPC) with a gated recurrent-unit deep neural network dynamics model for transient H2DF control. Trained on 99,800 engine cycles, the model predicts indicated mean effective pressure, nitrogen oxides (NOx), particulate matter (PM), and maximum pressure-rise rate. Single-cylinder Cummins 4.5 L experiments follow an unseen 4,900-engine-cycle trajectory. Compared with production diesel-only control, NMPC improves load-tracking mean absolute error by 27.8% and reduces mean PM by 61.1%, while mean engine-out NOx increases by 105.1% without exhaust-gas recirculation. Mean and peak HES reach 39.7% and 55.1%; a high-hydrogen setting achieves 77.8% peak HES without constraint violations. Robust to feedback noise and model-plant mismatch and executing in 3 to 7 ms per engine cycle on low-cost embedded hardware, learned-dynamics NMPC enables practical, real-time, constraint-aware transient H2DF control with substantial diesel substitution.
Read the full abstract
Artificial Intelligence
Atria Dawn: The Dawn of Agentic Superintelligence
As AI agents become participants in the development of their successors, they reshape both the production of intelligence and the role of human researchers. We introduce Atria Dawn Preview, a foundation agentic language model designed for scientific research and engineering workflows, with the goal of expanding the frontier of agent productivity in the real world. This model is trained via a Verifiable Experience Pipeline that connects tool-mediated interactions to executable environments and externally verified outcomes. Across 16 benchmarks spanning real-world research, engineering, and digital work, Atria Dawn Preview is competitive with frontier agents and achieves the highest reported score on five of them. Beyond standalone performance, we examine the real research-and-development process behind this model as a case study of human–AI collaboration, analyzing 769 task records from 56 participants together with agent logs. When asked to evaluate completed tasks under comparable conditions, participants rated about one-third of completed AI-assisted tasks as infeasible without AI. More strikingly, agents frequently propose methods and implement revisions, while humans retain most final decisions and guide exploration through judgment and feedback. These observations indicate a shift from task-level execution to project-level partnership, with human effort concentrating on what is worth pursuing and how evidence should guide research. Progress toward more autonomous AI research must therefore advance both the capacity for discovery and the capacity for meaningful human oversight, preserving accountable human authority over the risks and direction of continued development.
Read the full abstract
Functional Analysis
On Property (H) of c_0 and Lipschitz Problems of Gromov and Johnson
Let $X$ be a normed space and $S_X$ be the unit sphere of $X$. We say that $X$ has Property (H) if there exist two increasing sequences $V_j\subset X$ and $H_j\subset \ell_2$, $j=1,2,\ldots$, of finite dimensional subspaces such that $V=\bigcup_j V_j$ is dense in $X$, and there is a uniformly continuous mapping $\psi:S_V\to S_{\ell_2}$ such that the restrictions $\psi|_{S_{V_j}}$, $j=1,2,\ldots$, are isomorphisms from $S_{V_j}$ onto $S_{H_j}$. The Kasparov-Yu problem asks whether the Banach space $c_0$ of all null sequences admits Property (H), or rational Property (H). In this paper, we prove that $c_0$ admits neither Property (H) nor rational Property (H). We also obtain the following results. (1) The infimum $\mu_n$ of the Lipschitz constants of all nonzero-degree maps from $S_{\ell_\infty^n}$ to $S_{\ell_2^n}$ satisfies $\lim_{n\to\infty} \mu_n/\log n=1/2$. (2) Let $\lambda_n$ and $C_n$ denote the infima of the Lipschitz constants of homeomorphisms from $S_{\ell_\infty^n}$ onto $S_{\ell_2^n}$ and from $B_{\ell_\infty^n}$ onto $B_{\ell_2^n}$, respectively. Then $\lim_{n\to\infty} \lambda_n/\log n=\lim_{n\to\infty} C_n/\log n=1/2$. The first result solves a problem of Gromov on nonzero-degree maps between finite-dimensional spheres, while the second settles Johnson's problem on the Lipschitz distortion of finite-dimensional unit balls.
Read the full abstract
Dynamical Systems
Sharp iterated-logarithmic thresholds for quantitative measure and orbit equivalence between integer lattices
We identify the sharp iterated-logarithmic thresholds at the critical exponent for quantitative measure equivalence and orbit equivalence between integer lattices. To be more precise, let $n>m$ be two positive integers, $\alpha$ be a positive number and $( \beta_j)_{j \geqslant 1}$ be a finitely supported sequence of non-negative numbers. We show that there is a quantitatively $t^{\alpha}\cdot \prod_{j \geqslant 1} ( \log^{(j)}{t} )^{-\beta_j} $-integrable measure equivalence from $\mathbb{Z}^n$ to $\mathbb{Z}^m$ if and only if either $\alpha<m/n$ or $\alpha=m/n$ and there is $j_* \geqslant 1$ such that $\beta_j= 1$ for all $1 \leqslant j < j_*$ and $\beta_{j_*} > 1$. Here $\log^{(j)}$ is the $j$-fold iterated logarithm. The same characterization holds for quantitative orbit equivalence. This characterization greatly strengthens the previous best-known result, due to the work of Delabie, Koivisto, Le Maître and Tessera (2022) and the work of Correia (2025), which asserts that there is a $t^{\alpha}$-integrable ($\alpha>0$) measure equivalence (or orbit equivalence) from $\mathbb{Z}^n$ to $\mathbb{Z}^m$ if and only if $\alpha<m/n$. In particular, our result solves, in much stronger forms, two open problems posed respectively by Delabie, Koivisto, Le Maître and Tessera, and by Naryshkin and Petrakos.
Read the full abstract
Strongly Correlated Electrons
Hidden Magnetic Complexity Within a Simple van der Waals Ferromagnet Ce$_2$Te$_5$
Ce$_2$Te$_5$ is a layered $f$-electron van der Waals magnet in which reduced dimensionality and inequivalent Ce sites give rise to competing magnetic interactions. We investigate its magnetic ground state using muon spin relaxation ($\mu$SR), neutron powder diffraction (NPD), and inelastic neutron scattering (INS). Zero-field $\mu$SR reveals an onset of static magnetism below $T_{\mathrm{C}} = 5.0(1)$~K, followed by an additional anomaly in the internal field at $T_{\mathrm{2}} = 2.3(2)$~K, consistent with features observed in bulk thermodynamic and transport measurements. In contrast, NPD data collected between $0.05-8$~K reveal a single long-range ordered magnetic phase below $T_{\mathrm C}$, with the magnetic Bragg intensities vanishing at $T_{\mathrm C}$ with no evidence for additional structural or magnetic phase transitions down to base temperature. The ordered state is characterized by a commensurate propagation vector $\mathbf{k}=(0,0,0)$ and ferromagnetic alignment of Ce moments along the crystallographic $b$ axis. Remarkably, only one of the two crystallographically distinct Ce sites carries an ordered moment of $\sim 0.85(3)~\mu_{\mathrm B}$ per Ce at 0.05~K. INS measurements establish the crystal electric field (CEF) energy scale of Ce$^{3+}$, revealing low-lying excitations at 7.94 and 21.46~meV and a Kramers doublet ground state with strong single-ion anisotropy. These results demonstrate that Ce$_2$Te$_5$ undergoes a single symmetry-breaking magnetic transition, while additional low-temperature anomalies reflect subtle modifications of the ordered state driven by competing interactions and CEF effects.
Read the full abstract
Differential Geometry
Metrics with positive Ricci eigenvalue sums on $\mathbb{T}^k\times \mathbb{S}^m$
We construct metrics with $k$-positive Ricci curvature on $\mathbb{T}^k\times \mathbb{S}^m$, for $k,m\geq2$, arbitrarily close to the standard flat-round product in the smooth topology. The two-torus case disproves Wolfson's conjecture that every closed manifold with two-positive Ricci curvature has virtually free fundamental group. We also establish a local obstruction to two-positive Ricci curvature near closed backgrounds carrying three orthonormal parallel one-forms.
Read the full abstract
Computational Complexity
NP-hardness of ideal lattice problems
We establish the worst-case hardness of several ideal lattice problems (including SVP and CVP) in the $\ell_2$ norm by providing a dimension-preserving, deterministic polynomial time reduction from their generic lattice versions. The reduction constructs an ideal lattice in the canonical embedding of a number field that approximates some input lattice up to scaling and orthogonal transformation. The integers defining the ideal and the ambient number ring, in particular its discriminant, are all polynomial in bit length relative to the generic input lattice. Furthermore, the ideal is invertible, the ring is monogenic, and the number field is totally real. If the number ring is also required to be a full ring of integers, the reduction conjecturally succeeds in bounded-error quantum polynomial time.
Read the full abstract
Mathematical Physics
The Robust Ten Martini Problem
We solve the robust ten martini problem: Cantor spectrum with no arithmetic restriction on the irrational frequency for Type I analytic one-frequency quasiperiodic Schrodinger operators. We introduce Type I through a new acceleration-type quantity with stability under analytic perturbations. This yields a global open class of analytic potentials whose associated operators have Cantor spectrum for every irrational frequency.
Read the full abstract
Statistics Theory
When Is a Relevance Threshold Statistically Resolvable? Minimax Limits for Effect Classification
Statistical precision and scientific relevance operate on different scales. In regular problems, sampling uncertainty contracts at rate $n^{-1/2}$, whereas the magnitude below which an effect is scientifically negligible may be fixed or may vary with information. Let $\Delta_n$ denote a relevance threshold and $I_0$ Fisher information, and define $\lambda_n=\sqrt{nI_0}\Delta_n$. For separated negligible and meaningful parameter classes, we establish the minimax lower bound $\liminf R_n^*\ge 2\{1-\Phi(\varepsilon\kappa)\}$ when $\lambda_n\to\kappa$. If $\lambda_n\to0$, the experiments merge and consistent classification is impossible. If $\lambda_n\to\kappa\in(0,\infty)$, the problem converges to a Gaussian-shift decision problem. We characterize the exact minimax rule and risk in that limiting composite problem. As $\kappa\downarrow0$, the optimal cutoff converges to one statistical-error unit and the optimal improvement over trivial risk is $O(\kappa^2)$, while a simple relevance-boundary rule improves only at $O(\kappa^3)$; as $\kappa$ grows, that rule becomes asymptotically minimax. If $\lambda_n\to\infty$, consistent classification is attainable under a uniform estimation-resolution condition, verified for Gaussian and Bernoulli models. For $\Delta_n=dn^{-\gamma}$, the critical rate is $\gamma=1/2$. We also show that moving asymmetric relevance regions depend on standardized distances to their two boundaries rather than total width alone. For heterogeneous true effects, we derive a significance-saturation limit and show that point-null significance can be most selective for scientifically relevant effects at intermediate information. The framework characterizes when scientific relevance is statistically resolvable through a common information scale.
Read the full abstract
Computer Vision
VC-Attention: Value Smoothing and Softmax Casting for Low-bit Attention
Diffusion Transformers deliver state-of-the-art video generation, but their long spatiotemporal sequences make attention the dominant deployment cost, and a deployable low-bit kernel must be accurate and fast. Accuracy is limited by outliers: a block's quantization scale is set by its largest entries, leaving typical entries confined to a narrow range of representable values. Prior work smooths queries and keys, but value outliers follow no fixed channel or spatiotemporal structure and remain the dominant source of output error. Speed is limited by softmax: low-bit Tensor Cores accelerate only the two matrix multiplications, so the high-precision exponential between them becomes the longest pipeline stage on datacenter GPUs. We propose VC-Attention, a training-free low-bit attention framework that addresses both by pairing Value smoothing with a fused probability Cast. V-Smooth reorders value tokens by lightweight online clustering, so the tokens in a hardware block quantize well together. It quantizes only the residual after subtracting the block mean, and restores that mean from the row sum the online softmax already maintains. ExpCast-FP8 maps log-domain scores directly to E4M3 probability codes with one fused multiply-add, eliminating the FP32 exponential and the format conversion. We implement VC-Attention for B200, B300, H200, RTX PRO 6000, and RTX 5090. Across Wan2.2, LongCat-Video, HunyuanVideo-1.5, and MiniMax-H3, VC-Attention improves fidelity over low-bit baselines, speeds up the attention kernel over BF16 FlashAttention-4 by 1.46-1.59x on datacenter Blackwell and Hopper and by 2.3-3.6x on workstation cards, and generates a clip 1.13-1.19x and 1.36-1.70x faster end to end.
Read the full abstract
Analysis of PDEs
The semigroup generated by linear fractional divergence form operators in $L^p(\mathbb{R}^N)$: the constant coefficient case
Let $0<s<1$ and $A$ be a constant symmetric positive definite matrix. Given the quadratic form $$ Q_A(u)=\int_{\R^N} A\nabla^s u(x)\cdot \nabla^s u(x)\,dx, $$ where $\nabla^s$ denotes the Riesz fractional gradient, we establish necessary and sufficient conditions such that $Q_A$ defines a Dirichlet form. In strike contrast with the local case $s=1$, positive definiteness of a constant matrix is not sufficient for the Markov property when $0<s<1$. We also describe the associated symmetric Markov semigroup and its $L^p$ realizations, prove the existence of a smooth nonnegative heat kernel and establish a bounded $H^\infty$ functional calculus.
Read the full abstract
High Energy Physics (Phenomenology)
Existence of a second phase transition in A-A collisions at energies available at the CERN Large Hadron Collider: A physical interpretation
The observed fluid behavior of the matter produced in heavy ion and $pp$ collisions implies that the strings stretched between the constituents of the projectile and target should have a hard core giving rise to a repulsion between them. It is shown that in these collisions there are two critical string densities, one corresponding to the well studied percolation of strings and another to the percolation of the hard core of the strings. These two critical densities are related to two critical temperatures, one T $\sim $ 160 MeV would correspond to the restoration of chiral symmetry and partial deconfinement and a second transition T $\sim $ 220 MeV form the fluid behavior of QCD matter strongly interacting to a quasi free gas of quarks and gluons. This argument is supported by the recently published results on the sudden increase in the degrees of freedom above the deconfinement temperature in pp and A-A collisions.
Read the full abstract
Performance
Accelerating Transfer-Learning-Based Autotuning with Predictive LLVM IR Performance Ranking
As the complexity of High Performance Computing (HPC) ecosys- tems continually increases, achieving optimal performance becomes a challenge. Traditional performance autotuning techniques pro- vide promising means to navigate this complexity, these techniques remain computationally intensive and require many evaluations to find optimal configurations. This work proposes an autotuning framework that designs a machine learning-based ensemble LLVM Intermediate Representa- tion (IR) ranker, Neural Configuration Scorer (NCS). NCS ranks the performance of IRs sampled by a transfer-learning-based autotuner, improving the efficiency of the tuning process by reducing tuning overheads and circumventing subpar evaluations. By leveraging knowledge from related tasks, we are able to effectively exploit the transfer relationship to access high-performing configurations in fewer samples than traditional techniques that rely upon itera- tive refinement. Our framework can achieve similar performance improvements as state-of-the-art autotuning techniques with up to 61.67% fewer evaluations, averaging 27.85% fewer evaluations across various HPC benchmarks.
Read the full abstract
Optimization and Control
Revisiting Proximal Bundle Methods: Improved Rates under Hölder Smoothness
Proximal bundle methods (PBMs) are classical algorithms for nonsmooth convex optimization. Existing analyses of the classical PBM couple the null steps with the descent test. This coupling obscures how the bundle updates approximate the proximal subproblem. In this work, we consider composite objectives $F=f+h$ and view each null-step cycle as an inner bundle subroutine, called $\mathtt{ProxBundle}$. We analyze $\mathtt{ProxBundle}$ independently of any stopping criterion under general bundle model conditions and show that it automatically adapts to Hölder smoothness. Combining this inner-loop analysis with the descent-step analysis yields sharper complexity bounds for the classical PBM. For any fixed proximal parameter, its overall complexity is $\mathcal O\big(\epsilon^{-\frac{3-\nu}{1+\nu}}\big)$ for $\nu\in[0,1)$ and $\mathcal O(\epsilon^{-1})$ for $\nu=1$, where $\nu$ is the Hölder smoothness exponent. Choosing the proximal parameter proportional to $\epsilon$ improves the rate to $\mathcal O\big(\epsilon^{-\frac{2}{1+\nu}}\big)$ for $\nu \in [0,1)$. These are the first guarantees under Hölder smoothness with $\nu\in(0,1)$ for the classical descent test. We further introduce an absolute model-error test. The resulting PBM variant admits a clean inexact proximal-point analysis, and for any fixed proximal parameter, achieves the same complexity $\mathcal O\big(\epsilon^{-\frac{2}{1+\nu}}\big)$. Overall, our analysis separates the roles of the descent and null steps and gives a modular understanding of PBMs across different~tests.
Read the full abstract
Complex Variables
Shannon Integrals and Applications to Reimann-Hilbert Problems
We say that a function f can be represented by a Shannon integral if it can be written as the integral of sin(xs)/(xs), with respect to some finite measure. We develop a theory of Shannon integral representations for positive definite functions and show that under mild conditions f can be written as sum of two positive definite functions that constitute solutions to an additive Reimann_Hilbert problem on real line.
Read the full abstract
Solar and Stellar Astrophysics
Multi-chromatic observations of classical Cepheids using the CHARA Array interferometer: Surface brightness-colour relation, projection factor, and limb-darkening
The Baade-Wesselink (BW) method compares the linear and angular variations of Cepheids to derive their distance. This method is limited, however, by the projection factor, which relates the observed radial velocity to the true pulsation velocity of the star. Using simultaneous observations from the CHARA Array interferometer in the K, H, and R bands with the MYSTIC, MIRC-X, and SPICA combiners, respectively, we aim to understand the physics of Cepheid atmospheres better. Applying a specific method to multi-chromatic simultaneous interferometric observations of Cepheids, we derived robust limb-darkened angular diameters that were then used to calibrate the SBCR and study the projection factor. We also developed a strategy to measure the limb-darkening of Cepheids in R, H, and K bands. These measurements were then used to constrain the geometrical component of the projection factor. From the limb-darkened angular diameter curves of three Cepheids, we decreased the scatter of the SBCR in $V-K$ colour to 0.0011 magnitude and to 0.0040 for the SBCR in $G_{BP}-G_{RP}$. These SBCRs are particularly robust because they are based on multi-chromatic diameters and are homogeneous, which previous calibrations for Cepheids were not. For the very first time, we derived the limb-darkening of Cepheids in R, H, and K bands. For $\delta$ Cep, we derived an R-based projection factor of $1.275 \pm 0.051$, and the geometrical part obtained from the measured limb-darkening coefficient in R band was estimated to $1.420 \pm 0.016$ (on average), as expected from stellar static and hydrodynamical atmosphere models. The limb-darkening coefficients obtained in H and K band are consistent with models. These results demonstrate that multi-chromatic interferometry can improve the accuracy of the BW method. It is therefore essential to continue the CHARA survey of Cepheids in the coming years.
Read the full abstract
Game Theory
Delegating Authorization to Misaligned Agents: Coalitional Alignment and Safe Control
Long-running AI agents create a control problem: each action they take changes the state, which in turn affects the trajectory of future actions. If the agent is not fully aligned, then guaranteeing safety requires approving consequential actions before allowing them to be executed. But requiring human approval at every step makes attention a bottleneck. Delegating review to other AI agents raises the same alignment problem: the reviewers may themselves be misaligned. We identify a condition on a reviewing panel that is weaker than individual alignment yet necessary and sufficient for a guarantee that the principal fares at least as well in expectation as under a designated baseline policy. Each reviewer agent reports whether an action proposal made by a proposer agent improves its own utility relative to the baseline. We show that a threshold rule tolerating $k$ disapprovals is safe exactly when, after any $k$ reviewers are removed, the principal's utility can be written as a nonnegative combination of the remaining reviewers' utilities, plus a term that is nonnegative on every feasible proposal. We call this property $k$-robust coalitional alignment. The characterization lifts to sequential control: in a discounted MDP with an arbitrary proposer agent, safety at every state is both necessary and sufficient for the induced policy to match or improve on the baseline. When reviewers vote strategically, full-panel coverage in reward-function space guarantees that every Nash equilibrium is safe under the unanimous approval rule; in contrast, more permissive thresholds can admit unsafe equilibria even when reviewers are individually aligned. Experiments with existing reviewer models show that collective review can remain sound without an aligned individual, even when some disapprovals are tolerated.
Read the full abstract
General Economics
The Economics of Recursive Self-Improvement
We model the economics of recursive self-improvement (RSI) and assess its plausibility and impacts. First, we build a sequence of increasingly rich models of AI progress to highlight the feedback loops behind RSI. We represent our models as directed graphs and show that net acceleration in AI capabilities depends on the product of elasticities across each feedback loop. Second, we distinguish between "narrow" and "broad" AI capabilities, capturing the possibility that AI systems improve narrowly at optimizing AI R&D benchmarks without improving at broader economically valuable tasks. Third, we document existing estimates of key parameters and provide a wish list of empirical objects that AI companies can measure and feasibly share publicly. Finally, we calibrate the model with existing data. A back-of-the-envelope calculation suggests that feedback loops are not currently strong enough to generate a self-sustaining acceleration, though they appear to be strengthening. We conclude by assessing the plausibility and implications of such an acceleration.
Read the full abstract
Artificial Intelligence
When Should a World Model Move? Loss-Conditioned State Execution
We introduce loss-conditioned state execution, a model-agnostic method that decides whether to execute a world model's fixed feasible proposal or retain the current state. Predictive informativeness alone, however, does not establish whether an update will reduce downstream loss. Occurrence ranking can approach perfection while persistence remains the unique absolute-loss Bayes action. Two transition laws can also share occurrence information and conditional variance yet require opposite absolute-loss decisions. We formalize state movability as the existence of a loss-reducing feasible correction and distinguish it from the benefit of a particular proposal. Our method constructs a loss-specific feasible proposal from a predictive distribution and evaluates its groupwise bounded-loss gain over persistence on independent calibration units. The proposal is executed only in groups with a positive simultaneous lower confidence bound. For fixed proposals and groups with bounded unit losses, we prove that every accepted group has lower expected loss than persistence with high probability when calibration units are i.i.d. draws from the target population. Experiments on public forecasting and action-conditioned dynamics benchmarks show supported updates and a trade-off between certification and coverage. On 28,684 held-out M4 Monthly series, the method executes the proposal for 14.0% of series and achieves bounded loss 0.588, compared with 0.599 for persistence and 0.621 for always executing the proposal. The paired 95% bootstrap intervals for both comparisons lie below zero. In constrained forecasting of six unhealthy-inventory types from JD$\mbox{.}$com, a leading e-retailer in China, strong occurrence-ranking signal coexists with a loss-based preference for persistence, illustrating why event predictability and state execution must be evaluated separately.
Read the full abstract
Artificial Intelligence
Navigating Sparse Evidence: Agentic Visual RAG via Explicit Context Selection and Consolidation
Visual Retrieval-Augmented Generation (VRAG) empowers models to navigate and answer queries about visually rich documents by retrieving relevant page images as visual evidence and reasoning over their content. However, effectively utilizing this visual evidence is usually impeded by two main challenges. First, answer-relevant evidence is sparse and may be concentrated in a small region of one page or dispersed across multiple pages. Second, existing agentic methods often generate answers based on raw exploration trajectories or compressed textual memories rather than an explicitly organized set of supporting images, making answers susceptible to exploration noise and obscuring the evidence-backed reasoning trace. We argue that the bottleneck lies not only in evidence discovery but also in its preservation and organization before answer generation. We propose SCoRE (Selection and Consolidation for Robust Evidence), a unified agent loop for explicit evidence selection and consolidation. During exploration, SCoRE retains only query-relevant observations and their source pointers in a maintained textual ledger, preserving earlier evidence while keeping the visual context bounded. At termination, it reloads the referenced original images and consolidates the visual evidence for answering, arranging it into a logical sequence. This decouples final reasoning from exploratory trial-and-error while ensuring strict visual grounding via indexed claim-to-image linkages. To enable end-to-end optimization of this unified rollout, our training paradigm combines filtered cold-start trajectory distillation with evidence-aware reinforcement learning, whose reward promotes evidence coverage, consolidation compactness, and answer correctness.
Read the full abstract
Cryptography and Security
RAPID: A Real-Time Defense Against Unauthorized Model Distillation for Text-to-Image Services
Diffusion-based text-to-image (T2I) models are increasingly used for visual content creation, making their generation capability a valuable intellectual property asset. However, this capability is vulnerable to black-box output-based distillation, where an adversary queries the service, collects prompt-image pairs, and trains an unauthorized substitute model that mimics its generation behavior. Existing perturbation-based defenses apply sample-wise optimization to make generated images disruptive to unauthorized training, but introduce substantial computation and latency that reduce the usability of online T2I services. A natural solution is to integrate defensive perturbations into the VAE decoder, allowing the protected model to generate defended images directly without online sample-wise optimization. However, existing sample-wise objectives struggle to transfer to the shared decoder setting. We empirically find that a defensive shared decoder induces a substantially smaller latent shift than sample-wise optimization, suggesting that objective reachability matters more than destructiveness in this setting. To overcome this limitation, we propose RAPID, a self-referenced latent maximization framework that removes external dependencies and encourages the same model update to induce consistently disruptive effects across training samples, thereby improving reachability. We further introduce reconstruction-guided color regularization that blocks the latent shortcut and reinforces visual disruption. Extensive experiments on four T2I models and four datasets, with comparisons against five representative baselines, show that RAPID consistently degrades substitute-model generation quality while preserving service visual fidelity. Our work establishes a paradigm for real-time protection against unauthorized distillation in deployed T2I systems.