Peer Revenue Now accepting positions

Papers

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

All of arXiv

Saturday 7 November

Information Theory

Quantum synchronizable codes, polynomial and quasi-polynomial shifts

Quantum synchronizable codes protect quantum information against Pauli errors and against misalignments of the block boundaries. Almost all known constructions use cyclic codes, because the channel shifts the transmitted block cyclically. We study what happens when the algebraic shift of the code differs from the physical one. When one uses a twisted padding, constacyclic codes synchronize at no cost, and they are the only polycyclic codes with this property for a linear padding. For the other polycyclic codes, we measure the cost of each misaligned position, and we show that the phase damage it causes can be corrected as erasures. Using annihilator duality, we extend the construction to quasi-polycyclic codes, in which the cost is confined to a single block. We prove that the usual quasi-cyclic families have a distance bound that does not grow with the length, and show that quasi-polycyclic codes over repeated-root trinomials overcome it, giving synchronizable codes with larger dimension and distance than the known synchronizable codes of the same length.
Read the full abstract

Systems and Control

Bisect-Index-Choose: A Fast and Exact Algorithm for EV Charging under Minimum Charging Rates

Electric vehicle (EV) charging stations are increasingly operated with an aggregate power limit below the sum of their port ratings, and require scheduling algorithms to divide the available power among the connected EVs. Charging standards permit a current that is either zero or at least a minimum rate, typically 6 A, which makes this allocation problem non-convex. Existing methods either solve a mixed-integer program with a commercial solver, which can be computationally expensive, adds software dependencies, and incurs licensing costs, or apply heuristics without optimality guarantees. This paper proposes Bisect-Index-Choose (BIC), a fast and solver-free scheduling algorithm that computes a globally optimal allocation among EVs under mild assumptions on their utilities. For $n$ EVs, BIC runs in $O(n \log n)$ time when all EVs share one maximum rate and in $O(n^2)$ time for two, times a logarithmic factor in the numerical precision. In tests with up to 1000 EVs, BIC attains global optimality and is up to three orders of magnitude faster than the solver. On real-world charging data, it needs less capacity and delivers more energy than existing heuristics. We implement BIC on a low-cost microcontroller and show that it solves instances with 1000 EVs in less than 90 ms.
Read the full abstract

Artificial Intelligence

How Short Is Too Short? Length-Conditional Calibration and Abstention for Detecting AI-Generated Text

Detectors of AI-generated text are usually tuned and tested on passages of several hundred tokens, yet many real cases are short. We study score-based detection as a function of passage length. We give a closed form for how the false-positive rate of a threshold set on long text changes as length falls, and a formula for the number of tokens a detector needs to reach a target true-positive rate. We calibrate thresholds within length bins with conformal p-values, and we add a bin-level rule that abstains when the detector cannot be informative at that length. On synthetic scores the formulas agree with simulation, and binned calibration holds the false-positive rate at its nominal value at every length. On three zero-shot detectors (log-likelihood, log-rank, Binoculars), three corpora and two open-weight generators, binned calibration gives a mean false-positive rate of 5.0% in every detector-length cell from 10 to 200 tokens, averaged over 300 random document-level resplits. A single threshold set at 200 tokens gives 29% for Binoculars at 10 tokens and about 1% for log-likelihood, as the closed form predicts. No detector reaches an 80% true-positive rate, and the best, Binoculars at 200 tokens, reaches 59%, so the rule abstains in every cell at that floor and releases only the 200-token bins at floors of 0.3 and 0.5.
Read the full abstract

Analysis of PDEs

Sharp Constants and Maximizers for Weighted Moser–Trudinger Inequalities on the Heisenberg Group

This article concerns Moser-Trudinger inequalities on the entire Heisenberg group, focusing on sharp constants, maximizers, and applications to subelliptic PDEs. We address several questions on the existence of maximizers in the singular case $0<\beta<Q$, where the singular weight yields compactness. We first establish an Adimurthi-Druet-type weighted Moser-Trudinger inequality without blow-up analysis and determine its sharp constant. We also prove its equivalence with the critical weighted Moser-Trudinger inequality. For the subcritical inequality of Lam, Lu and Tang \cite{Lam-Lu-Tang-2014}, we obtain two-sided asymptotic bounds near the sharp threshold and prove existence of maximizers in a suitable regime. This inequality also yields insight into borderline Sobolev theory and an asymptotic Caffarelli-Kohn-Nirenberg interpolation-type inequality. We further establish a weighted Moser-Trudinger inequality under inhomogeneous norm constraints and prove its equivalence with the subcritical inequality, thereby characterizing the associated supremum and its maximizers. An improved subcritical weighted inequality is also derived, and its maximization problem is shown to admit a maximizer. Finally, these inequalities are applied to normalized subelliptic equations driven by the $Q$-sub-Laplacian with nonlinearities having singular critical exponential growth at infinity. We prove the existence of nontrivial ground states for every prescribed mass $a>0$, provided the nonlinearity is sufficiently large in a suitable sense depending on $a$.
Read the full abstract

All authors Somnath Kumar, Nguyen Lam, Deepak Kumar Mahanta, Abhishek Sarkar

Number Theory

New lower bounds for even and odd values of the partition function

Let $N_{\mathrm{odd}}(X)$ and $N_{\mathrm{even}}(X)$ denote the numbers of integers $0\leq n\leq X$ for which the ordinary partition function $p(n)$ is odd and even, respectively. The Parkin–Shanks conjecture predicts that $N_{\mathrm{odd}}(X)$ and $N_{\mathrm{even}}(X)$ are both asymptotic to $X/2$. Bellaïche and Nicolas proved that $N_{\mathrm{even}}(X)\gg \sqrt{X} \log \log X$, while Zheng, building on work of Ono and Swaminathan, recently proved that $N_{\mathrm{odd}}(X)\gg \sqrt{X}$. We improve both bounds simultaneously by proving $\min\{N_{\mathrm{odd}}(X), N_{\mathrm{even}}(X)\}\gg X^{3/4}/(\log X \log \log X)$. These results have been verified formally in Lean by AxiomProver.
Read the full abstract

Machine Learning

Optimal Byzantine-resilient Federated Learning with User-level Differential Privacy

Real-world deployment of federated learning (FL) requires resilience against adversarial clients and data privacy. Byzantine fault captures the worst-case behavior of adversarial clients, while differential privacy (DP) provides statistical guarantees against privacy loss. In particular, user-level DP guards a client's entire contribution and provides more realistic protection against information leakage. Despite intensive efforts, the minimax-optimal convergence bound of Byzantine-resilient FL with user-level DP remains open. We establish an information-theoretic lower bound for any Byzantine-resilient empirical risk minimization problem under user-level DP and heterogeneous clients. We propose ByzULDP, a central DP algorithm featuring a two-level momentum across the client and server. Client-side momentum mitigates stochastic gradient variance during robust aggregation. Unlike the fault-free case, naively combining non-linear Byzantine-resilient aggregation with central DP noise can be insufficient to guarantee convergence to a bounded neighborhood of a stationary point. Our key contribution is a server-side momentum that leverages a property specific to user-level DP, where privacy sensitivity depends on how client updates are constructed, to empower convergence. ByzULDP achieves the order-optimal convergence bound for strongly-convex objectives, with vanishing error terms decaying at the optimal $O(1/T)$ and $O(1/\sqrt{T})$ rates for strongly-convex and non-convex objectives, respectively. We corroborate our analysis with numerical experiments on real-world datasets across diverse privacy budgets.
Read the full abstract

All authors Ming Xiang, Stratis Ioannidis, Edmund Yeh, Carlee Joe-Wong, Lili Su

Systems and Control

Computationally Efficient Data Generation for Learning Nonlinear MPC Policies

Learning MPC requires large datasets generated by repeatedly solving MPC problems. For Nonlinear MPC (NMPC), dataset generation is computationally expensive because feasible initial states are difficult to sample efficiently. The feasible set is often non-convex or even disconnected, and many randomly sampled states are infeasible, leading to wasted optimization solves and unnecessary computational effort. In this work, we propose a data-generation framework for learning NMPC policies that combines active-set exploration with projection-based sampling. First, a bilevel formulation samples activity weights to promote solutions associated with different active sets while preserving optimality for the original NMPC problem; second, a simpler slack-based formulation uses weighted slack variables to target different active sets at the single level. These reference solutions are then used to construct local linearized feasible sets, from which candidate initial states are sampled and projected through the nonlinear NMPC problem to recover feasible and locally optimal training samples. Our approach is up to $37\times$ faster than baseline uniform sampling with $99\%$ acceptance rate, while maintaining a representative coverage of the feasible set.
Read the full abstract

All authors Pietro Gori, Anton Pozharskiy, Franco Angelini, Manolo Garabini, Moritz Diehl, Armin Nurkanović

Quantum Gases

Three-dimensional multipole states of the Bose-Einstein condensate of polar molecules in a shallow two-dimensional trapping potential

We investigate the formation, stability, and dynamics of three-dimensional (3D) elongated multipole quantum droplets (QDs) in a Bose-Einstein condensate (BEC) of polar molecules with long-range interactions between molecular electric dipoles. The condensate is confined by a 2D trapping potential, which is a combination of the harmonic-oscillator potential and a central Gaussian potential hill. The system is modeled by an extended Gross-Pitaevskii (GP) equation, incorporating the dipole-dipole interactions and beyond-mean-field Lee-Huang-Yang (LHY) correction, which play a crucial role in stabilizing self-bound QD structures. In regimes of weak and strong confinement, stable multipole states with the number of lobes $n=2S$ are created by means of the imaginary-time-propagation (ITP) method from inputs with vorticity $S\leq 2$. The so created multipole QDs carry no angular momentum. Stable higher-order multipole states can be produced by the ITP method from inputs with $S\geq 3$ up to $S=6$ (i.e., $2S=12$ lobes) only under strong confinement and when the molecule number $N$ exceeds a threshold value. We analyze the dependence of the chemical potential, energy per particle, and peak density on $N$ and $S$ in the weak- and strong-confinement regimes. A noteworthy fact is that, below another threshold value of $N$, the minimum-energy (ground) state is created by the input with $S=1$, rather than $S=0$. Simulations of perturbed evolution reveal that, despite satisfying energetic stability criteria, only a subset of the multipole QDs are truly stable, with higher-order ones tending to develop instabilities or relax toward lower-order states. These results highlight the essential roles of the trapping potential, anisotropy of the dipole-dipole interactions, and LHY-induced repulsion in the formation and structural stability of multipole QDs.
Read the full abstract

Machine Learning

Robust Satisficing Ensemble: Scalable Model Aggregation Under Distribution Shifts

Standard ensemble methods can improve robustness by mitigating individual model stochasticity, but they frequently fail to generalize beyond the training distribution, leaving systems vulnerable to out-of-distribution (OOD) data. We introduce the Robust Satisficing Ensemble (RSE), a scalable probabilistic aggregation framework designed to secure model combinations by minimizing their fragility against distributional changes. By bridging a robust satisficing objective with a quadratic risk surrogate, RSE derives an analytical inner adversary. This fundamental reformulation enables highly efficient weight updates without the computational bottleneck of standard minimax optimization, crucially decoupling optimization complexity from sample size to scale effectively to large datasets. In particular, we prove a geometrically decaying truncation error for the fragility approximation via Lagrange-Burmann expansion and establish stationarity guarantees for RSE. We evaluate RSE across a comprehensive spectrum of distribution shifts, including adversarial label shift (SST-5, TREC), natural subpopulation and temporal shifts (CivilComments, HuffPost). Our results demonstrate that RSE consistently outperforms strong robust baselines, including GroupDRO, IRM, and SRM. These findings establish RSE as a practical, theoretically grounded safeguard for reliable model aggregation in dynamic environments.
Read the full abstract

Robotics

Zero-Shot VLA Adaptation with Retrieved Guidance

Pre-training large parametric "Vision-Language-Action" (VLA) models on large demonstration datasets has recently yielded promising results for general-purpose manipulation. However, these models still struggle under even small domain shifts relative to the training data, such as changes in camera viewpoint, background, object identity, or robot morphology. Motivated by the success of Retrieval-Augmented Generation (RAG) in language models, we pursue retrieval to boost VLA performance. Specifically, we seek to leverage retrieval from domain-mismatched, "in-the-wild" demonstration datasets like DROID to guide a VLA's behavior in a new target domain. To achieve this, we introduce RAGu, which overcomes robotics-specific challenges for RAG: (1) It incorporates simple geometric calibration to transform source demonstrations into the target setup and reduce cross-domain differences. (2) It uses a coarse-to-fine retrieval strategy that selects, from a large pool of pre-warped candidates, the trajectory most compatible with the base VLA. (3) It guides the VLA's flow-matching inference procedure rather than seeking to retrain the base model or expand its limited context window. We evaluate RAGu across real-world, simulation, sim-to-real transfer, and cross-embodiment humanoid manipulation. We find it improves real-world VLA success rates from 20% to 80% and outperforms the closest-related baselines in sim. Project website: raguvla.github.io
Read the full abstract

Artificial Intelligence

Complementarity without ground truth?

Human-AI complementarity captures the idea that a human assisted by an AI system can outperform both the human and the AI acting alone. In prediction tasks, this improvement is assessed against ground truth labels produced through a reference construction procedure and intended to represent the state of affairs to be predicted. Many human-AI interactions, including those where medical AI provides treatment recommendations, however, concern tasks other than predicting what obtains. Their relevant output concerns adequacy relative to a practical goal and depends on several evaluative dimensions. In this work, we examine the ground truth dependence of complementarity and extend it to human-AI interactions where ground truth is not the appropriate reference. We show that tasks beyond prediction can still admit a reference construction procedure that generates adequacy labels. We develop a route for extending complementarity to such settings. It replaces ground truth labels with context-dependent adequacy labels generated through a two-step scoring methodology inspired by multi-criteria decision analysis frameworks in healthcare and it makes the normative choices underlying reference construction explicit. Because the construction of adequacy labels embeds evaluative choices, we draw methodological inspiration from computational reliabilism to identify technical, scientific, and social factors relevant to supporting their construction. Our approach expands the empirical study of human-AI complementarity beyond prediction tasks, enabling its quantitative evaluation in decision-support settings where ground truth is not the appropriate evaluative reference.
Read the full abstract

Optimization and Control

Hypervolume-based column-and-constraint generation for bi-objective two-stage robust mixed-integer programs

Extending column-and-constraint generation (C&CG) for two-stage robust mixed-integer programs from the single-objective to the bi-objective setting is challenging: for a fixed first-stage decision and scenario, the second stage yields an entire nondominated set rather than a single objective value. We develop a C&CG algorithm that uses the hypervolume indicator - one of the most widely studied quality indicators in multi-objective optimization - to evaluate these sets and seeks a first-stage decision maximizing the worst-case hypervolume over a finite set of scenarios. To avoid the nonlinearities of an exact hypervolume formulation, we introduce a slab decomposition of the objective space that yields a linear mixed-integer master problem with an explicit bound on the discretization error. This bound enables a computable optimality-gap certificate and supports an adaptive algorithm that alternates between adding scenarios and refining the decomposition. We prove the correctness of the algorithm and establish sufficient conditions for its finite termination. Computational experiments on two-stage robust selection and assignment problems show that only a small number of scenarios need to be included in the master problem, even when the full scenario set contains 128 scenarios. Moreover, we identify the number of initial slabs as a key factor influencing the running time and develop a simple, instance-dependent rule for choosing the initial slab width that outperforms fixed choices on average for the selection problem and remains competitive for the assignment problem.
Read the full abstract

Artificial Intelligence

Approximate Recall, Approximate Forecasts: Recall as a Diagnostic for LLM Forecasting Errors

Large Language Models (LLMs) exhibit approximate recall: they retain coarse associations about past events while losing fidelity on specific details. Past research on LLMs has quantified different models' forecasting ability in real-world events using prediction markets, but has not studied how losing said fidelity impacts their forecasting abilities. For example: while an LLM correctly recalls that the answer to: ”Did Anna Sawai win the Best Drama Actress in the 76th Emmy Awards?” is ”Yes,” it incorrectly states the event date as September 22 rather than September 15. In this paper, we introduce a novel metric called the Recall Fidelity Score to study how imprecision in internalized knowledge propagates to forecasting capability on future events. On resolved historical markets, we find that higher local recall fidelity around semantically similar past events predicts lower forecasting error on resolved markets, and the same signal predicts smaller distance from open markets prices.
Read the full abstract

Metric Geometry

Admissible geometries for Talenti's Pólya–Szegő principle

Talenti symmetrization rearranges functions on an open connected set $X\subset\mathbb{R}^N$, endowed with a measure $\mu$ with continuous positive density, along the superlevel sets of a given function $f$. Talenti proved a Pólya–Szegő principle for it, provided that these sets are $\mu$-isoperimetric, have connected boundaries, and are crossed by straight lines of steepest ascent. We show that, if $X$ is convex, these requirements are very rigid: they force the superlevel sets of $f$ to be either parallel half-spaces, or concentric balls, or complements of tubes around an affine subspace of codimension at least two. This identifies the only geometries in which Talenti's Pólya–Szegő principle can be applied. We also prove a family of weighted isoperimetric inequalities in convex cones, showing that complements of tubes around affine subspaces of every codimension at least two do occur.
Read the full abstract

Materials Science

A data-driven approach for systematic correction of electronic bandgaps

We demonstrate that a single machine-learning correction, applied to bandgaps from density functional theory calculations with two of the most frequently used energy functionals, results in bandgaps that are close to the experimental values. This solves a classical problem of computationally efficient electronic structure methods, to give reliable bandgaps that can be compared to experiments. The employed approach relies on materials specific feature engineering of machine learning models, and is shown to have close accuracy to that of much more expensive calculations based on the GW-approximation or hybrid functionals. The findings reported here open the possibility to use efficient calculations in materials design workflows that need reliable bandgaps, with implications for investigations targeting large computational databases or for investigations that are entirely predictive in nature.
Read the full abstract

All authors Pamela C. Carvalho, Asa Marif, Erna K. Delczeg-Czirjak, Sergei Simak, Utkarsh Singh, Olle Eriksson

Statistics Applications

Digital Twins for Clinical Development: A Five-Level Framework from Prognostic Adjustment to In-Silico Trials

Digital twins are increasingly discussed in drug development, often narrowly as virtual control arms or synthetic data. We propose a broader framework for data-driven clinical digital twins, organized into five levels according to how much of the trial data matrix of baseline covariates, treatment, and responses is generated by a statistical or machine-learning model trained on historical, external, or real-world data. The levels are: (1) prognostic score adjustment, (2) response trajectory prediction, (3) virtual control, (4) synthetic population, and (5) fully synthetic patients and responses. For each level we describe the estimation workflow, the development decision it can support, and the principal statistical guardrail. As the share of model-generated information increases, the inferential weight placed on the model grows and regulatory maturity generally declines. We argue that credibility requirements should therefore scale with context of use, characterized jointly by model influence and decision consequence, following a risk-based credibility framework. We identify four recurring credibility questions (drift and transportability, estimand alignment, compounding error, and data quality and provenance), summarize the current regulator landscape, and discuss bottlenecks in data, model transferability, collaboration, and lifecycle governance. The central message for statisticians is to start from the decision rather than the model, and to earn trust use by use.
Read the full abstract

Quantum Physics

Periodic Readout Induces a First-Order Generalization Transition in a Quantum Perceptron

Variational quantum classifiers can generate highly structured decision boundaries through the Fourier dependence of observable expectation values on circuit parameters. We show that expressivity can come at a sharp statistical cost, by considering a quantum perceptron whose readout is the simplest periodic classifier, namely a square-wave function of the weighted inputs, and determine its typical generalization in the thermodynamic limit using a teacher-student replica-symmetric theory. Above a critical oscillation frequency, increasing the training load induces a first-order transition from weakly to strongly teacher-aligned phases, resulting in a discontinuous improvement in generalization. At fixed load, however, increasing the frequency eventually drives the equilibrium generalization error to chance level, despite perfect classification of all the training examples. Thus, the same periodic structure that enhances random-label storage can delay the onset of learning. Our results expose a quantitative tradeoff between memorization and generalization generated purely by periodic readout, and provide a benchmark for understanding Fourier-rich variational quantum classifiers.
Read the full abstract

All authors Fabio Benatti, Masoud Gharahi, Giovanni Gramegna, Stefano Mancini, Vincenzo Parisi

Computational Engineering

The one-step surface-air lag in the EnergyPlus zone heat balance: order of accuracy, correction and consequences

The order of accuracy sets how fast numerical error falls as the timestep shortens. This study measured that order for the zone heat balance of EnergyPlus 26.2 against exact or converged references and tested a correction on 45 U.S. Department of Energy prototype cases. All three zone air formulas, the third-order default included, give first order or lower, because the surfaces use the previous timestep's air temperature. Solving surfaces, air and HVAC to agreement within each timestep raises the default to about second order, and in a room with an exact solution two extra passes cut the error eightfold at 1.65 times the runtime. In the prototype cases, the correction changed annual site energy by at most 0.45% but an outpatient clinic's heating by 5.1%. Window-upgrade savings changed by 10% or more in 13 of the 34 applicable cases, about as often as under a change of timestep.
Read the full abstract

Geometric Topology

The Gorsky-Oblomkov-Rasmussen conjecture over $\mathbb{Q}$

We prove a generation theorem for the homology of certain Koszul complexes which naturally arise as the $E^1$ pages of the stable Rasmussen spectral sequence of torus links. Combining this theorem with recent work of Ballinger, Gorsky, Hogancamp, and Wang, we compute the stable Khovanov-Rozansky $\mathfrak{gl}_N$ homology of torus links and determine its Poincaré series, resolving conjectures of Gorsky, Oblomkov, and Rasmussen and their $\mathfrak{gl}_N$ analogs, over $\mathbb{Q}$. Using these results, we determine the Khovanov homology of $T(n,n-1)$ and $T(n,n)$ over $\mathbb{Q}$, confirming conjectures of Shumakovitch and Turner and of the first author, which in turn allows us to compute the Khovanov skein lasagna module of $\overline{\mathbb{CP}^2}$ over $\mathbb{Q}$.
Read the full abstract

Artificial Intelligence

Keeping LLM Agents on Procedure: Executable Finite-State Control for Long-Horizon Instruction Fidelity

Much of the automation handed to language-model agents is procedural: each step needs little intelligence, but a long unattended run must follow a skill or SOP stated once at the start. We show that agents drift from such procedures over long runs, and that restating the instructions or resuming a stopped agent only moves the drift to another shortcut. We study an executable finite-state machine (FSM) that keeps the workflow state outside the model's context and enforces it at three points: tool calls, state transitions and the end of the agent's turn. In a pre-registered experiment on 200-account SOP-Bench batches, Claude Haiku 4.5 with the FSM decided on average 139 accounts correctly and by the procedure, against 48 for the strongest baseline (95% CI of the difference [46, 130]), and kept this rate to the end; across three models, the gain tracked how much the plain agent drifted. Because the runtime, not the model, supplies this long-horizon reliability, the FSM can let small models carry procedural work: Haiku 4.5 with it averaged more such decisions than plain Claude Sonnet 4.6 at 38% of the cost. Analyses on SWE-bench Verified, $\tau^2$-bench, and staged and cross-context workflows further suggest that reliable long-horizon automation can come from stronger execution control, not only stronger models.
Read the full abstract

All authors Yutao Sun, Haokun Li, Mingshuai Chen, Guanyan Li, Zhongyi Wang, Shengchao Qin, Jianwei Yin

Combinatorics

The Brualdi–Quinn conjecture on strong edge-coloring holds fractionally and asymptotically

Let $G$ be a bipartite graph with parts $A$ and $B$, and let $\Delta_A$ and $\Delta_B$ be the maximum degrees of the vertices in $A$ and in $B$, respectively. Brualdi and Quinn conjectured that the strong chromatic index of $G$ is at most $\Delta_A\Delta_B$. We prove that the fractional strong chromatic index of $G$ is at most $\Delta_A\Delta_B$, and that the conjecture holds asymptotically: $\chi_s'(G)\le(1+o(1))\Delta_A\Delta_B$ as $\max\{\Delta_A,\Delta_B\}\to\infty$. The error term is $12(\ln\Delta)^{1/2}\Delta^{-1/4}$ with $\Delta=\max\{\Delta_A,\Delta_B\}$. Both results rest on a smart partition of edges and Lovász Local Lemma.
Read the full abstract

Atmospheric and Oceanic Physics

Guiding future global temperature distributions with simulations and observations

Assessing climate risks requires projections of daily temperature distributions, including rare and spatially widespread extremes. Climate simulations are costly and biased, while correcting them toward observations risks suppressing inter-model differences. We train a single machine-learning emulator, a conditional diffusion model, on seventeen CMIP6 models and ERA5 reanalysis to generate spatially coherent daily temperature fields conditioned on model identity, date, and CO2-equivalent concentration. At sampling time, diffusion guidance reduces model biases by steering each emulated climate-model distribution toward observations, with its strength selected through perfect-model experiments using withheld simulated futures. Observational influence decreases with projection horizon and forcing. Across seventeen synthetic targets, guidance more than halves median biases in the 1st and 99th percentiles of daily temperature, improves probabilistic scores and reliability, and reduces median annual global-mean temperature RMSE by 32-57% across three emission scenarios. In a detailed benchmark, guidance also outperforms established bias corrections on daily-field probabilistic scores. Efficient sampling generates large ensembles without further climate simulations. For example, our framework corrects CMIP6's roughly threefold historical overestimate of extreme-heat coverage, yet projects that concurrent extremes will affect nearly half of global land on a typical day in the 2080s under SSP3-7.0.
Read the full abstract

All authors Francesco Immorlano, Elijah Tavares, Felix Draxler, Juan Nathaniel, Violette Launeau, Duncan Watson-Parris, Padhraic Smyth, Pierre Gentine, Stephan Mandt

Machine Learning

Robust and Efficient Continual Model Merging via Global Singular Subspace Separation and Restoration

Continual model merging incrementally integrates task-specific models without revisiting past models or data. However, repeated merging entangles past task components and distorts their singular geometry, making balanced task-wise allocation increasingly difficult. We propose Global Singular Subspace Separation and Restoration (GS3R), which augments the merged update with a compact restoration state. At each merge, GS3R restores the retained singular geometry, separates task-wise subspaces, and globally reallocates the available rank across tasks fairly. To reduce merging-time memory overhead, we further introduce Pre-merge Restoration and Flushing (PRF). Experiments on vision, NLP, and multimodal benchmarks show that GS3R provides a strong accuracy-efficiency trade-off.
Read the full abstract

All authors Juncheol Shin, Byeongho Yu, Semin Kim, Eunhyeok Park

General Relativity and Quantum Cosmology

A note on Positivity and Rigidity for an Expanding Quasi-Local Energy

We study positivity and rigidity for an expanding quasi-local energy $E_\lambda$ associated with constantly umbilic initial data $(\Omega,g,k=\lambda g)$, introduced in earlier work [ALP26] as an analogue of the Brown-York energy adapted to an expanding flat de Sitter background. Under nonnegative scalar curvature and suitable boundary assumptions, our previous results established positivity for $\lambda$ below a geometrically determined threshold $\alpha$, while vanishing $E_\lambda$ at $\lambda=\alpha$ forced the boundary to be round. In this paper, we strengthen this endpoint rigidity statement by showing that, if $E_\lambda=0$ when $\lambda=\alpha$, then $(\Omega,g)$ is isometric to a closed Euclidean 3-ball of radius at most $1/\lambda$. The proof combines the rigidity theorem of Shi-Tam for the Brown-York energy with a strict concavity argument. We also construct an explicit counterexample to the positivity and rigidity conjecture formulated in [ALP26], showing that the energy can become negative beyond this threshold and that vanishing energy need not imply rigidity.
Read the full abstract

Algebraic Geometry

The moduli stack of affine bundles with fixed linearisation type

The goal of this article is the classification of affine bundles with fixed linearisation type over a compact complex manifold $X$. More precisely, let $E$ be a fixed holomorphic vector bundle over $X$ and $\mathcal{E}$ the associated locally free sheaf. We will assume that there exists an effective divisor $D$ on $X$ such that $H^1(X,\mathcal{E}(D))=0$. This condition is always satisfied when $X$ is a projective complex manifold. We define in a natural way the moduli stack of affine bundles on $X$ whose linearisation is isomorphic to $E$, and we prove that this moduli stack is isomorphic to the quotient stack $\big[H^0(X,\mathcal{E}(D)_D)/H^0(X,\mathcal{E}(D))\ltimes \mathrm{Aut}(E)\big]$ of the complex vector space $H^0(X,\mathcal{E}(D)_D)$ by the semi-direct product of the automorphism group $\mathrm{Aut}(E)$ (which is always an affine algebraic group) by the complex vector space $H^0(X,{\cal E}(D))$. Our isomorphism is obtained by constructing explicitly comparison functors $\mathscr{A}$, $\mathscr{T}$ between the two stacks and explicit natural isomorphisms $\mathscr{A}\circ\mathscr{T}\stackrel{\simeq}{\longrightarrow} \mathrm{id}$, $\mathrm{id} \stackrel{\simeq}{\longrightarrow}\mathscr{T}\circ\mathscr{A}$.
Read the full abstract

Probability

Martingale representations without finite or well-defined mean

The classical Strassen theorem characterizes convex order through martingale representations under the assumption of finite first moments. We show that a natural reformulation of the theorem remains true on the real line for distributions without finite or even well-defined means. We further investigate the higher-dimensional setting, where this equivalence fails in general. Martingale representations are nonetheless guaranteed in higher dimensions when every nonzero linear projection of the second marginal has undefined mean.
Read the full abstract

Robotics

Sampling-based Model Predictive Control as Sequential Monte Carlo Inference

Sampling-based model predictive control (MPC) is treated as a family of distinct algorithms: predictive sampling, MPPI, the cross-entropy method, and their biased, colored-noise, and annealed variants all sample control sequences and reweight them by cost. We show a single Sequential Monte Carlo (SMC) inference algorithm implements them all, specified by three independent axes: the proposal, the update rule that reduces cost-weighted particles to a nominal plan, and the effective-sample-size threshold for resampling. We evaluate SMC-based planning on MuJoCo non-linear control, open-loop navigation through obstacle fields, 7-DoF motion planning, and non-prehensile push-T, in simulation and on a Franka FR3. Resampling combats particle degeneracy at long horizons. At a 28-knot horizon in our navigation domain it cut the collision rate by 30% over non-resampling baselines. Programmable proposals improve push-T success over perturbative baselines, improving from 1% to 100% in simulation and 10% to 90% on hardware.
Read the full abstract

Astrophysics of Galaxies

Studying the Fullerene Nest: JWST Observations of the Planetary Nebula Tc 1. II. A Spectroscopic and Spatial Study of C$_{60}$ and C$_{70}$ Emission

We present a detailed analysis of the fullerene emission bands in JWST MIRI/MRS observations of the planetary nebula Tc 1. The observations reveal the emission features due to neutral C$_{60}$ and C$_{70}$ in unprecedented detail. The combined fullerene emission represents $\sim$2.6% of the total infrared output, suggesting that $\sim$0.6% of the carbon budget is locked up in fullerenes. We find no clear evidence for C$_{60}^+$, C$_{70}^+$ or C$_{60}^-$ in the spectrum. The high spectral resolution allows us to separate the fullerene bands from atomic emission lines for the first time, allowing for a much better characterization of the band profiles and yielding more accurate band ratios than previous observations. We see no spatial variation in the individual C$_{60}$ and C$_{70}$ band profile shapes across all pixels. The 17.4 and 18.9 $\mu$m C$_{60}$ bands show clear resolved ro-vibrational structure in their profiles, and similar structured emission is seen in the 12.5, 14.8, 17.7, and 21.8 $\mu$m C$_{70}$ bands, further establishing that all emitting fullerene molecules are in the gas phase. We report a higher 7.0/18.9 $\mu$m C$_{60}$ band ratio ($\sim$1.4-1.9) than previously reported ($\sim$1.3). We note that all bands, except the 8.5 $\mu$m band, contain a significant contribution from C$_{70}$, complicating the analysis. The fullerene emission is spatially concentrated in a thin, asymmetric shell at a radius of 5-6$''$ from the central star. The shell lies at the outer edge of the core of the nebula, which was ejected in a single, brief episode $\gtrsim$4,000 years ago.
Read the full abstract

All authors M. M. Giese, J. Cami, S. Van Schuylenbergh, E. Peeters, C. Bhatt, D. Van De Putte, C. Baerts, M. J. Barlow, J. Bernard-Salas, A. Candian, P. B. Changala, N. L. J. Cox, H. L. Dinerstein, V. J. Esposito, D. A. García-Hernández, M. A. Gómez-Muñoz, K. Justtanont, K. E. Kraemer, E. Lagadec, A. Manchado, A. Monreal Ibero, R. Sahai, A. Sidhu, G. C. Sloan, N. C. Sterling, A. G. G. M. Tielens, J. R. Walsh, R. Wesson, J. C. Whitman, A. Zijlstra, L. Finazzi, J. Martens, G. Berden, J. Oomens

Computer Vision

REDiff-AD: No Training and No Target Masks for Industrial Anomaly Image Generation via Diversity-Driven In-Context MMDiT

Industrial anomaly detection is challenged by the scarcity and diversity of real defects, leaving insufficient anomalous data to effectively train downstream models. Existing anomaly generation methods often require product-specific fine-tuning, target mask annotation, or costly procedures to achieve realistic and diverse defects. We propose REDiff-AD, a framework that requires neither model training nor target masks and introduces Rectified Flow MMDiT into industrial anomaly generation for efficient sampling with fewer generation steps. Given a normal product image and a single masked reference anomaly, REDiff-AD employs in-context learning to transfer the defect appearance from the reference into the normal image. Internal attention signals generated during this process are further reused to localize the synthesized anomaly, while edge and color cues refine the localization into pixel-level pseudo-masks. Two lightweight modules further improve generation quality: adaptive reference conditioning enhances defect fidelity, while diversity guidance promotes varied defect patterns without additional training. Experiments demonstrate that REDiff-AD can well address the tradeoff between realism and diversity for anomaly pattern generation and achieve competitive downstream anomaly classification, detection, and localization performance. These results show that REDiff-AD is a simple, efficient, and effective solution for large-scale anomaly data augmentation and pixel-level data construction. Github code: https://github.com/efdefe4-droid/REDiff-AD
Read the full abstract

Analysis of PDEs

Superlinear problems related to Fully Non Linear singular or degenerate operators in punctured balls

We study Fully Nonlinear uniformly elliptic equations with a higher differential term degenerate or singular, coupled with a zero order term singular at the point $0$, and an absorbing super-linear term of $p$ power type. We will limitate to punctured balls and positive radial solutions. In the continuity of two previous works in \cite {BDL1}, \cite{BDL2}, we will provide a complete classification depending on $p$ of the asymptotic behaviour of the solutions near the singular point $0$.
Read the full abstract