arXiv subjects

Peng Wu

Publications and source records attributed to Peng Wu.

At least 19 recordsLinked to original sources

arXiv2026

AnomalyCraft-700K: Component-Level Controllable and Verifiable Synthetic Anomalies for Fine-Grained Video Anomaly Understanding

Progress in video anomaly understanding (VAU) has long been limited by inherent deficiencies of real-world anomaly videos, which are hard to collect and offer little control over their content. Synthetic anomaly approaches partially alleviate data scarcity, yet their generation remains largely controlled at the category or prompt level. They also lack component-level verification of video-text consistency and provide insufficient hard normal samples near the normal-anomaly boundary. To address this, we present AnomalyCraft-700K, a component-controllable synthetic anomaly dataset for fine-grained VAU, containing over 40K videos and over 700K task-level textual annotations. From fine-grained semantic components and a progressive three-stage pipeline, we craft anomaly events that are richly detailed, semantically controlled, and temporally structured, and additionally construct per-category hard normal samples to prompt the model to discriminate based on anomaly semantics rather than surface visual cues. Moreover, using the components as verification units, AnomalyCraft-700K further performs component-wise correction of video-text discrepancies introduced during generation, providing reliable annotations with verified cross-modal alignment for six tasks that progress from anomaly detection, through anomaly retrieval and captioning, to fine-grained anomaly reasoning. Evaluations of widely used methods under both traditional and MLLM-based protocols demonstrate that AnomalyCraft-700K serves as an effective source of supervision, from anomaly detection to fine-grained anomaly understanding.

cs.CV↗

arXiv2026

DPC-Net: Dual-Prior Collaborative Network for All-in-One Image Restoration

All-in-One Image Restoration (AiOIR) aims to handle diverse degradations within a unified model. However, existing methods often overlook image semantics in degradation modeling and lack low-level visual priors during reconstruction, leading to structural distortions and semantic inconsistencies. To address these issues, we propose a novel Dual-Prior Collaborative Network (DPC-Net), which achieves high-quality restoration by jointly exploiting degradation-semantic coupled priors and low-level visual priors. Specifically, degraded images are fed into a Degradation-Aware Network (DAN) to extract degradation-semantic coupled features. To this end, a Vision-Language Model (VLM) supervises DAN by constraining its features distribution, introducing image semantics into the encoding of degradation patterns. A Degradation-Semantic Modulation Module (DSMM) further translates this guidance into degradation-semantic coupling and propagates coupled representations to the decoder. During decoding, knowledge bases provide low-level visual priors, and the Dual-Prior Collaborative Reconstruction Module (DPCR) integrates dual-prior information to guide degradation removal while preserving structure and semantics, producing high-fidelity restored images. Extensive experiments on multiple restoration benchmarks demonstrate that DPC-Net achieves superior performance against state-of-the-art AiOIR methods.

cs.CV↗

arXiv2026

Rethinking Auxiliary Modalities in Multimodal Zero-shot Anomaly Detection: From Semantic Fusion to Conditional Modulation

Recent foundation model-based methods have endowed RGB images with strong zero-shot anomaly detection (ZSAD) through vision-language pretraining. However, RGB observations alone remain limited in perceiving anomalies dominated by geometric deformation, depth variation, or subtle surface changes. Auxiliary modalities can provide complementary structural information, but existing multimodal methods typically fuse them directly into a shared semantic space, which may disturb the text-aligned anomaly semantics established by RGB foundation models and often requires modality-specific architectures. To address this issue, we propose a plug-and-play auxiliary-conditioned enhancement framework for zero-shot anomaly detection. Instead of reconstructing a joint multimodal anomaly semantic space, our framework preserves the original RGB image-text anomaly matching pathway and uses auxiliary observations as conditional signals for RGB feature refinement, allowing auxiliary modalities to seamlessly enhance existing RGB-based zero-shot anomaly detectors. Specifically, a lightweight meta-learning module takes global RGB and auxiliary representations as input and generates sample-adaptive low-rank residual updates to determine how RGB features should be refined. We further construct uncertainty-aware spatial modulation from the initial RGB anomaly response and auxiliary reliability, which determines where local residual updates are strengthened or suppressed. This global-to-local conditional modulation enables selective multimodal enhancement while preserving the original RGB anomaly semantics. Extensive experiments on MVTec 3D-AD and Eyecandies demonstrate that our framework consistently improves multiple popular RGB-based zero-shot anomaly detectors, achieving state-of-the-art performance for multimodal zero-shot anomaly detection.

cs.CV↗

arXiv2026

Once Poisoned, Arbitrarily Controlled: A Programmable Backdoor in VLMs

Existing vision-language model (VLM) backdoors are usually treated as static vulnerabilities: one-to-one and N-to-N attacks bind one or more triggers to a finite set of targets before victim training. This assumption substantially underestimates the threat. We show that a single poisoning phase can implant a programmable backdoor into a VLM, allowing an attacker to choose previously unseen target-caption semantics at inference time and synthesize corresponding stealthy triggers on demand. Unlike fixed-mapping attacks, the proposed any-to-any caption-control paradigm decouples post-training target selection from poisoning, enabling dynamic control of target captions without retraining the VLM. Our method has two components. First, a heuristic poisoning strategy exposes the model to diverse trigger-caption pairs, encouraging it to learn a general trigger-as-instruction rule rather than memorize a specific backdoor pattern. Second, a feature-space trigger steganography method maps any attacker-specified target caption to a stealthy visual trigger, implemented as either a norm-controlled perturbation or a non-semantic patch. Once inserted into arbitrary images, these triggers cause the poisoned VLM to generate outputs semantically aligned with the chosen target caption, even when the target was unseen during poisoning. Extensive experiments show that our attack achieves high any-to-any caption-control success rates, preserves clean model utility, and remains effective under several classical backdoor defenses.

cs.CV↗

arXiv2026

Quark matter at finite temperature and proto-quark stars with the axion effects in SU(3) Nambu-Jona-Lasinio model

We investigate the thermodynamical properties of strange quark matter (SQM) and proto-quark stars (PQSs) within the SU(3) Nambu-Jona-Lasinio (NJL) model at finite temperature, specifically incorporating the effects of axion fields and vector interactions. Our results demonstrate that these interactions significantly influence the equation of state (EoS), constituent quark masses, entropy density, and the maximum star mass of PQSs at the isentropic stages along the star evolution line. Furthermore, we reveal a distinct thermodynamic signature in the early evolution: the presence of trapped neutrinos leads to a substantial increase in electron number density while simultaneously suppressing the core temperature compared to the neutrino-free case. These findings may highlight the crucial role of the axion effects, flavor-dependent vector interactions, and particle composition in determining the observable properties of compact stars at finite temperature.

nucl-th↗

arXiv2026

OmniPack: Unified Token Compression for Efficient Omni-modal Large Language Models

Omni-modal large language models (Omni-LLMs) have achieved remarkable performance on audio-visual understanding tasks, but processing long and highly redundant visual and audio token sequences incurs substantial computational overhead, demanding aggressive token compression for efficient deployment. Existing methods often degrade at low token budgets: pre-LLM compression may discard structurally important and globally distributed evidence, whereas inner-LLM compression often underexploits query-conditioned audio-visual collaboration. To address these limitations, we propose OmniPack, a training-free framework that coordinates structural compression before the LLM with task-relevant semantic refinement within the LLM. Before the LLM, OmniPack removes structural redundancy through modality-specific importance, global coverage, and similarity-aware merging. After sufficient multimodal interaction, it further consolidates diverse, task-relevant representations through textual guidance and audio-visual collaboration. Extensive experiments on five benchmarks with three Omni-LLM backbones demonstrate that OmniPack consistently achieves the best performance-efficiency trade-off across diverse retention ratios, outperforming all existing methods. Notably, on Qwen2.5-Omni-7B, OmniPack preserves 98.0% of the original performance while reducing FLOPs to 16.7%, and still retains 92.9% of the original performance with only 6.8% of the original FLOPs.

cs.CV↗

arXiv2026

A black-box-model-enhanced interaction method for water-wave scattering by large group of arbitrary-shaped ice floes in Arctic route planning

This study develops an enhanced interaction (EI) method for efficient prediction of the water-wave field among a large group of ice floes in Arctic route planning. A novel black-box model, termed the wave component detection (WCD) method, is proposed for constructing the diffraction transfer matrix (DTM) within the framework of interaction theory. The DTM, which is conventionally mathematically intractable for three-dimensional ice floes with arbitrarily complex geometry, can now be determined using this readily implementable and universally applicable approach. Without loss of generality, four ice-floe shapes are taken as example models to demonstrate the capability of the EI method. Operation rules are recommended for the practical implementation of the EI method. The error range of the EI method is identified in scenarios with multiple ice floes of different sizes and distances.The super-high efficiency of the EI method is demonstrated in cases involving an ultra-large group of ice floes. It takes less than 1.5 hours to calculate wave amplitudes at 160,000 locations in the wave field of 1,800 ice floes (based on 1,440,000 boundary elements) on an ordinary personal computer with a 2017-released CPU. Based on the wave field predicted by the EI method, users can take advantage of the wave-sheltering effect of the ice floes to optimize routes. For demonstration, the dynamic programming strategy is used to recommend optimized navigation routes among 1561 ice floes of mixed shapes. The average wave amplitude the ship encounters can be reduced to about half of the incident wave amplitude.

physics.ao-ph↗

arXiv2026

Temperature fluctuations in a realistic Polyakov-loop extended Nambu--Jona-Lasinio Model along the freeze-out line

We extend and reparameterize the Polyakov--Nambu--Jona-Lasinio (PNJL) model to reproduce lattice simulation data at zero baryon chemical potential and to position its critical endpoint (CEP) within the BES-II experimental energy range. Using this realistic PNJL model, we investigate the behavior of the second-order temperature cumulant $C_2$ along the freeze-out line, aiming to understand the non-monotonic energy dependence of the two-particle transverse momentum correlation $C_{p_T}$ observed by the STAR Collaboration. Our results show a distinct dip structure in $C_2$ near the CEP and the first-order phase boundary on the phase diagram. Along the experimentally extracted freeze-out line, the dip minimum occurs around 7.7 GeV, and its trend is consistent with that observed by STAR, suggesting that the non-monotonic dependence of $C_{p_T}$ may be related to the CEP. Our results also indicate that cumulant ratios such as $C_3/C_2^2$ or $C_4/C_2^3$ eliminate the influence of initial volume fluctuations and may better reveal the underlying critical fluctuations. Further verification could be pursued through hydrodynamic or transport simulations that incorporate critical dynamics. These results and predictions might provide an a priori theoretical basis for future experimental measurements of higher-order event-mean transverse momentum fluctuations.

nucl-th↗

arXiv2026

Model Merging to Evolution: Parameter Space Exploration for Expert Models

Model merging integrates the capabilities of multiple expert models to create strong models for multiple tasks without additional training, thereby reducing computational resource requirements. However, existing methods operate within the convex combination space of expert models, failing to explore high-performance regions outside this space. This paper proposes the MERGEvolve framework, which unifies model merging and evolution within an evolution strategy by treating the merged model as the initialization for evolutionary exploration of the parameter space. During the merging phase, expert models act as deterministic sources to build a strong initial point. The evolution phase then explores the parameter space using random noise. Theoretical analysis shows that MERGEvolve explores regions outside the convex combination space. Extensive experiments on single-task and multi-task benchmarks demonstrate that MERGEvolve consistently achieves performance competitive with advanced model merging baselines. Ablation studies confirm that a high-quality initial point is critical for efficient exploration of the parameter space.

cs.NE↗

arXiv2026

PhysGuard: Fisher-Guided Gradient Projection for Sim-to-Real Neural PDE Surrogates

Neural operator models trained on simulation data often lose accuracy when applied to experimental measurements due to the sim-to-real gap. Standard fine-tuning with limited real data can reduce this gap, but it may also damage the core physics-relevant representations learned during pretraining. Although knowledge-preserving adaptation has been widely investigated in vision or language tasks, it remains unclear whether these methods are suitable for neural operators whose architectures and protected knowledge are fundamentally different. Neural operators need to preserve core-scale physical structures rather than semantic or visual features. We propose PhysGuard, a physics-preserving framework for accurate sim-to-real adaptation of neural operators. Specifically, PhysGuard uses the empirical Fisher Information Matrix computed on simulation data to identify physics-critical parameter directions, then restricts fine-tuning updates to directions that do not interfere with them. A layer-wise Gram-matrix formulation makes this efficient for models with millions of parameters, while an adaptive threshold automatically determines the protected subspace size. A spectral probe experiment shows that the dominant Fisher directions are strongly associated with low-frequency output structures. Experiments on benchmark across four neural operator architectures and different physical systems show that PhysGuard performs strongly on most evaluation metrics compared to baselines. The benefits are most evident under severe domain shift, where it reduces low-frequency error by up to 32\% compared to standard fine-tuning while maintaining adaptability. Our code is available at https://github.com/ZhouChaunge/PhysGuard.

cs.LG↗

arXiv2026

Counterfactually Safe Reinforcement Learning

Reinforcement learning algorithms are generally designed to maximize the expected return across a population. However, a policy that is optimal on average may be suboptimal for certain individuals, leading to potential safety concerns. To address this, we first formalize the notion of individual harm from a counterfactual perspective and define harm as the event in which a chosen action results in a strictly worse outcome than a baseline alternative. We then propose a general two-stage procedure for learning policies that maximize the expected return while accounting for individual harm. We further establish the finite-sample properties of the learned policy, derive an upper bound on its sub-optimality gap, and show that the harm rate remains well-controlled. Numerical experiments on both simulated and real-world datasets demonstrate the effectiveness of the proposed approach.

stat.ML↗

arXiv2026

Diagnosing Training Inference Mismatch in LLM Reinforcement Learning

Modern LLM RL systems separate rollout generation from policy optimization. These two stages are expected to produce token probabilities that match exactly. However, implementation differences can make them assign different values to the same sequence under the same model weights, inducing Training-Inference Mismatch (TIM). TIM is difficult to inspect because it is entangled with off-policy drift and common stabilization mechanisms. In this work, we isolate TIM in a zero-mismatch diagnostic setting (VeXact), and show that small token-level numerical disagreements can independently cause training collapse. We further show that TIM changes the effective optimization problem, and identify a set of remedies that could mitigate TIM. Our results suggest that TIM is not benign numerical noise, but a systems-level perturbation that should be treated as a first-order factor in analyzing LLM RL stability.

cs.LG↗

arXiv2026

Double/debiased machine learning of quantile treatment effects on long-term outcomes in clinical trials

Long-term outcomes are often unavailable in randomized clinical trials, although short-term surrogate outcomes are commonly observed. External observational data may contain the long-term outcome, but causal comparisons based on such data alone are vulnerable to confounding. Existing surrogate-based data integration methods for long-term outcomes have focused primarily on average treatment effects. We study estimation of quantile treatment effects for long-term outcomes in the trial population by combining randomized trial data with external observational data. Under treatment randomization, positivity, and a surrogate-based transportability assumption, we establish identification and develop a doubly robust estimator for inference. The estimator accommodates flexible machine learning methods for nuisance estimation, remains consistent if either the score-related or outcome regression-related nuisance functions are consistently estimated, and is asymptotically normal under regularity conditions. Simulation and real-data results demonstrate that the proposed method performs well in finite samples and can reveal heterogeneous long-term treatment effects across quantiles.

math.ST↗

arXiv2026

Adaptive Influence-Based Borrowing Framework for Improving Treatment Effect Estimation in RCTs Using External Controls

Randomized controlled trials (RCTs) often suffer from limited sample sizes due to high costs and lengthy recruitment periods, compromising precision in treatment effect estimation. External real-world control data offer a valuable opportunity for augmentation, but na\"ive integration may introduce bias without careful compatibility assessment. This paper presents a practical tutorial on the adaptive influence-based borrowing framework~\citep{Yang-etal2026}, which addresses this challenge through a principled, individual-level borrowing strategy. The core intuition is straightforward: rather than indiscriminately pooling all external controls (ECs), the framework first asks how much each external patient would perturb the outcome model fitted using RCT controls. External patients whose inclusion barely changes this model are deemed comparable and prioritized for borrowing, whereas those who substantially shift it are flagged as potentially incompatible. This individual-level compatibility metric, based on the influence score, is then used to construct a sequence of nested candidate subsets of ECs, from which the optimal subset is selected by minimizing the mean squared error of the treatment effect estimator, balancing the competing risks of bias from over-borrowing and imprecision from under-borrowing. When systematic differences between ECs and RCT controls are substantial, an optional outcome calibration step can align the two groups before influence-based selection proceeds. We provide a clear, step-by-step workflow with emphasis on methodological intuition, practical considerations, and visualization, thereby offering a principled, transparent, and practical method for leveraging ECs when RCTs alone are underpowered. Implementation is supported by an accompanying \texttt{R} package InfluenceBorrowing.

stat.AP↗

arXiv2026

Probabilities of Causation for Continuous Outcomes: Bounds and Identification

The probability of necessity (PN), which quantifies the probability that an observed event would not have occurred in the absence of the treatment, is a central estimand in attribution analysis. While PN has been extensively studied for binary outcomes and has recently been developed for ordinal outcomes, a formal framework for continuous outcomes remains underdeveloped. To address this gap, we propose the general probability of necessity (GPN) for continuous outcomes, a setting that is substantially more challenging than the binary and ordinal cases. Rather than imposing strong identifiability assumptions, we adopt a partial identification perspective and derive sharp lower and upper bounds under standard assumptions of ignorability and monotonicity. We further introduce a copula-based framework that exploits dependence information between potential outcomes to tighten these bounds. Simulation studies and real-world applications demonstrate the effectiveness of our method.

stat.ME↗

arXiv2026

Improving Treatment Effect Estimation in Trials through Adaptive Borrowing of External Controls

Randomized controlled trials (RCTs) often suffer from limited inferential efficiency in estimating treatment effects due to their small sample sizes. In recent years, incorporating external controls (ECs) has gained increasing attention as an effective way to augment small RCTs and thereby enhance estimation efficiency. However, ECs are not always comparable to RCTs, and direct borrowing without careful evaluation can introduce substantial bias and, paradoxically, undermine the accuracy of treatment effect estimation. In this paper, we propose a novel adaptive influence-based sample borrowing framework to improve average treatment effect (ATE) estimation in RCTs. The framework quantifies the ``comparability'' of each sample in ECs using influence functions and identifies the optimal subset of ECs that minimizes the mean squared error of the ATE estimator. The proposed framework is assumption-lean regarding the distribution of ECs and is robust to outliers, making it broadly applicable across diverse settings. Moreover, we develop an outcome calibration method to improve the data utilization efficiency of ECs, further strengthening the adaptive influence-based sample-borrowing framework. We demonstrate the effectiveness of the proposed method using both simulated and real-world datasets.

stat.ME↗

arXiv2026

Efficient Matrix Implementation for Rotary Position Embedding

Rotary Position Embedding (RoPE) has become a core component of modern Transformer architectures across language, vision, and 3D domains. However, existing implementations rely on vector-level split and merge operations that introduce non-negligible computational overhead, often overlooked in attention optimization. The problem is further amplified in multi-dimensional settings (e.g., 2D and 3D RoPE), where additional vector operations and uneven feature partitions degrade hardware utilization. To overcome these limitations, we propose RoME (Rotary Matrix position Embedding), a mathematically equivalent yet computationally efficient reformulation of RoPE that replaces vector operations with unified matrix transformations. RoME eliminates dimension-specific operations, simplifies implementation, and enables fused parallel execution across Cube and Vector units on modern NPUs. Experiments show that RoME delivers substantial acceleration at both the operator and full-model levels. The implementation is available at https://gitcode.com/cann/ops-transformer/blob/master/experimental/posembedding/rope_matrix/README.md.

cs.LG↗

arXiv2026

Towards Video Anomaly Detection from Event Streams: A Baseline and Benchmark Datasets

Event-based vision, characterized by low redundancy, focus on dynamic motion, and inherent privacy-preserving properties, naturally fits the demands of video anomaly detection (VAD). However, the absence of dedicated event-stream anomaly detection datasets and effective modeling strategies has significantly hindered progress in this field. In this work, we take the first major step toward establishing event-based VAD as a unified research direction. We first construct multiple event-stream based benchmarks for video anomaly detection, featuring synchronized event and RGB recordings. Leveraging the unique properties of events, we then propose an EVent-centric spatiotemporal Video Anomaly Detection framework, namely EWAD, with three key innovations: an event density aware dynamic sampling strategy to select temporally informative segments; a density-modulated temporal modeling approach that captures contextual relations from sparse event streams; and an RGB-to-event knowledge distillation mechanism to enhance event-based representations under weak supervision. Extensive experiments on three benchmarks demonstrate that our EWAD achieves significant improvements over existing approaches, highlighting the potential and effectiveness of event-driven modeling for video anomaly detection. The benchmark datasets will be made publicly available.

cs.CV↗