Papers

Representation alignment with pretrained visual encoders improves diffusion transformer training but requires auxiliary encoders and extra training overhead. Recent approaches instead exploit representations from the generative model's own variational autoencoder (VAE). However, reconstruction-oriented VAE features are not optimized for the semantic discrimination and spatial structure needed for generative learning. We introduce CoVA (Composed VAE Representations), a lightweight framework that composes VAE features online into stage-wise guidance for flow-matching training. A small latent composer is jointly trained with class supervision, masked latent reconstruction, and horizontal-flip consistency, yielding semantically discriminative, spatially structured dense features and sample-dependent class posteriors. CoVA aligns intermediate SiT features with the composed representations early in training, then removes the increasingly detrimental alignment and reuses detached composer posteriors for sample-dependent soft label conditioning. We further uncover a hidden confound in soft label conditioning: averaging class embeddings attenuates their magnitude, coupling semantic interpolation with conditioning strength. This attenuation alone can substantially improve FID while degrading Recall, revealing a non-semantic source of the apparent gain. CoVA corrects this confound by restoring the hard-label magnitude, isolating posterior-dependent directional semantics. CoVA adds only 8.5M training-only auxiliary parameters (1.3% of SiT-XL/2) and introduces no inference-time overhead. On ImageNet 256$\times$256, CoVA reduces SiT-XL/2 FID from 17.2 to 7.4 (57%) in 400K steps. With classifier-free guidance, it achieves an FID of 2.03 in only 400 epochs, versus 2.06 after 1,400 epochs for vanilla SiT-XL/2.

Web crawling is a major source of large language model (LLM) pretraining data, yet data-selection pipelines discard a large share of crawled pages. We investigate how a crawler should allocate its budget across and within hosts to collect more high-quality pretraining data. Combining document-level quality classifier scores, web-graph centrality metrics, and Common Crawl's per-host crawl coverage, we find that roughly half of the variation in classifier scores occurs between hosts. In the crawls we analyze, a few pages from a host also provide a useful estimate of the quality of its remaining pages. Common Crawl uses harmonic centrality, a graph centrality metric, to guide its crawl allocation. In our simulations, using centrality to choose which hosts to crawl yields limited gains over random ordering and performs worse for some classifiers. For deciding how deeply to crawl each host, sampling a few pages and then fetching more from hosts with higher-scoring samples collects several times more high-quality pages per fetch than uniform depth or centrality-based allocation. Finally, Common Crawl's observed per-host coverage has little correlation with the evaluated quality signals. We estimate that tens of millions of high-quality pages remain among known but uncrawled URLs in each crawl examined, with most concentrated on a small share of hosts.

The road transportation sector is undergoing a radical transformation, working for a future where all vehicles will be autonomous and connected. A key challenge in this vision concerns intersections, which will no longer rely on traffic lights or traditional right-of-way rules. Instead, algorithms and communication protocols will be employed to optimize traffic flow. In this context, we are developing a testbed where cars equipped with wireless connectivity exchange information with a remote controller and coordinate their maneuvers. As the vehicles navigate through different types of intersections with various routes, they communicate with the controller, adjusting their speed to pass through intersections without stopping. This demo specifically showcases a hardware-in-the-loop (HiL) implementation of such intelligent intersection, using the SUMO traffic simulator to emulate vehicle positions. Real hardware and software components, which are designed for use in vehicles and roadside infrastructure, are employed to enable real-time data exchange of contextual and control information, running the cooperative coordination algorithm.

We present a training framework for developing core competencies in the practical operationalisation of AI ethics: (1) identifying ethical concerns; (2) determining which standard ethics requirements they implicate; (3) selecting and prioritising concrete technical and organisational measures to address them; and (4) recognising ethics tensions and reasoning about trade-offs where such tensions cannot be fully resolved. The framework is intended for researchers and engineers in AI, Research Ethics Committee members, policymakers, and students, providing them with practical tools for translating abstract ethical requirements into concrete actions. The framework provides a step-by-step approach organised around four interconnected tasks: issue identification, requirement mapping, measure selection, and tension identification. These tasks are supported by a suite of training materials, including an Ethics Checklist and a Portfolio of Measures. Here, we complement them with a set of practical exercises of three different types as well as trainer guides. We also discuss lessons learned in pilot training sessions and offer recommendations for trainers.

The efficiency of serverless computing fundamentally depends on rapid function startup, a challenge increasingly addressed by restoring from virtual machine snapshots. However, because content deduplication techniques can open timing-channel attacks that leak sensitive tenant data, production platforms like AWS Lambda instead enforce strict, zero-sharing isolation between snapshots of different tenants. While secure, this isolation precludes the deduplication of common infrastructure components, wasting per-worker cache capacity and exacerbating cold-start latency through inefficient remote fetching. Our characterization reveals that non-sensitive, shareable infrastructure accounts for roughly 70% of a typical snapshot's contents and, on average, about 50% of its working set. We present SnapShare, a cluster-wide snapshot management system that co-designs storage and caching to exploit this commonality without compromising sensitive data isolation. SnapShare classifies a snapshot page as shareable only when its full content matches that of a page from a provider-approved source; any non-matching content is treated as isolated private content. Common infrastructure components are selectively deduplicated and pinned across the cluster's worker nodes, while private data remains strictly isolated and is eagerly fetched as part of the tracked working set, falling back to on-demand retrieval only for pages that fall outside it. Implemented as an extension to Knative, SnapShare reduces end-to-end tail request slowdowns by 10-29% compared to traditional working-set and chunk-based platforms. It also reduces remote storage requirements by 39% and local cache footprints by 38%, reducing network retrieval payloads and bounding worst-case fetch delays by 10x.

Partial differential equation (PDE) surrogate modeling often relies on either dense observations or access to governing equations/physics constraints. In practice, observations can be sparse and partial, and the underlying physics may be partially specified or unavailable. We introduce SBNO, an equation-free paired-data diffusion bridge matching framework that learns conditional operators for both forward and inverse PDE problems under partial observations. SBNO uses diffusion Schrödinger bridge matching to learn a pair of time-dependent drifts from a mask-aware endpoint coupling that preserves empirical input-solution pairing during training and supports both PDE-task directions within a single framework. At inference time, SBNO produces a deterministic point reconstruction by integrating the probability-flow ODE constructed from the learned drift pair, without inference-time observation or PDE guidance-weight tuning. On five PDE benchmarks with 5% observations, SBNO achieves superior or comparable accuracy to baselines while demonstrating lower test-instance error dispersion. Furthermore, SBNO is at least $9.7 \times$ faster and reduces GPU memory by at least $4.3 \times$ compared to guided-diffusion baselines.

Drag-based image editing provides an intuitive interface for spatially controlled image manipulation by allowing users to specify handle and target points. Existing methods have made substantial progress through optimization-based, one-step, and feed-forward formulations, but they often interpret drag instructions as geometric displacements. This limits their effectiveness when the desired edit depends on interpreting image content and inferring a plausible semantic transformation from sparse point constraints. We introduce ThinkDrag, a unified multimodal framework for drag-based image editing with visual reasoning. ThinkDrag is trained to associate point constraints with meaningful object- and scene-level transformations, and can optionally produce an explicit reasoning trace that interprets the intended edit before image generation. This reasoning trace improves interpretability, particularly for ambiguous drags. Drag Endpoint Markers, motivated by an analysis of cross-image attention, provide an explicit correspondence signal that grounds each drag in image space. To support this framework, we construct a supervised dataset of semantic drag transformations paired with reasoning traces and introduce DragBench++, a benchmark targeting challenging drag-based editing scenarios with reference edit solutions. Experiments show that ThinkDrag achieves state-of-the-art performance, improving generation quality and plausibility while maintaining competitive point-following precision.

This paper considers channel state information (CSI) prediction using neural networks for time-varying multi-antenna channels from noisy historical observations. The main challenge is to exploit temporal correlation while distinguishing channel evolution from observation noise. We propose JEPA-kMMVAE, which combines a joint-embedding predictive architecture with a k-memory Markov variational autoencoder to predict future CSI through temporally structured latent representations. The framework supports direct prediction and residual corrections relative to the latest latent state and noisy CSI observation. Simulation results show that the residual configuration outperforms the evaluated baselines at moderate and high input CSI signal-to-noise ratios, while sequence baselines remain more accurate under stronger noise.

Memory retrieval aims to identify evidence from stored records to answer a query. Stored records with different factual details may share topics, entities, or events. Such shared contents would make records without the required evidence more likely to receive high similarity scores and outrank the correct evidence. In this paper, we use the information shared by these records to construct local references for query-memory matching. We propose NEighborhood-Aware Residuals (NEAR), a training-free framework that uses these references to construct residual representations that complement the original embeddings. Specifically, each memory derives a reference direction from its neighbors, and each query obtains its reference from a precomputed table. Using these references, NEAR constructs projection residuals for both queries and memories. These residuals are combined with the original embeddings for a single approximate nearest-neighbor search over a fixed memory index. Experiments on eight long-context retrieval benchmarks show consistent improvements across nine embedding models. Downstream evaluations show that NEAR improves question-answering performance.

Speech anonymization is evaluated with automatic speaker verification attackers whose performance depends on the attribute distribution of both their training and evaluation data, which may bias privacy estimates for some speaker groups. To alleviate this issue, we introduce VoxAnon, a dataset built from VoxCeleb1 and 2 with training and evaluation splits balanced across gender, age, accent, and pitch. We extract and cross-check attributes using two complementary methods, balance trial lists per subgroup, and draw test/enrollment utterances from distinct video sources in order to limit linguistic and session bias. Comparing VoxAnon with unbalanced training and test data across four attackers and three anonymizers, we show that attribute imbalance changes which subgroups appear hardest to re-identify, in directions that depend on the attribute. In the balanced setting, residual disparities persist, mostly tied to accent and pitch, and depend on the anonymizer/attacker pair. Code and trial lists are publicly available.

Adult attachment is commonly represented by attachment anxiety and avoidance, but these person-level dimensions do not by themselves specify how momentary relational conditions are perceived, compared with desired conditions, translated into action, and updated after interpersonal feedback. We present the Attachment Regulation and Control (ARC) framework as an integrative feedback-regulation account that makes these temporal assumptions explicit and empirically testable. In ARC, control refers to feedback-based regulation toward relational goals in the control-theoretic sense, not interpersonal dominance, coercion, or a generalized need for control. ARC does not claim a novel control-system architecture. Instead, it synthesizes established control-theoretic attachment ideas, secure-base representations, perceived partner responsiveness, relational discrepancy, dyadic regulation, and recovery into a parsimonious sequence for adult romantic relationships: perceived relational state, desired relational state, discrepancy, regulatory action, partner feedback, state updating, and recovery. The framework treats the discrepancy score itself as a falsifiable modeling restriction by comparing it with perceived-state-only, unconstrained perceived-plus-desired-state, and asymmetric discrepancy alternatives. Attachment anxiety and avoidance form core regulation moderators. Secure-base representation is a candidate requiring incremental prediction of recovery outcomes. Mentalized affectivity and regulation flexibility remain optional moderators, while adult disorganization is an independent extension outcome.

Inside a federated learning system a participant can quietly study the shared model and learn about others' private data. The usual defense, differential privacy, often adds so much noise that the model is no longer worth training. In cross-silo FL, where participants are a few accountable organizations and the server is trusted, we ask whether the server can instead interrogate clients. Each round it sends a secret challenge batch and asks every client to return a checksum of the gradient that batch induces, together with a timing. A client that mutated the model state, as an active generative-adversarial-network discriminator attacker does, returns a wrong checksum. We make this checksum robust to benign cross-platform floating-point non-determinism with a quantized fingerprint compared under a small tolerance. It flags two distinct active attacks at a true-positive rate of $1.0$ with no false positives on the checksum axis, stable across the fingerprint's parameters and up to $50$ clients at malicious ratios to $40\%$, for $32\,\mathrm{ms}$ of server work and $4.35\,\mathrm{kB}$ per response. Against a purely passive attacker the only signal is timing, and it is not robust: a cohort-relative test detects only in low-jitter conditions while a false-positive-free per-client baseline is blind, so no configuration delivers both. An attacker that answers the challenge on an untouched copy of the model while tampering with a replica evades the audit entirely. Therefore the checksum is a reliable in-band signal against attackers that do not decouple the audited computation from the attacked one, and it composes with differential privacy, whose accuracy cost we quantify on Texas100, rather than replacing it.

In this paper we propose and study arctan-skewing, a general approach to skew any symmetric distribution on the unit hypertorus. Arctan-skewing retains the main advantages of the popular sine-skewing mechanism, but improves on it by overcoming its two major limitations. These are a condition on the skewness parameter and a singular Fisher Information matrix in the vicinity of symmetry when combined with well-known symmetric models from the literature such as the circular von Mises or bivariate Cosine models. Arctan-skewing thus represents a compelling alternative to model skewed data on the circle and torus, notably from the field of protein bioinformatics. After a thorough investigation of its stochastic properties, we present a maximum likelihood estimation procedure and devise optimal tests for symmetry against our new class of skewed alternatives. Finally we show that arctan-skewed models better fit protein data compared to their sine-skewed counterparts.

6D pose estimation of novel objects without CAD models remains an open challenge: model-free methods must build the object representation from scratch during onboarding and have long trailed their model-based counterparts. We present NovaSplat, a model-free pipeline that closes this gap. From a short smartphone video captured at onboarding, we train a 3D Gaussian Splatting (3DGS) model that serves both as a template source for detection and coarse pose estimation and as a differentiable backbone for pose refinement. Our refiner optimizes the object-to-camera pose with combined photometric and geometric losses under a continuous rotation parametrization free of gradient singularities. On a household-object benchmark, NovaSplat surpasses all model-based methods. On a challenging industrial benchmark with textureless and specular surfaces, it outperforms all model-free competitors while substantially narrowing the gap to model-based ones. We further introduce NovaSplat-Lite, an inference variant that is $3.1\times$ faster at negligible accuracy cost. Code will be made available at \href{https://github.com/POSE-Lab/novasplat}{https://github.com/POSE-Lab/novasplat}.

Quantitative Gaussian approximation estimates on path space are established for the nonequilibrium density fluctuations of the simple symmetric exclusion process (SSEP) on a $d$-dimensional discrete torus of mesh size $1/n$, started from a product measure with a smooth density profile. A novel $n$-dependent modified Ornstein-Uhlenbeck process is constructed that incorporates higher-order lattice information while remaining linear and Gaussian. It is shown that, for sufficiently high modification orders, this modification yields quantitative Gaussian approximation error bounds of order $n^{-\min\{d/2,2\}}$ for a class of smooth functionals of the entire fluctuation trajectory. In addition, it is shown that retaining an explicit microscopic correlation correction leaves a remainder of order $n^{-d/2}$, thereby distinguishing contributions to the bound due to correlations from those due to spatial approximation. For nonconstant initial profiles in dimensions $d\ge3$, a separate argument yields matching upper and lower bounds of order $n^{-2}$ for the direct weak error of suitable quadratic test functionals, provided that the modification orders are sufficiently high. The proof is based on a novel path-dependent Trotter formula. Its derivation builds on the introduction of a concept of horizontal absolute continuity for non-anticipative functionals and a corresponding martingale formula for Markov jump processes. This formula is subsequently combined with an integral backward Kolmogorov equation for Gaussian continuation functionals. A second ingredient of the proof is a refinement of the exchangeable-pair argument introduced by the authors in a previous work, thereby improving the central-limit rate $n^{-\min\{d/2,1\}}$ to $n^{-d/2}$ for the initial empirical fluctuations.

The escalating environmental footprint of the Information and Communication Technol- ogy (ICT) sectorwith global Data Center electric- ity usage potentially reaching 3% by 2030 and the imperative for real-time responsiveness in smart buildings demand a foundational shift towards Edge Computing. This article proposes a holis- tic architectural and algorithmic approach for cre- ating autonomous, resilient, and privacy-preserving learning systems within smart buildings, advancing the core concept of the building itself as the data center. Our contributions, demonstrating originality compared to conventional centralized systems, are structured around three main pillars: 1) Validation of resilient data acquisition chains using the af- fordable Arduino ecosystem (Uno R4 WiFi/BME280) and secure persistence solutions including centralized archiving (Node-RED/MySQL) and secure remote transfer (Python/rsync/SSH). 2). The proposal of a lightweight, KOptim orchestration framework, inter- operable with the FIWARE standard, for the dynamic management of Service Level Agreements (SLA). 3) The adaptation and experimental valida- tion of the Generalized Hebbian Algorithm (GHA) for Principal Component Analysis (PCA) on low- power microcontrollers,We introduce and compare an iterative learning strategy using data buffers, featuring sorting and explicit preservation of extreme samples to maintain diversity. Experimental results on the high-dimensional Tour Perret sensor data confirm that this novel method, despite introducing a necessary temporal complexity of O(W log W ) due to sorting, achieves convergence and final result quality that is visually equivalent to processing the entire dataset in full memory, reliably identifying two distinct operational states.

Local online training on resource-constrained platforms requires low computational and training-state overhead. We study the numerical stability and computational efficiency of half-precision proximal policy optimization (PPO) on commodity GPUs. Directly converting PPO training to FP16 is unreliable because numerical errors feed back through policy updates and subsequent data collection, while PPO repeatedly compares a current policy with a fixed behavior policy. We propose PPO-FP16 with three structure-aware modules. Global scalar protection stabilizes statistics shared across a minibatch or parameter update. Policy-change residual protection extracts and preserves small differences between the current and behavior policies. Efficient persistent-update protection stabilizes optimizer state and parameter writeback while reducing state storage. These changes preserve the PPO objective, hyperparameters, and training budget. Across six continuous-control tasks and five seeds, PPO-FP16 achieves the lowest mean return-AUC rank (1.90) among the four methods and ranks first in 17 of 30 environment-seed comparisons. Eight of nine performance configurations yield PPO update speedups ranging from 1.16 to 2.03 times. Relative to Full FP32 Adam, optimizer-state memory is reduced by 71.7 to 74.9 percent, and peak allocated GPU memory is reduced by up to 44.5 percent. The results show that protecting a small set of structurally important quantities can deliver stable, efficient FP16 PPO without retuning.

We introduce MORAD, an online reinforcement learning framework for backbone-conditioned RNA diffusion models that trains a single shared policy across diverse target backbones. Structure-conditioned RNA inverse design seeks nucleotide sequences compatible with a target three-dimensional backbone, yet existing reinforcement learning approaches either optimize a separate policy for each target backbone or rely on offline preference optimization, leaving online post-training of a reusable backbone-conditioned model largely unexplored. MORAD combines backbone-conditioned group sampling, group-relative reward normalization, and a forward-process diffusion objective. Beyond existing RL rewards that primarily target tertiary structural quality, MORAD jointly optimizes six measurements spanning tertiary geometry, base pairing, ensemble behavior, and sequence composition. To aggregate these heterogeneous objectives despite differences in scale and optimization direction, MORAD maps each measurement to a bounded desirability score and combines them with a weighted geometric mean. Compared with a tertiary-structure-only reward, this multi-objective formulation improves 10 of the 11 evaluation metrics that have a preferred direction. Applied to pretrained RIDE, MORAD improves GDT-TS by 16.0% (from 0.3300 to 0.3827) and pairing MCC by 19.4% (from 0.6110 to 0.7295), while reducing normalized ensemble defect by 27.5% (from 0.3906 to 0.2830). On the test set, MORAD outperforms RiboDiffusion, gRNAde, and RDesign across tertiary structural accuracy, secondary-structure pairing, and thermodynamic quality, establishing a new state of the art in structure-conditioned RNA inverse design.

DAG structures can be learned from observations by incorporating acyclicity via smooth constraints, such as DAGMA's log-determinant function. Typically, observations and the underlying DAG are related through linear structural equation models, with Gaussianity assumption on the disturbances at each node. Extending DAG learning methods that account for scale and tail behavior of disturbances that vary across nodes remains a challenge. We propose a node-varying Student-$t$ DAGMA that jointly estimates the graph structure, node-specific disturbance scales, and the degrees-of-freedom parameter. A Gaussian scale-mixture representation yields an iteratively reweighted formulation that down-weights large standardized residuals. Building on existing results, we also establish identifiability of the graph and disturbance parameters. Synthetic experiments demonstrate improved structural recovery under heterogeneous heavy-tailed disturbances, and experiments on real finance data show that the proposed method yields lower test negative log-likelihood than the evaluated baselines.

LLM judges are increasingly deployed as panels, and disagreement among panel members is often read as a symptom of an unclear rubric. We test a label-free procedure that turns this disagreement into rubric repairs. The procedure decomposes each rubric into sub-criteria, ranks them by cross-judge disagreement on a development split, and has an editor model rewrite only the most contested ones. An additional stage that moves text-checkable sub-criteria to deterministic checkers transferred none in our evaluation. Human labels are never used to select or edit clauses; they are used only to evaluate the outcome. We pre-specified four hypotheses in a version-controlled protocol and evaluated six rubric conditions with a panel of five judges from three providers on SummEval, the HealthBench meta-evaluation, and USR Topical-Chat, for a total of 129,770 judge requests, excluding retries. On SummEval, the procedure raised inter-judge Krippendorff's $\alpha$ by 0.069 (95% CI 0.046 to 0.092) over the original rubric and by 0.074 over a control that applied the same edits to an equal number of randomly chosen clauses that were not selected. Agreement with expert ratings showed no detectable change (Kendall $\tau$ difference -0.007, 95% CI -0.037 to 0.021). On HealthBench and USR, neither inter-judge agreement nor human agreement met the preregistered effect criteria. In a preregistered secondary comparison, human agreement after editing the contested clauses was lower than after editing the control clauses on all three datasets in point estimate, although no interval excluded zero. Selecting clauses by disagreement makes judges agree with each other more than editing arbitrary clauses does, but in our setting the added consensus did not bring the judges closer to human annotators. We will release code, frozen rubrics, item identifiers, and judge scores.

Motivated by semi-Markov models of anomalous diffusion, we study obliquely reflected diffusions in generalized parabolic domains under independent inverse-subordinator time changes. We focus on infinite-mean subordinators, whose inverse clocks can slow the superdiffusive growth induced by reflection. For domains with cross-sectional radius asymptotic to $a_\infty x^\beta$, where $-1<\beta<1$, we strengthen the almost-sure growth laws of [MMW23] to obtain precise second-moment asymptotics when the subordinator's Laplace exponent is regularly varying at zero, and matching bounds more generally. Inverse $\alpha$-stable clocks yield mean-square growth of order $t^{2\alpha/(1+\beta)}$, giving subdiffusive, diffusive, or superdiffusive regimes. For $\beta<-1$, explosion persists almost surely, with finite mean explosion time if and only if the subordinator has finite mean. Supporting results establish strong existence, pathwise uniqueness, and a time-change representation, also allowing dependent clocks.

Co-speech gesture video generation aims to synthesize realistic, audio-driven portrait videos from a single reference image and a speech signal. Despite advances in visual quality, diffusion-based approaches still suffer from degraded local details and imperfect audio-visual alignment. We present PAGE (Parts-Aware GuidancE), a general, backbone-agnostic optimization framework for diffusion-based generation that enhances local spatial fidelity and global audio-visual consistency at both the training-objective and denoising-trajectory levels. During training, we propose Parts-Specific Reweighting (PSR), which adaptively increases supervision in semantically important regions such as lips, eyes, and hands, improving visual fidelity in perceptually critical areas. During inference, we introduce Consistency Classifier Guidance (CCG), an audio-conditioned latent classifier that identifies and corrects audio-visual inconsistencies in the generated video. We further design Differential Guidance (DG), an efficient inference strategy that refines facial and body motion separately while avoiding cross-region interference. To enable large-scale evaluation, we construct the Chinese Announcer Speech (CAS) dataset, the first comprehensive Chinese benchmark for co-speech gesture portrait video generation. Compared with the widely used PATS dataset, CAS includes substantially more identities, video clips, and scene diversity. Extensive experiments on both PATS and CAS demonstrate that PAGE achieves state-of-the-art performance in perceptual quality, gesture diversity, and audio-visual alignment. Project page: https://aigc-explorer.github.io/PAGE/.

Modeling human reactions to spatial audio requires understanding both what is heard and where it comes from. Prior work generates motion offline from complete acoustic and spatial input sequences. Humans, however, react to sounds without knowing what they will hear next. We present Sonic-AMG, to our knowledge the first framework for streaming human motion generation from spatial audio. The model encodes acoustic content and body-relative source geometry as separate tokens. Conditioned on these tokens, a causal transformer with a flow head generates continuous motion latents, from which a causal decoder reconstructs motion segments. Standard teacher-forced training derives motion histories and body-relative source locations from ground-truth motion, whereas streaming inference relies on generated motion for both, resulting in a training-inference mismatch. To reduce this history discrepancy, we introduce rehearsal training, which combines full-sequence teacher-forced supervision with supervised prediction along short autoregressive self-rollouts. To keep spatial conditioning consistent with generated motion, we propose spatial feedback, which recomputes body-relative source locations from generated root poses during both rehearsal training and inference. On SAM, Sonic-AMG achieves state-of-the-art FID and R-precision. It further supports real-time online generation of coherent reactions across sound events and source locations. Project page: https://jingchengni.com/sonic-amg/.

Pretrained vision transformers learn structured image representations spanning spatial patch tokens and non-spatial tokens. Recent analyses suggest that non-spatial register tokens encode global scene properties such as illumination, blur, and style. Yet patch-based image decoders discard these outputs, leaving the decoder to recover image appearance from patches alone. We investigate whether retaining register features improves decoding when the complete spatial patch representation is already available. We study this question using representation autoencoders (RAEs) with frozen register-equipped encoders. Our decoder processes four register tokens together with all patch tokens through joint self-attention. Appearance probes, paired reconstruction-error analysis, and fixed-decoder register substitutions examine which global image properties are accessible from registers and how their values affect reconstructed pixels. For generation, we train a multi-head attention pooling (MHAP) predictor to recover register features from sampled patches, keeping the pretrained diffusion model and its samples unchanged. On ImageNet-1K, our ViT-B decoders achieve mean reconstruction FIDs of 0.472 with DINOv2 and 0.442 with DINOv3, reductions of 9.5% and 5.2% relative to their respective patch-only baselines. The corresponding PSNRs are 19.73 and 20.54 dB. For generation, our DINOv2 models achieve a mean FID of 1.641, 9.8% below the patch-only baseline on identical sampled patches.

Multi-source semantic information fusion enables edge networks to support global inference using observations collected by distributed terminals. However, limited communication and computation resources restrict semantic uploading, leaving the edge server with incomplete tokens that differ in modality, semantic granularity, arrival time, and channel condition. The value of these tokens also depends on the complementary or redundant information provided by other terminals. In this paper, we propose a multi-source and multi-dimensional semantic information fusion framework that coordinates terminal uploading and edge-side fusion under resource constraints. We develop Hyper-Slot to combine semantic content with available token attributes through iterative token-to-slot assignment and slot-to-token feedback, producing a compact representation for global inference. To encourage useful semantic uploading, we further design a contribution-based incentive mechanism that uses Hyper-Slot assignments to evaluate information coverage and cross-type complementarity. A bi-level Stackelberg-Nash game determines the server reward pool and terminal transmission-quality strategies while accounting for computation and communication energy costs. Experiments on a tri-modal extension of COCO2014 and image-text retrieval on Flickr30K show that Hyper-Slot improves mAP by 3.66 percentage points and RSum by 15.9 points over the strongest evaluated baselines, respectively, while reducing the number of parameters by more than 80% compared with a naive concatenation-based fusion baseline.

We study the (typical) singularities of the distance function from a point in the ambient space to a polytopal complex in Euclidean space. By means of the Goresky-MacPherson splitting we generalize the existing linear theory to the quadratic distance function. Critical points occur on `selected' strata of the PL-space, where normal Morse Data determine the change of topology. On the homology level there exist indices with multiplicities, combined in a Morse vector. We focus mainly on PL-manifolds. We give bounds, depending on the dimension of the strata. In low dimensions we describe singularity types. At the end we discuss relations with Discrete Morse Theory.

High-precision retrieval over external knowledge bases is essential for LLM agents to operate effectively. However, existing Information Retrieval (IR) research primarily optimizes query-side retrievers (e.g., ColBERT, BGE) and generators (e.g., Self-RAG, GraphRAG), while reducing corpus-side representation to chunking. This leaves index-side quality an underexplored lever. We propose Pre-Retrieval Index Modeling & Enhancement (PRIME), an index-side optimization framework built on two complementary tracks: an LLM-driven Semantic Restructuring (SR) track that transforms long documents into structured, information-dense sections, and a deterministic Metadata Synthesis (MS) and Hierarchical Retrieval (HR) track that extracts semantic anchors and heading structure for coarse-to-fine section retrieval. Our evaluations show that deterministic metadata matches structure-aware chunking's retrieval quality while cutting embedding computation by 45.8 $\times$ and index vector count by 22.4 $\times$. On top of these gains, Semantic Restructuring (SR) further improves Answer-F1 by 13.7% in full deployment, with characterized quality-fidelity-cost trade-offs. Across QASPER, BRIGHT, and NFCorpus, we further characterize three qualitative regimes that determine whether index-side optimization helps or degrades retrieval, while showing that these index-side effects are neutralized under a strong reranker.

We consider Hecke modules of orthogonal modular forms attached to definite ternary lattices of half-discriminant $p^2$ with $p$ prime and $4p^2$ with $p$ an odd prime. Where possible, we prove explicit Jacquet-Langlands style correspondences with spaces of classical modular forms (in the spirit of Birch and Hein-Tornaria-Voight), and give conjectures for the remaining genera. This is done by applying sequences of Watson-style transformations that reduce the discriminant while preserving the adelic orthogonal group.

Extreme mass ratio inspirals (EMRIs) are promising sources for future space-based gravitational-wave detectors. We investigate EMRIs produced through the Hills mechanism, in which a stellar-mass black-hole binary (BBH) is tidally separated by a massive black hole (MBH), leaving one component bound to it. We present the first self-consistent modelling of this channel, combining an orbit-averaged Fokker-Planck treatment of the nuclear star cluster (NSC), three-body modelling of the binary-MBH encounter, Monte Carlo evolution of captured black holes (BHs), and fully relativistic EMRI trajectories to follow the evolution from binary supply down to final plunge. We find that Hills EMRIs are substantially more circular than classical EMRIs formed through two-body relaxation. For an MBH of $4 \times 10^6 M_\odot$ and circular, equal-mass BBHs of $20 M_\odot$ with initial separation of $0.1$ AU, EMRI mean eccentricities are approximately $0.097$ at our adopted LISA-band entry threshold and $0.034$ at plunge, compared with $0.478$ and $0.228$, respectively, for classical EMRIs. Of the initial BBHs, approximately $0.4$ per cent become EMRIs, whereas approximately $60$ per cent of captured BHs do. For comparable NSC initial compact-object mass, we find approximately $1.6 \times 10^{-8}$ Hills EMRIs per year per galaxy, approximately $12$ per cent of the classical EMRI yield. The Hills channel contribution to the overall eccentricity distribution depends strongly on the binary mass fraction in NSCs. We discuss implications for identifying EMRI formation channels with LISA and for quasi-periodic eruptions.

We investigate the quantum dynamics of a single ion confined in a radio-frequency (RF) Paul trap in the presence of weak octupole-induced anharmonicity, as can occur in realistic ion-trap potential. The explicitly time-dependent quadrupole potential is treated in a reference-oscillator Fock basis, allowing the effects of the RF drive and the quartic correction to be resolved directly at the level of motional-state couplings. While the quadratic RF interaction generates $\Delta n=\pm2$ transitions and Gaussian squeezing, the quartic contribution opens additional $\Delta n=\pm4$ coupling pathways. Starting from the reference-oscillator vacuum, we show that the octupole interaction broadens the even-parity Fock-state distribution and enhances the phonon-number fluctuations more strongly than the mean occupation. At the same time, the RF-induced sub-vacuum quadrature squeezing remains robust, although its magnitude and orientation are modified by the quartic interaction. The number statistics and quadrature fluctuations therefore respond differently to the same nonlinear perturbation, reflecting the distinct roles of Fock-space redistribution and phase coherence. The resulting dynamics provide a quantum-mechanical characterization of weak octupole corrections beyond the ideal quadrupole approximation and clarify how realistic anharmonicities modify nonclassical trapped-ion motion. Importantly, the motional nonclassicality remains robust against the weak octupole perturbation, as evidenced by persistent sub-vacuum quadrature squeezing and Klyshko-inequality violation. These results characterize how realistic anharmonic corrections reshape trapped-ion motion beyond the ideal quadrupole approximation.