arXiv:2510.04930v3 Announce Type: replace Abstract: Grokking is the phenomenon whereby, unlike the training performance, which peaks early in the training process, the test/generalization performance of a model stagnates over arbitrarily many epochs and then suddenly jumps to usually close to perfect levels. In practice, it is desirable to reduce the length of such plateaus, that is to make the learning process "grok" faster. In this work, we provide new insights into grokking. First, we show both empirically and theoretically that grokking can be induced by asymmetric speeds of (stochastic) gradient descent, along different principal (i.e singular directions) of the gradients. We then propose a simple modification that normalizes the gradients so that dynamics along all the principal directions evolves at exactly the same speed. Then, we establish that this modified method, which we call egalitarian gradient descent (EGD) and can be seen as a carefully modified form of natural gradient descent, groks much faster. In fact, in some cases the stagnation is completely removed. Finally, we empirically show that on classical arithmetic problems such as modular addition and sparse parity problem which this stagnation has been widely observed and intensively studied, that our proposed method eliminates the plateaus.
Science Journals
arXiv:2509.22510v3 Announce Type: replace Abstract: Alignment of Large Language Models (LLMs) is the ability to satisfy desired objectives during generation, which is critical for trustworthy deployment. In practice, alignment is often operationalized through multiple objectives such as Helpfulness, Harmlessness, and Honesty (HHH). Prior works study alignment via steering vectors in standard Transformer decoders but treat objectives in isolation, where optimizing a single objective can overwrite others, leading to interference. Recent works attempt to address this limitation by extending steering to a 1-to-N Transformer setting by replicating representations into objective-specific pathways, but apply transformations independently, resulting in inconsistent responses across objectives. Similarly, approaches such as safe RLHF and MoE-based designs study trade-offs across objectives but do not constrain objective-specific transformations within a shared representation during inference. As a result, even aligned State-of-the-Art (SOTA) LLMs can struggle to jointly satisfy HHH objectives in complex settings. To address this, we propose Adaptive Multi-Branch Steering (AMBS), a two-stage framework in a 1-to-N Transformer setting that parameterizes objective-specific transformations relative to a shared representation. In Stage I, a shared hidden representation is computed once. In Stage II, this representation is replicated into N pathways and updated relative to a shared reference, capturing objective-specific deviations while restricting divergence. This produces N objective-specific responses within a single forward pass, which can be combined at decoding to obtain a single response across objectives. Across multiple backbones, AMBS improves performance across HHH, with consistent gains in WR, TI, and SS (e.g., Avg 56.5% on LLaMA-2-7B) while maintaining efficiency (e.g., 189 Tok/s, 9 GPU-hrs).
arXiv:2602.22941v2 Announce Type: replace Abstract: Pacing strategies, defined by velocity and stroke rate profiles, are essential for peak performance in canoe sprint. While GPS is the gold standard for analysis, its limited availability necessitates automated video-based solutions. This paper presents an extended framework for reconstructing performance metrics from panned and zoomed video recordings across all sprint disciplines (K1-K4, C1-C2) and distances (200m-500m). Our method utilizes YOLOv8 for buoy and athlete detection, leveraging the known buoy grid to estimate homographies. We generalized the estimation of the boat position by means of learning a boat-specific athlete offset using a U-net based boat tip calibration. Further, we implement a robust tracking scheme using optical flow to adapt to multi-athlete boat types. Finally, we introduce methods to extract stroke rate information from either pose estimations or the athlete bounding boxes themselves. Evaluation against GPS data from elite competitions yields a velocity MAPE of 0.011 [0.008 0.014] (Spearman rho=0.974) and a stroke rate MAPE of 0.009 [0.006 0.013] (Spearman rho = 0.975). The methods provide coaches with highly accurate, automated feedback with minimal manual initialization work required, and without requiring sensors.
arXiv:2505.19155v2 Announce Type: replace Abstract: Due to the auto-regressive nature of current video large language models (Video-LLMs), the inference latency increases as the input sequence length grows, posing challenges for the efficient processing of video sequences that are usually very long. We observe that during decoding, the attention scores of most tokens in Video-LLMs tend to be sparse and concentrated, with only certain tokens requiring comprehensive full attention. Based on this insight, we introduce Sparse-to-Dense (StD), a novel decoding strategy that integrates two distinct modules: one leveraging sparse top-K attention and the other employing dense full attention. These modules collaborate to accelerate Video-LLMs without loss. The fast (sparse) model speculatively decodes multiple tokens, while the slow (dense) model verifies them in parallel. StD is a tuning-free, plug-and-play solution that achieves up to a 1.94$\times$ walltime speedup in video processing. It maintains model performance while enabling a seamless transition from a standard Video-LLM to a sparse Video-LLM with minimal code modifications.
arXiv:2605.18625v1 Announce Type: cross Abstract: We study the near-kernel sector of the Hamiltonian constraint operator in the one-vertex model of quantum reduced loop gravity using variational Monte Carlo methods with neural quantum states. The analysis is based on the symmetric Hamiltonian containing both Euclidean and Lorentzian contributions, and on the variational minimization of the positive quadratic operator $\hat{\mathcal Q}=\hat C \hat C^\dagger$ in truncated Hilbert spaces with spin cutoff up to $j_{\mathrm{max}}=1001$. The resulting near-kernel states are found to organize into three qualitatively distinct classes. At low cutoffs, we find solutions that do not factorize across the three edge degrees of freedom. At larger cutoffs, we find two different factorized branches, both described to very high accuracy by products of one-edge wavefunctions but localized in different spin regimes. One of these branches is matched with near-unit fidelity by reduced Thiemann coherent states, providing evidence for an emergent semiclassical organization of the near-kernel sector. The other is likewise strongly factorized, but its one-edge factors are not well described by the same coherent-state family.
arXiv:2605.18204v1 Announce Type: cross Abstract: Discrete diffusion models are a powerful class of generative models with strong performance across many domains. For efficiency, however, discrete diffusion typically parameterizes the generative (reverse) process with factorized distributions, which makes it difficult for the model to learn the target process in a small number of steps and necessitates a long, computationally expensive sampling procedure. To reduce the gap between the target and model distributions and enable few-step generation, we propose Forward-Learned Discrete Diffusion (FLDD), which introduces discrete diffusion with a learnable forward (noising) process. Rather than fixing a Markovian forward chain, we adopt a non-Markovian formulation with learnable marginal and posterior distributions. This allows the generative process to remain factorized while matching the target defined by the noising process. We train all parameters end-to-end under the standard variational objective. Experiments on various benchmarks show that, for a given number of sampling steps, our approach produces a higher quality samples than conventional discrete diffusion models using the same reverse parameterization.
arXiv:2605.18180v1 Announce Type: cross Abstract: Wide neural networks in the feature-learning regime drive modern deep learning, and yet they remain far less studied than their kernel-regime counterparts. We consider a critical yet under-explored difference between these two regimes: the regulariser and prior implied by gradient flow training. This canonical regularisation property is well-studied in kernel regime networks -- of all the infinite global minima, gradient flow selects exactly the vanishing ridge solution -- and underpins the celebrated NN-GP correspondence, precisely allowing the modelling of noise during training. However, we prove ridge regularisation biases gradient flow in feature-learning regime networks, even in the infinitesimal limit of vanishing regularisation. Over training, ridge distorts the inductive bias of the network, with a particular damage done to pretrained networks where the implicit prior is informative. We resolve this by axiomatising the canonical regulariser as a regime-agnostic function-space energy and lift, which uniquely identifies ridge in the kernel regime, and crucially generalises to the feature-learning regime. By studying the Riemannian geometry of feature-learning networks, we derive geodesic ridge from our framework, generalising ridge to the feature-learning regime. Correspondingly, we prove the canonical function-space prior is a Riemannian Gibbs Process, generalising the more familiar Gaussian Process. As a practical contribution, we propose arc ridge as a minimax-robust, scalable surrogate to geodesic ridge, revealing a deep relationship between early stopping and canonical regularisation across learning regimes. Finally, we demonstrate the consequences of our theory empirically on both image processing and NLP transfer-learning problems.
arXiv:2605.18159v1 Announce Type: cross Abstract: The Toffoli gate is a fundamental building block for quantum arithmetic and reversible logic, yet its efficient realization remains a major challenge in both near-term and fault-tolerant quantum architectures. Recent advances in dynamic quantum circuit capabilities, including mid-circuit measurement and classical feedforward, provide new opportunities for reducing the resource overhead of non-Clifford operations. In this work, we propose a set of dynamic decomposition strategies for multi-controlled Toffoli gates that exploit adaptive circuit execution and ancilla-assisted constructions. Our methods systematically reduce entangling-gate count, T-count, and T-depth compared with conventional static decompositions, while preserving fault-tolerance guarantees. Through analytical cost models and experimental evaluation, we demonstrate that relative-phase primitives and measurement-conditioned corrections enable scalable implementations with improved depth and resource efficiency.
arXiv:2605.18006v1 Announce Type: cross Abstract: Because LiDAR sensors acquire point clouds with a fixed angular resolution, the resulting data can be systematically parameterized and efficiently compressed in the spherical coordinate system. Traditional spherical coordinate-based point cloud compression methods have demonstrated strong rate-distortion (RD) performance, with the predictive geometry coding (PredGeom) method in the geometry-based point cloud compression (G-PCC) standard being a prominent example. Although PredGeom includes an inter-frame prediction mode, it relies on a simple linear model, which limits its ability to capture complex motion patterns and structural dependencies. Meanwhile, existing learning-based compression methods in the spherical domain do not exploit inter-frame correlations to reduce geometry redundancy. To address these limitations, we propose a learning-based inter-frame predictive coding method, termed Inter-LPCM. For azimuth prediction, we employ a delta coding strategy based on the predefined angular resolution. To improve radius compression, we introduce an inter-frame radius predictive (Inter-RP) model that estimates the current point's radius using neighboring points from both the current frame and the registered reference frame. In addition, we design a lightweight attention-based prediction (LAEP) model to predict elevation angles by capturing long-range geometric correlations across different coordinates. For quantization, we propose an RD-optimized method to select quantization steps in the spherical coordinate system. For entropy coding, we design distinct models for each spherical coordinate component. These models are adapted to the statistical priors of each coordinate, enabling more accurate probability estimation. Our source code is publicly available at https://github.com/SDUChangSun/Inter-LPCM
arXiv:2605.18473v1 Announce Type: cross Abstract: Electric double-layer capacitors (EDLCs), consisting of an ionic fluid between two metallic electrodes, are electrochemical energy storage devices complementary to batteries, allowing for a faster charge/discharge. The charging dynamics in response to a voltage step features a variety of regimes and relaxation timescales, depending on the applied voltage and the various lengths characterizing the system, most importantly the inter-electrode distance and the Debye length over which electrostatic effects are screened in the electrolyte. Inspired by recent works on "shortcut to adiabaticity" in colloidal systems, here we investigate the possibility to control the charge and discharge of planar EDLCs using time-dependent voltages. Specifically, our aim is to achieve a full charge or discharge within a finite time shorter than their intrinsic relaxation timescales. Within the Poisson-Nernst-Planck model and the small-voltage regime, we derive time-dependent protocols that can eliminate an arbitrary number of relaxation modes. This permits to approach the equilibrium charged state within a finite time, that can be in practice an order of magnitude faster than the natural equilibration time. We illustrate the relevance and efficacy of the method on polynomial drivings and show that the surface charge density, charge-density profiles, and global deviation from equilibrium (quantified by a Kullback-Leibler-like divergence) can all be significantly accelerated, even for driving times comparable to or shorter than the natural RC time of the system.
arXiv:2605.12070v2 Announce Type: replace Abstract: Asynchronous reinforcement learning improves rollout throughput for large language model agents by decoupling sample generation from policy optimization, but it also introduces a critical failure mode for PPO-style off-policy correction. In heterogeneous training systems, the total importance ratio should ideally be decomposed into two semantically distinct factors: a \emph{training--inference discrepancy term} that aligns inference-side and training-side distributions at the same behavior-policy version, and a \emph{policy-staleness term} that constrains the update from the historical policy to the current policy. We show that practical asynchronous pipelines with delayed updates and partial rollouts often lose the required historical training-side logits, or old logits. This missing-old-logit problem entangles discrepancy repair with staleness correction, breaks the intended semantics of decoupled correction, and makes clipping and masking thresholds interact undesirably. To address this issue, we study both exact and approximate correction routes. We propose three exact old-logit acquisition strategies: snapshot-based version tracking, a dedicated old-logit model, and synchronization via partial rollout interruption, and compare their system trade-offs. From the perspective of approximate correction, we focus on preserving the benefits of decoupled correction through a more appropriate approximate policy when exact old logits cannot be recovered at low cost, without incurring extra system overhead. Following this analysis, we adopt a revised PPO-EWMA method, which achieves significant gains in both training speed and optimization performance.
arXiv:2602.08206v2 Announce Type: replace Abstract: Open-vocabulary semantic segmentation has become an important direction in remote sensing, as it enables recognition beyond predefined land-cover categories. However, existing methods mainly depend on passive visual-text matching and often struggle with semantic ambiguity in geographically complex scenes, especially when different classes exhibit similar spectral or structural patterns. To address this issue, we propose a Geospatial Reasoning Chain-of-Thought (GR-CoT) framework for remote sensing open-vocabulary semantic segmentation. GR-CoT consists of an offline knowledge distillation stream and an online instance reasoning stream. The former constructs category interpretation standards for confusing classes, while the latter performs macro-scenario anchoring, visual feature decoupling, and knowledge-driven decision synthesis to generate an image-adaptive vocabulary for downstream segmentation. Experiments on the LoveDA and GID5 benchmarks indicate that the proposed framework improves overall segmentation performance and yields more semantically coherent predictions in complex scenes.
arXiv:2605.14292v2 Announce Type: replace Abstract: KV-cache compression at small budgets is a crowded design space spanning cache representation, head-wise routing, compression cadence, decoding behavior, and within-budget scoring. We study seven mechanisms across these five families under matched mean cache on long-form mathematical reasoning (MATH-500~\cite{hendrycks2021math}) with two distilled-reasoning models (Qwen-7B and Llama-8B variants of DeepSeek-R1-Distill~\cite{deepseek2025r1}) at budgets $b \in \{64, 128\}$. All seven were rejected. We then propose $\alpha$, a one-function modification to the TriAttention~\cite{mao2026triattention} retention scorer that replaces argmax-top-$k$ with greedy facility-location-inspired selection under a V-space redundancy penalty controlled by a single weight $\lambda$. A pre-registered protocol tunes $\lambda$ on a frozen development split and confirms on a disjoint held-out split; with $\lambda = 0.5$, $\alpha$ clears Bonferroni on two of the four (model, budget) cells (Qwen $b{=}128$ and Llama $b{=}64$), no cell is significantly negative, and the pre-registered Branch~A triggers. The finding is asymmetric: a minimal scoring modification beat heavier structural redesigns in this regime, and the combined matched-memory, sympy-graded, held-out confirmation protocol is the evidence standard that made the asymmetry visible.
arXiv:2601.14506v3 Announce Type: replace Abstract: Large language models are increasingly deployed in STEM education for personalized instruction and feedback across institutions in high- and low-income countries. These systems are designed to adapt content to student needs, but whether they adapt based on demonstrated ability or demographic signals remains untested at scale. Here we establish that LLM-generated STEM content systematically disadvantages marginalized student profiles across two cultural contexts, with the gap between the most privileged and most marginalized profiles reaching 2.55 grade levels. We audited four LLMs (Qwen 2.5-32B-Instruct, GPT-4o, GPT-4o-mini, GPT-OSS 20B) using synthetic profiles crossing dimensions specific to Indian education (caste, medium of instruction, college tier) and American education (race, HBCU attendance, school type), alongside income, gender, and disability, across ranking and generation tasks with FDR-corrected significance testing and SHAP feature attribution. Income produces significant effects across every model and context, medium of instruction drives the largest single effect in the Indian context, and disability status triggers simpler explanations. Effects compound non-additively: marginalization across multiple dimensions produces gaps larger than any single dimension predicts, and biases persist within elite institutions. Bias is consistent across all four architectures and persists through model selection, making intersectional, cross-cultural auditing a structural requirement before deployment.
arXiv:2605.18285v1 Announce Type: new Abstract: A key requirement on any well-behaved process language is its compositionality: behavioural equivalence of processes should be respected by the constructors of the language. Turi and Plotkin's abstract GSOS provides an elegant bialgebraic framework for modelling rule formats that guarantee compositionality from the outset. Their original results, however, are restricted to compositionality of strong bisimilarity, a rather fine-grained notion of process equivalence. In the present paper, we demonstrate that Turi and Plotkin's approach also applies to trace equivalence, which only observes external actions of processes. To this end, we revisit the general compositionality result of their original theory and present it in a refined form with regard to the required naturality conditions. This step makes abstract GSOS applicable over Kleisli categories and thereby enables reasoning about compositionality in the setting of coalgebraic trace semantics. As our main contribution, we introduce De Simone laws, a type of GSOS laws over Kleisli categories, and prove that their operational models are compositional for coalgebraic trace equivalence. This result recovers and explains compositionality of the well-known De Simone rule format for labelled transition systems in a natural categorical setting. As a further application, we derive from our general framework a novel De Simone-type format for probabilistic systems, compositional for probabilistic trace equivalence.
arXiv:2601.10705v3 Announce Type: replace Abstract: We study a semi-asynchronous client-server perceptron trained via iterative parameter mixing (IPM-style averaging): clients run local perceptron updates and a server forms a global model by aggregating the updates that arrive in each communication round. The setting captures three system effects in federated and distributed deployments: (i) stale updates due to delayed model delivery and delayed application of client computations (two-sided version lag), (ii) partial participation (intermittent client availability), and (iii) imperfect communication on both downlink and uplink, modeled as effective zero-mean additive noise with bounded second moment. We introduce a server-side aggregation rule called staleness-bucket aggregation with padding that deterministically enforces a prescribed staleness profile over update ages without assuming any stochastic model for delays or participation. Under margin separability and bounded data radius, we prove a finite-horizon expected bound on the cumulative weighted number of perceptron mistakes over a given number of server rounds: the impact of delay appears only through the mean enforced staleness, whereas communication noise contributes an additional term that grows on the order of the square root of the horizon with the total noise energy. In the noiseless case, we show how a finite expected mistake budget yields an explicit finite-round stabilization bound under a mild fresh-participation condition.
arXiv:2511.00392v2 Announce Type: replace Abstract: Accurate 3D reconstruction in visually-degraded underwater environments remains a formidable challenge. Single-modality approaches are insufficient: vision-based methods fail due to poor visibility and geometric constraints, while sonar is crippled by inherent elevation ambiguity and low resolution. Consequently, prior fusion technique relies on heuristics and flawed geometric assumptions, leading to significant artifacts and an inability to model complex scenes. In this paper, we introduce SonarSweep, a novel, end-to-end deep learning framework that overcomes these limitations by adapting the principled plane sweep algorithm for cross-modal fusion between sonar and visual data. Extensive experiments in both high-fidelity simulation and real-world environments demonstrate that SonarSweep consistently generates dense and accurate depth maps, significantly outperforming state-of-the-art methods across challenging conditions, particularly in high turbidity. To foster further research, we will publicly release our code and a novel dataset featuring synchronized stereo-camera and sonar data, the first of its kind.
arXiv:2605.17519v1 Announce Type: new Abstract: We present a comprehensive framework for gravity-driven, surfactant-laden thin films flowing over slippery substrates, elucidating how wall slip modifies the coupled hydrodynamics and interfacial transport. A long-wave model is formulated with a conservative bulk-surface mass balance and a Navier slip condition. The Orr-Sommerfeld eigenvalue problem governs the linear regime, while a weighted-residual model captures the nonlinear evolution over a range of equilibrium surfactant coverages, Marangoni strengths, and adsorption kinetics. The analysis predicts a non-monotonic variation of the critical Reynolds number with equilibrium coverage, exhibiting a maximum at intermediate $\Gamma_e$, and a slip-induced transition from single- to double-hump solitary structures with increasing Marangoni number, accompanied by attenuated capillary ripples. Under fast adsorption kinetics, the surface field homogenizes, preserving the mean film shape and flux while flattening both the surface concentration $\Gamma$ and the bulk inventory $\chi + h\phi$. A spurious interfacial mass growth reported by Pascal et al.(PRF, 2019) and D'Alessio et al.(JFM, 2020) is resolved through a revised surface balance ensuring strict conservation. Wall slip thus emerges as a key control parameter, reducing viscous resistance and mitigating Marangoni back-stress. The slip parameter $\beta$ is a useful control knob for surfactant-laden films. Slip prevents fragile multi-hump bound states, promoting a single broad crest or an almost flat, uniform sheet by carefully bonding $\beta$ to wave selection, ripple damping, and the bulk-surface surfactant balance.
arXiv:2605.17518v1 Announce Type: new Abstract: Spectral deconvolution is essential for extracting peak structures that encode material properties and chemical structures, but conventional automated methods often fail when spectra contain high-intensity noise or unknown background components. In practice, scientists rarely interpret spectra in isolation. Instead, they identify physically meaningful peaks by relating spectral structures to auxiliary information such as physical-property values, chemical structures, and trends across related measurements. Here, we propose a Bayesian framework that integrates spectral deconvolution with a model of expert scientific reasoning. In this work, expert scientific reasoning refers to the practice of evaluating candidate spectral structures by their consistency with independently measured physical-property values, rather than to manual expert intervention during inference. We formalize this reasoning as a physical-property regression layer, implemented using Gaussian process regression, and couple it with Bayesian spectral deconvolution. By averaging the physical-property likelihood over posterior predictive spectra inferred from Bayesian spectral deconvolution, the proposed method selects spectral models according to the consistency between inferred spectral structures and physical-property information. We validate the framework using synthetic spectra with high-intensity noise or unknown backgrounds and infrared spectra of poly(lactic acid). The method recovers physically meaningful peak structures that conventional Bayesian spectral deconvolution misses or misidentifies from spectra alone, including weak peaks in poly(lactic acid) IR spectra related to measured degradation rates. These results demonstrate that integrating expert scientific reasoning with Bayesian spectral deconvolution enables robust peak estimation under conditions where spectrum-only inference is unreliable.
arXiv:2503.10812v2 Announce Type: replace Abstract: Contrastive learning effectively clusters data despite a loss landscape filled with poor solutions, a success that is heavily dependent on the choice of data augmentations. How optimization consistently finds meaningful patterns remains an open question. We show this success stems from training dynamics rather than the loss function alone. Crucially, under a highly specific structural assumption governing the connectivity and variance of the data augmentations, we prove that once a critical spectral alignment threshold is reached, data features inevitably and rapidly separate into distinct clusters. We establish this mechanism for both discrete datasets and the macroscopic continuum limit, modeling latent dynamics as a Wasserstein gradient flow to demonstrate that this separation persists as the number of data points approaches infinity. We hypothesize that natural training dynamics inherently drive the system toward this critical state. We extensively validate this empirically across four diverse domains (synthetic shapes, images, text, and PDEs). In every setting, a sharp increase in this spectral quantity consistently precedes clean data separation, revealing that contrastive learning's success is governed by a dynamically emerging trigger tightly coupled to the underlying augmentation structure.
arXiv:2605.17491v1 Announce Type: new Abstract: Lasers that combine narrow linewidths with rapid tunability are critical for applications such as coherent optical ranging, distributed fiber-optic sensing, and precision spectroscopy. Despite significant progress in integrated laser technologies, the concurrent realization of low phase noise and frequency agility on a single integrated platform remains challenging owing to a fundamental architectural trade-off: conventional integrated laser designs typically suppress phase noise via high-$Q$ resonators, yet the extended photon lifetimes inherent to such resonators intrinsically constrain tuning speed. Here, we address this noise-agility trade-off by introducing a laser architecture that achieves ultralow phase noise and ultrafast tunability simultaneously. Rather than relying on ultrahigh-$Q$ resonators for self-injection locking, our design employs strong synthetic feedback within a Pockels-tunable, resonator-enhanced distributed Bragg reflector to suppress phase noise. As a proof of concept, we demonstrate a hybrid integrated laser with a short-term linewidth of 29 Hz, realized using a lithium niobate external cavity with a loaded $Q$ of only 0.62 million. The adoption of a moderate resonator $Q$ relaxes the photon-lifetime constraint on tuning speed, enabling sub-exahertz-per-second tuning rates and a chirp nonlinearity as low as 0.14%. Leveraging this laser, we implement a frequency-modulated continuous-wave LiDAR system that achieves a relative ranging precision of $1.7 \times 10^{-4}$ at a measurement rate of $1\,\text{MSa s}^{-1}$, without requiring complex chirp linearization techniques. We further demonstrate fiber-optic acoustic sensing capable of detecting sub-$\mu\epsilon$ dynamic strain, underscoring the platform's versatility for high-speed precision optical measurements. Our work provides a route toward cost-effective yet high-performance sensing and metrology systems.
GraphMAR: Geometry-Aware Graph Learning Framework for Spatially Adaptive CT Metal Artifact Reduction
arXiv:2605.17343v1 Announce Type: new Abstract: Computed tomography (CT) metal artifact reduction (MAR) aims to reduce the severe streaking artifacts induced by metallic implants and other high-density objects. Effective MAR generally requires both accurate artifact localization and artifact removal. Sinogram-domain methods can exploit explicit geometric cues, such as metal traces, to identify metal-corrupted measurements, while requiring raw projection data, which is often unavailable in clinical and practical scenarios. Image-domain methods are more flexible and widely applicable, yet they usually lack comparable geometric guidance, limiting their ability to localize artifacts and leading to suboptimal results. To address this limitation, we propose GraphMAR, a geometry-aware learning framework for explicit artifact identification and spatially adaptive MAR in the image domain. The key idea is to introduce graph-based geometric modeling as an image-domain analogue of sinogram metal traces. Specifically, we first construct a geometric graph from the metal mask and derive a geometric density graph that coarsely localizes artifact-prone regions according to inter-implant geometry. We then design GraphMoE, a graph-routed mixture-of-experts module that builds a polar-coordinate artifact graph in feature space and adaptively routes different experts to different spatial regions for MAR. By aligning the learned routing maps with the geometric density graph, GraphMAR provides explicit and interpretable artifact localization while enabling region-adaptive artifact reduction. Experiments on both simulated and real-world datasets demonstrate that GraphMAR achieves superior MAR performance compared with existing methods. To the best of our knowledge, this is the first work to introduce graph-based modeling for CT MAR and to enable explicit artifact identification in the image domain, improving both restoration quality and interpretability.
arXiv:2605.17488v1 Announce Type: new Abstract: The landscape of joint audio and video generation has been fundamentally transformed by the advent of powerful foundation models. Despite these strides, achieving cohesive multimodal customization for the simultaneous preservation of visual identities and vocal timbres across multiple interacting subjects remains largely underexplored. To bridge this gap, we present Omni-Customizer, an end-to-end framework targeted at the precise binding and seamless fusion of multimodal identity information. Specifically, we introduce an Omni-Context Fusion (OCF) module that effectively enriches the base textual prompt with dense, multimodal identity cues, along with a Masked TTS Cross-Attention (MTP-CA) mechanism explicitly designed to prevent the severe "speech leakage" problem. Within this architecture, we propose Semantic-Anchored Multimodal RoPE (SA-MRoPE) to anchor visual and audio reference tokens, along with TTS embeddings, to their corresponding semantic descriptions, enabling structured multimodal fusion and robust identity binding. Furthermore, we devise a comprehensive training strategy that incorporates interleaved audio-video scheduling to rapidly adapt the audio branch to multilingual scenarios without degrading foundational priors, and a progressive in-pair to cross-pair curriculum to facilitate the learning of high-level and robust identity features. Extensive experiments demonstrate that Omni-Customizer achieves state-of-the-art performance in dual-modal customized generation, excelling across visual identity similarity, timbre consistency, precise audio-video synchronization, and overall video-audio fidelity.
arXiv:2605.17486v1 Announce Type: new Abstract: Recent progress in Reinforcement Learning (RL) provides a principled approach to optimizing Vision-Language-Action (VLA) models, facilitating a shift from trajectory imitation to active learning in the task environment. Despite improvements in control precision, most RL optimizers remain task-specific, which reduces VLA models from generalist controllers to policies that overfit to a narrow set of tasks. In this study, we conduct an in-depth analysis of this phenomenon and highlight the importance of cross-task feature representations for improving the generalizability of VLA models. Motivated by this finding, we introduce DyGRO-VLA, a two-stage optimization framework that 1) effectively captures cross-task latent representations based on information-theoretic principles, and 2) dynamically refines policy optimization via a mixture-of-RL-residuals. DyGRO-VLA enables the RL optimizer to exploit task-relevant latent information while strategically mitigating adverse interference on the learned representations throughout the optimization process. We evaluate our approach on LIBERO, RoboTwin2 benchmarks, and further validate it on real world, demonstrating consistent improvements over strong baselines under multi-task training and distribution shift.
arXiv:2605.17483v1 Announce Type: new Abstract: Facial Expression Recognition faces two core challenges. The first is class imbalance in public datasets, which skews the learning process and weakens generalization. The second is related to privacy and data collection constraints, which limit the sharing of facial images and restrict the creation of large, balanced datasets. To address these issues, we examine three complementary strategies for constructing privacy-preserving FER datasets in the standard seven discrete facial expression classes setting. Our strategies are: (i) pseudo-labeling large unlabeled face collections with a teacher model under a confidence-thresholding scheme, (ii) prompt-driven synthesis using diffusion models conditioned on demographic attributes, and (iii) task-aware GAN-based expression editing that modifies facial expression while preserving identity and realism. For training and evaluation, we employed widely adopted datasets, including AffectNet, RAF-DB, and FER2013. We utilized the synthetic datasets DigiFace, DCFace, and EmoNet-Face BIG as unlabeled sources for pseudo-labeling. Additionally, we utilized the FFHQ dataset as the source for generative synthesis. The main experiments are conducted using a classic CNN backbone, IR50, and we also explore a more complex architecture, POSTERv1, to assess its feasibility and robustness. Using cross-dataset evaluations, we analyze the trade-offs each strategy presents in curated datasets. The findings demonstrate how synthetic data can effectively substitute or be combined with real datasets to mitigate imbalance and privacy limitations. Code and generated datasets:https://www.github.com/AliAZ98/SyntFER