Analysis of PDEs
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 5 November
Several examples of one-homogeneous classical solutions to the one-phase free boundary problem with an isolated free boundary singularity are known. Adapting the approach of Caffarelli, Hardt, and Simon for minimal surfaces, we construct non-conical classical solutions with an isolated free boundary singularity by perturbing a one-homogeneous solution $U$. Moreover, if $U$ is strictly stable (respectively, non-degenerate), then any solution sufficiently close to $U$ is also strictly stable (respectively, non-degenerate).
Artificial Intelligence
ChainFlow: Few-Step Latent Flow Matching for Speculative Decoding
Speculative decoding has become a standard way to accelerate LLM inference: a lightweight drafter proposes tokens that the target verifies in one parallel pass. Existing drafters face trade-offs that limits their speedups: autoregressive drafters generate tokens one at a time, so drafting cost grows with draft length, while diffusion drafters generate tokens in parallel but predict each position independently, ignoring inter-token dependencies. We introduce ChainFlow, a new speculative decoding paradigm and the first flow-matching drafter conditioned on the target's hidden states. ChainFlow treats drafting as a short transport over hidden states, needing one or two flow steps instead of the many denoising steps of generation from noise. A variational autoencoder compresses these states so drafting cost barely grows with target size, lightweight feedback heads restore inter-token dependencies, and tree drafting raises acceptance. On Qwen3.5 models, at matched drafter size and training data, ChainFlow outperforms EAGLE-3 and DFlash in vLLM by 39% and 11%, respectively. These gains further widen as drafters are adapted to new languages sequentially. Our ablations cover drafter design and concurrent serving, and we open-source our code and drafter models.
Read the full abstract
Mathematical Physics
Standard Subspaces and Modular Theory – General Theory and Applications to QFT
This is a preview of an upcoming monograph on standard subspaces and modular theory. At the moment, it consists of the first chapter. This starts with a gentle introduction to the subject, and then describes the modular theory of standard subspaces, the characterization of their modular data by the KMS condition and reflection positivity, and various constructions of standard subspaces. Further topics in Chapter 1 include the connection of standard subspaces to complexifications and symplectic structures, including their polarizers, and operations with standard subspaces, namely direct sums, direct integrals, and tensor products. Many examples and classification results are included as well. This first chapter also contains a description of the free U(1)-current and some other applications to quantum field theory, such as KMS states of type I factors, the local subspaces of the free Klein–Gordon field, and the entropy operator defined by a standard subspace.
Read the full abstract
Soft Condensed Matter
Devitrification: crystal nucleation and growth mechanisms in glasses
If a liquid is cooled rapidly, it may fall out of equilibrium before it can crystallize. It may, however, crystallize later from the nonequilibrium glassy state. Using the Random First-Order Transition (RFOT) theory of glasses, we discuss a variety of mechanisms of crystallization as a route to devitrification. Ordinary nucleation is sped up by the frozen in energy of the glass relative to the liquid. At sufficiently low temperatures, a mechanism of nanonucleation appears, producing nanometer scale crystal inclusions in the glass. Nanocrystallization depends strongly on the local fluctuations in stability, and the transition from classical nucleation and growth to nanocrystallization occurs dominantly through rare, less stable glassy regions that percolate through space, producing an additional dramatic speedup in the crystal nucleation and growth rates. Similar fluctuations in local stability also explain anomalies of crystal growth in deeply supercooled systems.
Read the full abstract
Artificial Intelligence
Pdb or Not to Be: Comparing Debugging Interfaces With Pass$_{X}$@$k$
Interactive debuggers are indispensable to human software development, yet coding agents overwhelmingly rely on edit-execute-observe loops, using common bash tools, test scripts, and print statements to debug via repeated shell commands. Stepping through a debugger provides live access to runtime state, which may enable different problem-solving capabilities. However, prior work typically introduces the debugger as an additional tool or through a new subagent, making its individual contribution difficult to isolate from the associated scaffold changes. To address this, we compare a debugger-only agent against a bash-only agent on all 500 SWE-bench Verified instances, holding the rest of the scaffold fixed. We evaluate both architectures on two models, GPT-OSS-120B and Nemotron-3-Ultra-550B, with three independent runs per architecture $\times$ model configuration. Crucially, to distinguish interface specialization from expanded coverage under repeated sampling, we introduce pass$_X$@$k$, a cross-agent pass@$k$ that provides an unbiased estimator of the probability that at least one of $k$ agents is successful. Although the bash and debugger agents solve partially disjoint subsets of the benchmark in any single pair of runs, our pass$_X$@$2$ estimator reveals that this is largely due to run-to-run variance rather than tool complementarity. Across both models, we find that two bash rollouts outperform the combination of one bash and one debugger rollout. Our findings highlight the importance of accounting for stochastic variance in evaluating agent and tool diversity, and we hope that pass$_X$@$k$ will serve as a useful framework for the community.
Read the full abstract
Graphics
Dynamic GPU Voxel Grids and Voxel-sequential Processing in NanoVDB
We introduce two fundamental advances to the NanoVDB data structure: dynamic GPU-native topological updates, and a voxel-sequential access paradigm that dramatically accelerates point-wise and stencil-based operators on sparse grids. NanoVDB's linearized memory layout has delivered strong GPU performance, and the Index-Grid concept introduced in the fVDB framework [Williams et al 2024] provided topology/data separation and optimal compactness of storage in sparse machine learning workloads. However, all these prior iterations of NanoVDB assume a static topology that has precluded use of this data structure in simulation and dynamic geometric processes. We eliminate this limitation by introducing a suite of dynamic topology operators – morphological dilation, grid merging, pruning, and cross-resolution coarsening and refinement. These operators use bit-parallel algorithms to reconstruct grids containing hundreds of millions of voxels in just a few milliseconds. We also introduce a voxel-sequential access paradigm and data structure for GPU kernels operating on sparse data. We stream active voxels as a flat linear sequence, minimizing or eliminating the need for tree traversal in the process. This achieves compute density and regularity that allow even complex stencil-based operators to approach peak memory bandwidth, with minimal storage overhead. Together, these contributions remove the read-only constraint from NanoVDB, establishing it as a high-performance primary representation for dynamic GPU simulation and geometry manipulation. We evaluate the efficacy of our newly introduced operators in benchmarks sourced from applications such as 3D reconstruction, fluid simulation, and large-scale mesh voxelization.
Read the full abstract
Robotics
ForPC: Force-Aware Whole-Body Control for Precise Humanoid Cart Pushing
Humanoid robots are expected to take over strenuous material handling in manufacturing, logistics, and healthcare, where operating existing wheeled carts would let them move heavy loads without modifying the workplace. However, humanoid cart pushing remains challenging because the bimanual grasp closes a force-coupled loop in which the contact wrenches that drive the cart also disturb the robot's balance, and payload changes alter the coupled dynamics. We propose ForPC, a force-aware whole-body control framework that regulates cart velocity through two-stage reinforcement learning. A locomotion prior is first trained under varied hand loads and upper-body postures to provide balance support. A force latent and a state latent are then inferred from onboard observations, without external force sensors, and fed to a residual policy that coordinates bounded arm, waist, and leg adjustments with base-velocity commands. Multi-objective rewards penalize bimanual internal opposition, grasp slip, and actuator saturation, encouraging the policy to apply contact force on demand without human demonstrations. Sim-to-sim experiments show that ForPC reduces cart-velocity RMSE by up to 68.2% relative to a locomotion-only baseline (Loco-only) and cross-track RMSE by up to 74.2% on the multi-bend route. Trained solely in simulation, the policy also transfers to a physical Unitree G1 without real-world training or fine-tuning and, with no payload-specific adjustment, pushes a cart carrying up to 30 kg at 0.5 m/s. These results suggest that humanoid robots could eventually replace humans for efficient material handling in daily life.
Read the full abstract
Strongly Correlated Electrons
Disentangling Multiple Scaling Dimensions via Operator Covariance in Critical Quantum Spin Systems
We present a general proof of different power-law decays versus distance of the eigenvalues of two-point covariance matrices in critical lattice quantum models, allowing extraction of multiple scaling dimensions. We demonstrate the practical utility of this method in the context of a spin-1/2 chain with critical ground state described by the SU(2)_1 Wess–Zumino–Witten field theory. Using quantum Monte Carlo simulations to compute covariance matrices in different symmetry sectors, we resolve the expected relevant and marginal primary operators as well as some descendants. The covariance method can be integrated into many existing numerical schemes and promises to significantly enhance computational studies of quantum criticality.
Read the full abstract
Analysis of PDEs
Blow-Up and Life-Span for Nonlinear Hardy-Hénon Parabolic Equations
We consider the nonlinear Hardy-Hénon parabolic equation $ u_t=\Delta u+|x|^l|u|^\alpha u,$ $t>0,$ $x\in\mathbb{R}^N$ or $x\in\mathbb{R}^N\setminus\{0\}$, with $N\ge1$, $l>-\min\{2,N\}$, $\alpha>0$, with nonnegative initial datum $u_0$. We first establish new criteria for finite-time blow-up. We then derive optimal upper bounds for the life-span $T_{\max}(\lambda\varphi)$ of solutions arising from the initial datum $u_0=\lambda\varphi$. These estimates establish a precise dependence of $T_{\max}(\lambda\varphi)$ on both the initial profile $\varphi$ and the parameter $\lambda$ in the small and large-amplitude regimes.
Read the full abstract
Spectral Theory
Liouville transforms for ordinary differential equations with distributional coefficients
We study Liouville transforms for the differential equation $Ju'+qu=wf$ on the real interval $(a,b)$ where $J$ is a constant, invertible, skew-hermitian matrix and $q$ and $w$ are matrices whose entries are distributions of order $0$ with $q$ Hermitian and $w$ non-negative. In particular, we examine their use in the treatment of quasi-regular endpoints and their impact on the Weyl-Titchmarsh $\mathbb M$-function associated with the differential equation.
Read the full abstract
Distributed Computing
AutoParallel: Composable, PyTorch-Native Automatic Parallelism for Production Training
Training large models requires multiple interacting forms of parallelism. As models and clusters scale, finding and maintaining efficient parallelization plans becomes essential for training performance. Diverse model architectures, workloads, and hardware topologies demand flexible strategies, yet production PyTorch systems typically rely on hand-written recipes that are costly to adapt and tend to reuse familiar patterns, which may be suboptimal for a given workload. We present AutoParallel, a PyTorch-native planner for distributed model training. AutoParallel searches parallelization choices jointly across the training computation, allowing established techniques to be combined and unconventional strategies to emerge. It expresses its decisions using native PyTorch abstractions and produces ordinary PyTorch programs, avoiding a separate execution runtime while composing directly with the PyTorch compilation and training stack. AutoParallel also composes automatic decisions with model-specific distributed code, allowing dynamic or specialized regions to remain under user control. Plans are inspectable and replayable artifacts. Across dense, hybrid, and mixture-of-experts models on meshes of up to four dimensions and 128 NVIDIA H100 GPUs, AutoParallel matches or exceeds expert-written TorchTitan recipes on most configurations, with throughput gains of up to 18.3%, and discovers plans that uniform expert recipes do not express.
Read the full abstract
Atmospheric and Oceanic Physics
The Re-Foundation of Radar Meteorology
Since the early development of radar meteorology in the 1940s, rainfall retrieval has traditionally relied on empirical scaling relationships between radar variables and rainfall rate. Here, we revisit the premise that horizontal and differential reflectivity sufficiently constrain the rainfall rate despite the variability of the underlying drop size distribution. Motivated by evidences that this variability requires at least three degrees of freedom, we develop a nonparametric statistical retrieval algorithm that considers different combinations of drop count and distribution shape compatible with the same radar observations. The method separates the contributions of drop count and distribution shape. It then empirically estimates how the shape factors for rainfall rate and horizontal reflectivity, as well as differential reflectivity, vary with the mean and skewness of the drop diameter distribution. These relationships identify compatible combinations of drop count and shape, from which rainfall rates are calculated and their median returned as the estimate. Neither a prescribed analytical drop size distribution nor a scaling rainfall estimator or explicit classification of precipitation regimes is required. We evaluate the method using a worldwide database (149 disdrometer datasets from 76 locations), with radar observables calculated from measured drop spectra. In cross-dataset validation, the method achieves lower median dispersion of relative retrieval errors across all 16 combinations of training and validation instrument types with respect two benchmark methods consisting of the traditional scaling law relationship, and a direct nonparametric fit. These findings motivate a reassessment of the foundations of radar rainfall retrieval by incorporating the range of rainfall rates compatible with radar observation.
Read the full abstract
Signal Processing
Physical Sensing Verification for Cognitive ISAC against UAV Spoofing via Digital Twins
As integrated sensing and communication (ISAC) lets 5G and 6G base stations detect drones, software agents will act on these detections, so a replayed signal can plant a phantom drone that triggers real actions. This paper proposes a wall-based physical check for such cognitive ISAC systems. A digital twin (DT) predicts the delay and angle of a genuine target's wall echo. A relay can fake the direct echo's range but, standing elsewhere, not the direction of its own wall echo. The paper proves that line-of-sight beam steering cannot expose a matched relay, and that, under stated conditions, a single-antenna relay faking a wall echo leaves a telltale echo where a genuine target leaves none. In a pre-registered ray-tracing study, the twin's evidence separated genuine targets and replays perfectly (AUC 1.000), and a later-registered passive check without extra transmission accepted every held-out genuine target and no replay. The primary active probe was inconclusive (5.3% false acceptances against a 5% limit), lacking threshold margin. At the evaluated relay positions, the conditional bound certifies 3 of 24 held-out streets against adaptive or fixed single-antenna relay signals meeting its conditions, reproduced on fresh streets. Doubling the bandwidth raises this to 11 of 24.
Read the full abstract
Computer Vision
JEPAhead: From Inpainting to Causal Forecasting for Urban Driving
We fine-tune a pre-trained V-JEPA1 predictor from bidirectional latent inpainting to action-free causal forecasting while keeping the encoder frozen. Using contiguous past observations, horizon-aligned targets, and group-level causal attention, the model forecasts eight future token groups, corresponding to sixteen frames, in one forward pass. On CARLA Town12, it retrieves the correct future among 591 candidates in 69.0% of cases, compared with 55.3% for persistence, with greater gains at longer horizons and on tokens exhibiting greater change. It also outperforms a compact predictor trained from scratch with six times fewer parameters. Semantic analysis indicates driving-relevant information in predicted features: a fixed linear probe trained on real encoder tokens achieves 0.16 mIoU on patches undergoing semantic-class changes, versus 0.13 for persistence. Vehicle-conditioned retrieval improves over persistence by 15 percentage points at the longest horizon. Pixel-space evaluation shows modest L1 gains but little improvement under LPIPS, motivating direct evaluation through latent and semantic measures alongside an oracle reconstruction reference. These findings support further investigation of latent forecasting for driving perception, while downstream utility remains unverified.
Read the full abstract
Machine Learning
Higher Order Component Attribution via Learned Surrogates in High Energy Physics
Experimental High Energy Physics (HEP) has adopted the transformer architecture for identification of jets, particularly the state-of-the-art Particle Transformer (ParT) which injects explicit information about particle correlations into the attention mechanism. While able to achieve state-of-the-art results in accuracy, interpretability to its effectiveness has been limited. In this paper we propose component attribution methods using Kolmogorov-Arnold Networks (KANs) that directly tie back to the physical descriptors of jets, to probe whether mechanisms such as Multi-Head Attention (MHA), or the ParT architecture, are able to learn these as discriminating features.
Read the full abstract
Computer Vision
Scoring, Not Learning: Decomposing Catalog-Consolidation Gains in Web-Scraped Retail Product Recognition
Retail product catalogs assembled from web image search are cheap to build and pervasively noisy: queries return packaging shots instead of products, and entries proliferate under names that cannot be distinguished visually. Consolidating such a catalog can look like an alternative to scaling the model. We show this is not a comparison between two ways of improving a recognizer, but between improving a recognizer and changing what counts as correct. We study a public archive of 165,769 cheese images across 1,824 query-named entries. A packaging filter and near-duplicate removal reduce it to a 998-class recognition set on one fixed, group-aware split. Canonicalizing entries to their base type (844 classes) raises top-1 accuracy by 4.0 points. Scoring the unchanged model against the merged label space, by summing its predicted probability within each canonical group, recovers +4.27 points with no retraining (+4.25 over six seeds). Retraining on the consolidated catalog then adds -0.10$\pm$0.53, +0.08$\pm$0.65 and -0.10$\pm$0.65 points at 28, 50 and 88 M parameters, indistinguishable from zero at every capacity ($\pm$ is the across-seed standard deviation). That component is specific, not mechanical: random merges with the identical size profile yield only +0.68 points, and the three backbones agree to within 0.1 points. But it is overwhelmingly category coarsening, not alias correction: across the 37 merged groups, type coarsenings supply 94% of the gain and alias collapses only 7%. We also detect no interaction between consolidation and backbone size: tripling backbone parameters is worth +1.50$\pm$0.57 points on the raw catalog and +1.42$\pm$00.44 on the canonical one, a within-seed interaction of -0.08 [-0.79,+0.63]. On this archive, the reported consolidation gain primarily reflects the evaluation label space, and separating that out costs one evaluation
Read the full abstract
Robotics
CS-Pi: Constraint-Satisfying Flow Policy Sampling
A flow-matching policy deployed in the field must often satisfy a constraint that was absent from its demonstrations, such as avoiding an obstacle it has never seen. Retraining for every new constraint is impractical, so the constraint must be enforced at inference time. Enforcing it must not discard what the policy learned: an action the policy would rarely sample may avoid the obstacle yet fail the task. We therefore want actions that are both safe, satisfying the constraint, and plausible, remaining likely under the policy. Gradient guidance, the most common approach, steers noisy states, so it must evaluate its cost on the policy's running estimate of the clean action; when the policy is uncertain between passing left or right of an obstacle, this estimate is their average and can pass straight through it. We instead build on rejection and projection. Rejection redraws actions until one is safe: every output is plausible, but the wait is intractable when safe actions are rare. Projection moves an action to the nearest safe one: it is efficient, but the result need not be plausible. Our key insight is that flow-matching sampling lets us combine the two, keeping projection's efficiency and rejection's plausibility. CS-Pi, our draw-project-renoise sampler, projects each draw to the nearest safe action, adds noise back scaled to how far the projection moved it, and re-runs the unmodified policy until a draw is safe. Replacing only the projection step extends CS-Pi to continuous costs. Across a 2D diagnostic, a 7-DoF shelf pick-and-place, an orientation preference, and a zero-shot Franka deployment, CS-Pi reaches 81% safe success on the shelf and 55% on hardware, against 55% and 40% for projection, while staying plausible.
Read the full abstract
Classical Analysis and ODEs
Fractional Fefferman–Stein inequalities on infinite rooted $k$-ary trees
We study weighted endpoint estimates for the centered fractional maximal operator on the infinite rooted $k$-ary tree, $T$. Our first result is a Fefferman–Stein type estimate \[ w\bigl(\{M_\alpha f>\lambda\}\bigr) \lesssim_{\alpha,s} \frac1\lambda\sum_{x\in T}|f(x)|M_{\alpha,s}w(x), \qquad s>1, \] with a constant independent of the branching parameter $k$. Although the proof is inspired by the non-fractional argument of Ombrosi, Rivera-Ríos and Safe in \cite{ORRS}, the fractional normalization introduces a genuinely new scale interaction. At radius $r$, one must renormalize the function by $k^{r\alpha}$ while keeping the dyadic cutoff at $k^r$; this leads to a new summation mechanism involving simultaneously $k^{r\alpha}$ and $k^{r(1-\alpha)}$. We then study sharpness and the geometry of admissible weights. We construct counterexamples showing that the estimate cannot in general be reduced to $s=1$ for any $0<\alpha<1$. In the subcritical range $0<\alpha<1/2$, we characterize power-type radial weights satisfying the natural fractional $A_1$ condition and prove that the failure persists after any fixed finite number of iterations of $M_\alpha$. We also prove that $M_\alpha w\lesssim w$ is necessary for the corresponding weighted weak type estimate but is not sufficient in the subcritical range. Finally, we identify a phase transition: for $1/2\le\alpha<1$ no nonzero global weight can satisfy $M_\alpha w\lesssim w$.
Read the full abstract
Computer Vision
An Anatomy-guided 3D Segmentation Pipeline for Prostate Surgical Phantoms from CT Scans
Full autonomy has been a long-term goal in robotic surgery and a near-to-medium-term goal for prostatectomy. Due to the relatively closed structure of the procedure, its standard-of-care robotic/laparoscopic solutions, and procedure economics, prostate procedures are an optimal early target for surgical automation. Prototyping and testing of possible solutions is done in constructed laboratory phantoms, either synthetic tissues or animal phantoms. To enable measurement of perceptual and robotic motion error, phantoms undergo volumetric imaging; these images are critical to both a potential deployed automatic system, used in planning motion and treatment (cutting, ablation, suturing, etc.), as well as post-hoc assessment of system performance. Rapid iteration on these phantom models yields small, inconsistent datasets, while the current workflow for human annotation is often anatomically unreliable and prohibitively time-consuming, rendering deep learning unsuitable at reasonable scales. To address these limitations, we present an automated 3D segmentation pipeline that uses classical medical image-processing techniques to encode the known anatomy of the phantoms as explicit geometric and intensity priors, which requires no training corpus and is transferred to each new phantom revision without re-annotation. Across 57 scans spanning multiple phantom revisions, our pipeline recovers the capsule interior at 0.83 median Dice on 11 contrast-suppressed scans and detects all 32 annotated focal lesions without false positives, completing a full scan in 2-8 minutes on commodity CPU hardware versus 1-2 hours by hand.
Read the full abstract
Computer Vision
Turkish Multimodal Video Fact-Checking: A New Dataset and a Comparison of Zero-Shot Agentic and Fine-Tuned Systems
Video-based misinformation spreads rapidly on social media, but almost all existing fact-checking systems focus exclusively on English text. Tools for low-resource languages such as Turkish are severely limited. In this work, we present TeViD (Turkish Video Fact-checking Dataset), to the best of our knowledge the first multimodal video fact-checking dataset for Turkish. We compiled the dataset from articles published by Teyit.org, an IFCN-certified organisation, and enriched it with translated Snopes articles and synthetic examples to reduce class imbalance. Building on TeViD, we develop two verification pipelines that share the same open-source base model. First, in 3MFact-TR (Pipeline 1), we adapt the multi-role 3MFact framework to Turkish with the Qwen2.5-7B large language model and the Qwen2-VL-7B vision-language model, and add three modules of our own: a semantic inconsistency checker, a 5W1H question manager, and a bilingual evidence retriever that excludes fact-checking sites to prevent leakage. Second, as our Fine-Tuned Model (Pipeline 2), we fine-tune a single Qwen2.5-7B with Low-Rank Adaptation (LoRA) to produce closed-book verdicts from the claim and the video's textual description. We report accuracy, macro F1, True class F1, and True class recall for both systems, together with ablation studies.
Read the full abstract
Data Structures and Algorithms
Differentially Private All-Pairs Min Cut Values
We study the problem of privately releasing the min cut values of all pairs of vertices in a given $n$-node weighted graph, under the edge-neighboring model. Straightforward application of advanced composition yields an approximate DP algorithm that outputs all min cut values up to $\tilde{O}(n)$ additive error. Prior works [Dalirrooyfard, Mitrović and Nevmyvaka, 2023; Aamand et al., 2026] have established that $\tilde{\Theta}(n)$ is the correct bound if the algorithm also needs to output cuts that realize the approximate min cut values. On the other hand, simply outputting the values could be an easier problem, but prior algorithms have been stuck on the $\tilde{O}(n)$ bound. In this work, we give an approximate DP algorithm outputting all-pair min cut values with $\tilde{O}(\sqrt{n})$ additive error, beating the linear error barrier. Our algorithm applies the sparse vector technique [Dwork and Roth, 2014] to maintain a hierarchical decomposition of the vertices based on the min cut structure. We also prove an $\Omega(\log^{3/2} n)$ lower bound for this problem.
Read the full abstract
Computer Vision
VoluCompose: Constructing Compact Volumetric Visual Representations for Medical Image Understanding
Medical vision–language models can interpret volumetric images, but processing full volumes is costly because hundreds of slices produce long, redundant visual-token sequences. Existing compression methods often couple content selection with budget determination. We observe that neighboring slices share visual features and that attention concentration varies across visual encoders. We propose VoluCompose, a training-free, model-agnostic framework that separates what visual content to retain from how many tokens to allocate. Its hyperbolic token composition module groups contiguous slices into ordered support units and uses the Lorentz distance to account for feature magnitude and direction during token allocation, anchor selection, and refinement. Its capacity control module uses visual-encoder attention entropy to allocate more tokens when attention is more dispersed. We evaluate VoluCompose on two tasks across three datasets and four backbones. Across nine calibrated settings, VoluCompose attains the highest mean quality–retention hypervolume among the compared methods. It achieves a mean relative quality of 121.68% against the uncompressed references while retaining 22.78% of visual tokens on average. Compression remains robust even at a 10% token budget. Across settings, the lowest ROUGE-L ratio to the uncompressed reference is 96.4% for VoluCompose. The corresponding minima for the compression baselines range from 15.0% to 70.5%. Code is available at https://github.com/yezanting/VoluCompose.
Read the full abstract
Astrophysics of Galaxies
TRGB Distances to the Andromeda Dwarf Satellites from the INT Monitoring Survey
We measure tip of the red giant branch (TRGB) distances to 17 dwarf satellites of M31 from Sloan $i$-band photometry of the Isaac Newton Telescope (INT) monitoring survey of Local Group dwarf galaxies, and add the distances of And I, And VII and And IX from earlier papers of the survey. We locate the tip with a Sobel edge-detection filter applied to the luminosity function of stars within two half-light radii of each galaxy. The distance moduli range from $23.38 \pm 0.06$ mag (And XVI, 474 kpc) to $25.35 \pm 0.06$ mag (And XVIII, 1177 kpc), and 15 of the 20 satellites lie between 650 and 850 kpc. For most galaxies, our distance agrees with at least one published value. Compared with Conn et al. (2012), our tip is brighter for 12 of the 18 galaxies in common, by up to 0.39 mag, possibly because of crowding in the central regions. And XII lies at $844^{+40}_{-38}$ kpc, about 60 kpc behind M31, consistent with a first infall into the Local Group. And XXII lies at $753^{+28}_{-27}$ kpc, closer than all earlier estimates, and about 59 kpc from M33 in three dimensions (for an M33 distance of 794 kpc), consistent with an association with M33. And XVIII lies about 400 kpc behind M31, too far for a secure association with M31.
Read the full abstract
Artificial Intelligence
IWIBench: Evaluating Gradient-Free Model Driven Intra-weight Injection
Intra-weight injection lets one model change another model's behaviour by editing its weights directly, without gradients. Such edits can range from altering lexical associations to enforcing output formats or updating factual knowledge. Despite growing interest in these approaches, there has been no systematic way to evaluate how well models perform them. We introduce IWIBench, which measures a large editor model's ability to modify the weights of a smaller target model (typically Gemma 3 4B IT) to achieve one of ten classical text-editing objectives while preserving its other capabilities. Editors are limited to four tools for reading and writing weights, with bounded tensor access and a fixed budget for generating and inspecting rollouts. A validator applies each proposed patch in a clean environment and compares the edited model against the unmodified target. Evaluating frontier closed- and open-weight models reveals sharply diverging capabilities, both across models and across tasks. A small number of models, most notably Astra and GPT-6.1 Sol, perform far ahead of the rest. The recency of this ability (little of it is evident in GPT-5.6 Sol, for instance), together with qualitative analysis, suggests it has been acquired through recent training. Tasks also vary widely: structured response resisted every attempt, and answer-length control yielded only one successful solution, whereas alias acquisition and recoding proved most amenable. Success in one task family correlated poorly with success in others, indicating an early and uneven capability frontier. We discuss strategies that editors repeatedly adopted and conclude with an outlook on intra-weight injection as an emerging but already potent approach to gradient-free behaviour modification.
Read the full abstract
Quantum Physics
Sign-hedged Beam QAOA:Scaling Quantum Optimization Algorithms
Quantum Approximate Optimization Algorithm (QAOA) encodes a combinatorial objective as an Ising Hamiltonian and alternates cost and mixer operations for p rounds. Recursive QAOA (RQAOA) scales further by using the resulting correlations, whether variables tend to agree or disagree, to eliminate variables one at a time. But each inferred sign is accepted irrevocably, so a single early mistake propagates through every later reduction and can discard the global optimum. As elimination proceeds, residuals shrink while effective edge density typically rises, so later decisions face denser Ising problems, a regime that fully connected Sherrington Kirkpatrick (SK) instances stress test directly. This paper proposes sign-hedged beam RQAOA, which retains the variable or pair selected by standard RQAOA, explores both possible sign assignments, keeps the best B residual problems according to a classical score, and completes each survivor with ordinary RQAOA. On frustrated SK instances, where standard recursion collapses to AR 0.474, beam recovers near-optimal AR 0.998 at the same p. We evaluate the workflow on 50+ instances from HamLib and QOBLIB, spanning 4 and 5 regular and Erdos Renyi MaxCut graphs, Sherrington Kirkpatrick (SK) spin glasses, and cardinality-constrained portfolio optimization, with up to 60 qubits and hardware runs on ibm kingston and ibm marrakesh. Matched simulator and device cells show consistent beam rankings: simulation and hardware agree that beam recovers near-optimal approximation ratio (AR) where standard recursion fails. These gains come from additional residual-circuit evaluations, classical pruning, and exact solution of a small core, rather than from increasing p, beam is a decision-side optimization atop fixed-p QAOA/RQAOA, not a claim that quantum circuits alone scale with qubit count.
Read the full abstract
Computer Vision
From Geometry to Function: A Multi-Task LiDAR Benchmark Dataset for Semantic and Functional Parsing of Urban Facades
Understanding building facades is central to urban digital twins, heritage modelling, and sustainable city analysis, yet existing LiDAR benchmarks stop at geometry and semantic segmentation and offer little support for reasoning about the functional role of each architectural element. We introduce HFX3D, a large-scale urban facade benchmark of approximately 492 million LiDAR points from 26 buildings in Halifax, Canada, captured across four seasons, day and night, and varied weather. Each facade is annotated at three tiers, semantic, instance, and per-instance functional attributes, making it, to our knowledge, the first facade dataset to pair geometry and semantics with functional labels. Annotations are produced by a semi-automatic pipeline that combines geometry-guided heuristics with expert validation. We benchmark representative point-based, graph-based, convolutional, and transformer architectures and find that semantic segmentation is already challenging under severe class imbalance, while instance separation of repetitive structures and functional reasoning from geometry are substantially harder. The benchmark establishes a new foundation for function-aware 3D urban understanding. Project page: https://jijup.github.io/hfx3d-benchmark/ ; code: https://github.com/jijup/hfx3d-benchmark
Read the full abstract
Artificial Intelligence
Decision Context: Persistent Behavioral Learning for AI Agents
LLM agents do not learn from corrections across sessions. A user fixes the same mistake at the start of every conversation. Existing agent memory systems primarily handle facts, such as preferences, entities, and conversational history, rather than procedural knowledge. We present Forge Decision Context (forge_dc), a system that captures behavioral signals from human corrections and self-assessed tool-call failures, compiles them into imperative RULE directives, and injects them at the decision points where they apply. We evaluate forge_dc alongside four predecessor approaches on DC-Bench, a new benchmark of 71 scenarios across nine behavioral dimensions, built from real executed agent traces in a deterministic mocked-tool environment. forge_dc reaches 95% compliance, up from 55% with naive retrieval, while injecting approximately 10 times fewer tokens per query. Across the five-stage progression, the strongest systems represent behavioral memory as explicit directives and deliver it at the decision points where it applies. The results identify representation, activation, and placement as central design choices for persistent behavioral learning.
Read the full abstract
Computation and Language
FinEvolveKG: Controlled and Attributable Prompt Evolution for Financial Knowledge Graph Construction
Financial institutions construct knowledge graphs (KGs) from shared SEC 10-Ks to serve risk analysis, portfolio management, and investment research, whose analytical needs differ. Extraction must adapt without fragmenting shared grounding, typing, and temporal rules. Hand-authored persona prompts and fine-tuning add cost. Weight-free prompt optimization offers an alternative, but unconstrained evolution of a monolithic prompt can rewrite shared rules alongside persona guidance, obscuring changes and complicating review. We introduce FinEvolveKG, a KG-construction harness separating reusable general and persona-policy blocks and mechanically confining GEPA, a reflective prompt optimizer, to the designated block. During persona adaptation, the approved general block remains byte-stable and every optimizer edit is localized to a separately versioned persona policy. A dual objective balances generic temporal-KG quality with persona-relevant facts, motifs, and multi-hop connectivity. We evaluate five mutation scopes on ticker-disjoint 10-K splits, with three seeds for the persona-conditioned arms (v3-v5) across three personas. On the three-seed mean, the complete two-stage block-constrained workflow (v4) improves over the frozen prompt baseline (v1) by 19.4-30.9% and over the tested unconstrained full-prompt workflow (v5) by 2.2-6.4%, leading v5 on both objective components and winning all nine paired-seed comparisons with the lowest seed variance in every persona. Without further optimization, v4 also exceeds v5 across all evaluated out-of-distribution (OOD) shifts. An SME audit supports using the generated triplets as ground truth; a blinded two-SME audit favors v4 over v5 in 70.1% of contrast-enriched judgments. These results show that mechanically attributable evolution can coexist with strong empirical performance, without attributing the gains causally to the boundary.
Read the full abstract
High Energy Astrophysical Phenomena
Testing influence of tides on gravitational wave emission with SPH code PHANTOM
In this article, we demonstrate how a Newtonian SPH (Smoothed Particle Hydrodynamics) code PHANTOM can be used to estimate the tidal contribution to gravitational wave (GW) waveforms. Using a binary neutron star (BNS) system with typical masses of 1.5 + 1.5 Msun as an example, we show — by comparing our results with waveforms calibrated against full numerical General Relativity (GR) simulations — that the tidal contribution to GWs can be reliably extracted using as few as about $10^4$ smoothed particles. Naturally, one must account for the difference in neutron star compactness between Newtonian gravity and GR. At separations greater than $\sim 80$ km, the relative tidal contribution falls below the level of numerical noise inherent in SPH. However, at such distances, GW waveforms for point-like sources are sufficient for detection purposes. It is also crucial to perform a hydrostatic/hydrodynamic relaxation of the initial models; otherwise, the entire GW signal is lost in the numerical noise.
Read the full abstract
Astrophysics of Galaxies
Digging into the Interior of Hot Cores with ALMA (DIHCA). VIII. Feeding a high-mass binary system in the making
We analyze the gas kinematics in the proto-binary NGC 6334I(N) ALMAe9-ALMAe14 system with the Atacama Large Millimeter/submillimeter Array (ALMA) at 1.33 mm with an angular resolution of 66 mas (~90 au). With a separation of 740 au between the system members, these observations allow a detailed study of the gas motion around this likely binary system hosting at least one growing high-mass star. Some CH$_3$OH lines tracing cold gas ($E_u/k_B$~100 K) reveal the presence of ordered circumbinary gas motion around the binary system within a ~1000 au radius, and rotation around the individual sources with rotation axes pointing roughly in the same direction. We estimate a central source mass, uncorrected for inclination, of 2 M$_\odot$ for the secondary, while the primary has an estimated mass of 5 M$_\odot$ as reported in previous works. The CH$_3$OH transitions with higher $E_u/k_B$ ($\gtrsim400$ K) values reveal circumstellar gas rotation in ALMAe9 consistent with the cold gas tracers. The vibrationally excited SO emission also traces rotation around ALMAe9 and potentially an accretion shock from a flow feeding the circumstellar disk of ALMAe14. We identify two potential inflows feeding the circumstellar disks from the circumbinary gas at rates of ~$10^{-6}-10^{-4}$ M$_\odot$ yr$^{-1}$ from CH$_3$OH emission. An additional third inflow seems to be feeding the primary from gas at larger scales.