arXiv subjects

Zhao Wang

Publications and source records attributed to Zhao Wang.

At least 55 records · Page 3Linked to original sources

arXiv2026

SAMRI: Segment Any MRI

Summary: SAMRI is an MRI-specialized adaptation of the Segment Anything Model achieving superior whole-body MRI segmentation, particularly for small and clinically critical structures, through box and point prompts for rapid annotation. Purpose: Existing SAM adaptations treat MRI as a generic modality, overlooking variable tissue contrast, intensity inhomogeneity, and clinically important small structures. We propose an MRI-specialized foundation model with strong whole-body segmentation and zero-shot generalization for direct use on any MRI annotation task. Methods: SAMRI fine-tunes only the mask decoder of SAM (ViT-B/16), keeping encoders frozen to preserve pretrained representations and eliminate redundant passes-reducing training time by 94%, trainable parameters by 96%, and FLOPs by ~99% versus full-model retraining. Training used 1.1 million 2D slice-mask pairs from 30 datasets spanning 47 targets, T1/T2/FLAIR/DWI contrasts, and whole-body anatomy, with focal-Dice loss and bounding-box (with optional point) prompts. Sizes were stratified by mask area (small: <0.5%; medium: 0.5-3.5%; large: >3.5%), and significance assessed by the Wilcoxon signed-rank test. Results: SAMRI with box+point prompts achieved mean DSC 0.87 +/- 0.11 across 47 targets, outperforming MedSAM (0.74 +/- 0.24) by 17.6% (p < 0.05), with largest gains for small (+42.4%) and medium (+26.9%) structures. On six zero-shot datasets, SAMRI achieved mean DSC 0.85, outperforming baselines. Inference requires only ~4.5 GB VRAM through an interactive interface on standard hardware. Conclusion: Decoder-only fine-tuning on a large, MRI-specific corpus delivers superior whole-body segmentation with strong zero-shot generalization, particularly for small and clinically salient structures. Public code, pretrained models, and an interactive interface make SAMRI deployable for MRI segmentation research and clinical workflows.

eess.IV↗

arXiv2026

Seeing is Understanding: Unlocking Causal Attention into Modality-Mutual Attention for Multimodal LLMs

Recent Multimodal Large Language Models (MLLMs) have demonstrated significant progress in perceiving and reasoning over multimodal inquiries, ushering in a new research era for foundation models. However, vision-language misalignment in MLLMs has emerged as a critical challenge, where the textual responses generated by these models are not factually aligned with the given text-image inputs. Existing efforts to address vision-language misalignment have focused on developing specialized vision-language connectors or leveraging visual instruction tuning from diverse domains. In this paper, we tackle this issue from a fundamental yet unexplored perspective by revisiting the core architecture of MLLMs. Most MLLMs are typically built on decoder-only LLMs consisting of a causal attention mechanism, which limits the ability of the earlier modalities (e.g., images) to incorporate information from the latter modalities (e.g., text). To address this problem a MLLM that unlocks causal attention into our proposed modality-mutual attention (MMA) to enable image tokens to attend to text tokens. This simple yet effective design allows MMA to achieve state-of-the-art performance in 12 multimodal understanding benchmarks (+6.2% on average across 3 LLMs backbones) without introducing additional parameters. Our MMA design is intended to be generic, allowing for applications across various modalities, and scalable to accommodate diverse multimodal scenarios.

cs.CV↗

arXiv2026

Tstars-Tryon 1.0: Robust and Realistic Virtual Try-On for Diverse Fashion Items

Recent advances in image generation and editing have opened new opportunities for virtual try-on. However, existing methods still struggle to meet complex real-world demands. We present Tstars-Tryon 1.0, a commercial-scale virtual try-on system that is robust, realistic, versatile, and highly efficient. First, our system maintains a high success rate across challenging cases like extreme poses, severe illumination variations, motion blur, and other in-the-wild conditions. Second, it delivers highly photorealistic results with fine-grained details, faithfully preserving garment texture, material properties, and structural characteristics, while largely avoiding common AI-generated artifacts. Third, beyond apparel try-on, our model supports flexible multi-image composition (up to 6 reference images) across 8 fashion categories, with coordinated control over person identity and background. Fourth, to overcome the latency bottlenecks of commercial deployment, our system is heavily optimized for inference speed, delivering near real-time generation for a seamless user experience. These capabilities are enabled by an integrated system design spanning end-to-end model architecture, a scalable data engine, robust infrastructure, and a multi-stage training paradigm. Extensive evaluation and large-scale product deployment demonstrate that Tstars-Tryon1.0 achieves leading overall performance. To support future research, we also release a comprehensive benchmark. The model has been deployed at an industrial scale on the Taobao App, serving millions of users with tens of millions of requests.

cs.CV↗

arXiv2026

OceanPile: A Large-Scale Multimodal Ocean Corpus for Foundation Models

The vast and underexplored ocean plays a critical role in regulating global climate and supporting marine biodiversity, yet artificial intelligence has so far delivered limited impact in this domain due to a fundamental data bottleneck. Specifically, ocean data are highly fragmented across disparate sources and inherently exhibit multi-modal, high-noise, and weakly labeled characteristics, lacking unified schemas and semantic alignment. Although Multimodal Large Language Models (MLLMs) have achieved remarkable success in general domains, their application to ocean science remains severely constrained by the absence of large-scale, well-aligned multimodal datasets tailored to marine environments. To bridge this gap, we introduce OceanPile, a large-scale multimodal corpus designed for ocean foundation models. It comprises three key components: OceanCorpus, a unified collection integrating sonar data, underwater imagery, marine science visuals, and scientific text from diverse authoritative sources; OceanInstruction, a high-quality instruction dataset synthesized via a novel pipeline guided by a hierarchical Ocean Concept Knowledge Graph; and OceanBenchmark, a manually curated evaluation benchmark for rigorous assessment. We establish a multi-stage quality control process to ensure scientific validity and alignment across modalities. Experimental validation demonstrates significant performance improvements for models trained on our data. All datasets are publicly released to advance the field of marine artificial intelligence and empower domain-specific MLLMs.

cs.MM↗

arXiv2026

How Personal Characteristics Shape User Exploration of Diverse Movie Recommendations with a LLM-Based Multi-Agent System

Diversity is an important evaluation criterion for recommender systems beyond accuracy, yet users differ in their willingness to engage with novel and diverse content. In this work, we investigate how a Large Language Model (LLM)-based multi-agent system supports users' exploration of diverse recommendations, and how individual characteristics shape user experiences. We conducted a between-subjects user study (N = 100) comparing a single-agent system (baseline) with a multi-agent system for movie recommendations. We measured Perceived Accuracy, diversity, novelty, and overall rating, and examined the influence of personal characteristics, including personality traits, demographics, GenAI recommendation experience, and GenAI skepticism. Results show that the multi-agent system significantly increases Perceived Novelty and Shannon Diversity. Conscientiousness is positively associated with Perceived Accuracy and diversity, whereas extraversion is negatively associated with Perceived Diversity. Prior experience with GenAI-based recommendations is positively associated with Shannon Diversity, while skepticism toward GenAI is negatively associated with it. We also observe significant interaction effects between system design and user characteristics. These findings highlight the importance of personality-aware conversational recommender systems and caution against one-size-fits-all multi-agent designs.

cs.HC↗

arXiv2026

Reasoning-Aware AIGC Detection via Alignment and Reinforcement

The rapid advancement and widespread adoption of Large Language Models (LLMs) have elevated the need for reliable AI-generated content (AIGC) detection, which remains challenging as models evolve. We introduce AIGC-text-bank, a comprehensive multi-domain dataset with diverse LLM sources and authorship scenarios, and propose REVEAL, a detection framework that generates interpretable reasoning chains before classification. Our approach uses a two-stage training strategy: supervised fine-tuning to establish reasoning capabilities, followed by reinforcement learning to improve accuracy, improve logical consistency, and reduce hallucinations. Extensive experiments show that REVEAL achieves state-of-the-art performance across multiple benchmarks, offering a robust and transparent solution for AIGC detection. The project is open-source at https://aka.ms/reveal

cs.AI↗

arXiv2026

Unposed-to-3D: Learning Simulation-Ready Vehicles from Real-World Images

Creating realistic and simulation-ready 3D assets is crucial for autonomous driving research and virtual environment construction. However, existing 3D vehicle generation methods are often trained on synthetic data with significant domain gaps from real-world distributions. The generated models often exhibit arbitrary poses and undefined scales, resulting in poor visual consistency when integrated into driving scenes. In this paper, we present Unposed-to-3D, a novel framework that learns to reconstruct 3D vehicles from real-world driving images using image-only supervision. Our approach consists of two stages. In the first stage, we train an image-to-3D reconstruction network using posed images with known camera parameters. In the second stage, we remove camera supervision and use a camera prediction head that directly estimates the camera parameters from unposed images. The predicted pose is then used for differentiable rendering to provide self-supervised photometric feedback, enabling the model to learn 3D geometry purely from unposed images. To ensure simulation readiness, we further introduce a scale-aware module to predict real-world size information, and a harmonization module that adapts the generated vehicles to the target driving scene with consistent lighting and appearance. Extensive experiments demonstrate that Unposed-to-3D effectively reconstructs realistic, pose-consistent, and harmonized 3D vehicle models from real-world images, providing a scalable path toward creating high-quality assets for driving scene simulation and digital twin environments.

cs.CV↗

arXiv2026

One-to-More: High-Fidelity Training-Free Anomaly Generation with Attention Control

Industrial anomaly detection (AD) is characterized by an abundance of normal images but a scarcity of anomalous ones. Although numerous few-shot anomaly synthesis methods have been proposed to augment anomalous data for downstream AD tasks, most existing approaches require time-consuming training and struggle to learn distributions that are faithful to real anomalies, thereby restricting the efficacy of AD models trained on such data. To address these limitations, we propose a training-free few-shot anomaly generation method, namely O2MAG, which leverages the self-attention in One reference anomalous image to synthesize More realistic anomalies, supporting effective downstream anomaly detection. Specifically, O2MAG manipulates three parallel diffusion processes via self-attention grafting and incorporates the anomaly mask to mitigate foreground-background query confusion, synthesizing text-guided anomalies that closely adhere to real anomalous distributions. To bridge the semantic gap between the encoded anomaly text prompts and the true anomaly semantics, Anomaly-Guided Optimization is further introduced to align the synthesis process with the target anomalous distribution, steering the generation toward realistic and text-consistent anomalies. Moreover, to mitigate faint anomaly synthesis inside anomaly masks, Dual-Attention Enhancement is adopted during generation to reinforce both self- and cross-attention on masked regions. Extensive experiments validate the effectiveness of O2MAG, demonstrating its superior performance over prior state-of-the-art methods on downstream AD tasks.

cs.CV↗

arXiv2026

StreetForward: Perceiving Dynamic Street with Feedforward Causal Attention

Feedforward reconstruction is crucial for autonomous driving applications, where rapid scene reconstruction enables efficient utilization of large-scale driving datasets in closed-loop simulation and other downstream tasks, eliminating the need for time-consuming per-scene optimization. We present StreetForward, a pose-free and tracker-free feedforward framework for dynamic street reconstruction. Building upon the alternating attention mechanism from Visual Geometry Grounded Transformer (VGGT), we propose a simple yet effective temporal mask attention module that captures dynamic motion information from image sequences and produces motion-aware latent representations. Static content and dynamic instances are represented uniformly with 3D Gaussian Splatting, and are optimized jointly by cross-frame rendering with spatio-temporal consistency, allowing the model to infer per-pixel velocities and produce high-fidelity novel views at new poses and times. We train and evaluate our model on the Waymo Open Dataset, demonstrating superior performance on novel view synthesis and depth estimation compared to existing methods. Furthermore, zero-shot inference on CARLA and other datasets validates the generalization capability of our approach. More visualizations are available on our project page: https://streetforward.github.io.

cs.CV↗

arXiv2026

ABounD: Adversarial Boundary-Driven Few-Shot Learning for Multi-Class Anomaly Detection

Few-shot multi-class industrial anomaly detection identifies diverse defects across multiple categories using a single unified model and limited normal samples. Although vision-language models offer strong generalization, modeling multiple distinct category manifolds concurrently without actual anomalous data causes feature space collapse and cross-class interference. Consequently, existing methods often fail to balance scalability and precision, leading to either isolated single-class retraining or excessively loose decision margins. To address this limitation, we present a one-for-all learning framework called ABounD that unites semantic concept anchoring with geometric boundary optimization. This method employs two lightweight mechanisms to resolve multi-class ambiguity. First, the Dynamic Concept Fusion module generates class-adaptive semantic anchors via query-aware hierarchical calibration, disentangling overlapping category concepts. Second, using these anchors, the Adversarial Boundary Forging module constructs a tight, class-tailored decision margin by synthesizing adversarial boundary-level fence features to prevent cross-class boundary blurring. Optimized in a single stage, ABounD removes the requirement for isolated per-category retraining in few-shot settings. Experiments on seven industrial benchmarks show that the proposed method achieves state-of-the-art detection and localization performance for multi-class few-shot anomaly detection while maintaining low computational costs during training and inference.

cs.CV↗

arXiv2026

Hardware-Software Co-design for 3D-DRAM-based LLM Serving Accelerator

Large language models (LLMs) have been widely deployed for online generative services, where numerous LLM instances jointly handle workloads with fluctuating request arrival rates and variable request lengths. To efficiently execute coexisting compute-intensive and memory-intensive operators, near-memory processing (NMP) based computing paradigm has been extensively proposed. However, existing NMP designs adopt coarse-grained KV cache management and inflexible attention execution flow. Such limitations hinder these proposals from efficiently handling \textit{highly dynamic} LLM serving workloads, limiting their ability to accelerate LLM serving. To tackle these problems, we propose Helios, a Hybrid-bonding-based \uline{L}LM \uline{S}erving accelerator. Helios aims to bridge the fundamental gap between the dynamic nature of KV cache management in LLM serving and the distributed, non-uniform memory abstraction among NMP processing engines (PEs). To this end, we design both the intra-PE execution flow and the inter-PE communication primitives for distributed tiled attention execution. We further propose \textit{spatially-aware} KV cache allocation mechanism to balance the attention workload distribution while minimizing the inter-PE data transfer overhead. Compared with existing GPU/NMP designs, Helios achieves 3.25 times (geomean) speedup and 3.36 times (geomean) better energy efficiency, along with up to 72%/76% P50/P99 time-between-tokens degradation.

cs.AR↗

arXiv2026

Improving Anomaly Detection with Foundation-Model Synthesis and Wavelet-Domain Attention

Industrial anomaly detection faces significant challenges due to the scarcity of anomalous samples and the complexity of real-world anomalies. In this paper, we propose a foundation model-based anomaly synthesis pipeline (FMAS) that generates highly realistic anomalous samples without fine-tuning or class-specific training. Motivated by the distinct frequency-domain characteristics of anomalies, we introduce aWavelet Domain Attention Module (WDAM), which exploits adaptive sub-band processing to enhance anomaly feature extraction. The combination of FMAS and WDAM significantly improves anomaly detection sensitivity while maintaining computational efficiency. Comprehensive experiments on MVTec AD and VisA datasets demonstrate that WDAM, as a plug-and-play module, achieves substantial performance gains against existing baselines.

cs.CV↗

arXiv2026

Graph Neural Network Prediction of Infrared Spectra of Interstellar Polycyclic Aromatic Hydrocarbons

Polycyclic aromatic hydrocarbons (PAHs) are recognized as the primary contributors to the aromatic infrared bands (AIBs) widely observed in space. However, analyzing these AIBs remains challenging because of the immense structural diversity within the PAH family, which makes the computation of reliable reference spectra difficult. To address this, we developed an efficient graph neural network (GNN) framework that can predict PAH absorption spectra up to 10,000 times faster than traditional quantum chemical methods. We evaluated four representative GNN architectures, including graph convolutional network (GCN), graph attention network (GAT), message passing neural network (MPNN), and attentive fingerprint (AFP). The AFP model is found to deliver the best overall performance and is further trained using five different spectral distance metrics as loss functions, among which the Jensen-Shannon divergence yields the most accurate and stable results. The model performs best for PAHs containing 20-40 carbon atoms, while accuracy decreases for larger molecules, reflecting the limited availability of training data. Overall, this framework offers a fast method to generate approximate reference spectra for small- to medium-sized PAHs, supporting future AIB analysis.

astro-ph.GA↗

arXiv2026

Boolean lattice without small rainbow subposets

A Boolean lattice $\mathcal{B}_n=(2^X, \leq)$ is the power set of an $n$-element ground set $X$ equipped with inclusion relation. For two posets $\mathcal{P}$ and $\mathcal{Q}$, we say that $\mathcal{Q}$ contains an \emph{induced copy} of $\mathcal{P}$ if there exists an injection $f : \mathcal{P} \to \mathcal{Q}$ such that $f(X) \le f(Y)$ if and only if $X \le Y$ in $\mathcal{P}$. A $k$-coloring is exact if all colors are used at least once. For posets $\mathcal{Q}$ and $\mathcal{P}$, the \emph{Boolean Gallai-Ramsey number} $\operatorname{GR}_{k}(\mathcal{Q}:\mathcal{P})$ is defined as the smallest $n$ such that any exact $k$-coloring of the sets in $\mathcal{B}_n$ contains either a rainbow induced copy of $\mathcal{Q}$ or a monochromatic induced copy of $\mathcal{P}$ and the \emph{Boolean rainbow Ramsey number} $\operatorname{RR}(\mathcal{Q}:\mathcal{P})$ is defined as the smallest $n$ such that any coloring of the sets in $\mathcal{B}_n$ contains either a rainbow induced copy of $\mathcal{Q}$ or a monochromatic induced copy of $\mathcal{P}$. In this paper, we first study the structural properties of exact $k$-colorings of the sets in Boolean lattice without rainbow induced copy of small posets. As the application of these results, we give exact values and some bounds of Boolean Gallai-Ramsey numbers and Boolean rainbow Ramsey numbers, which improve a result of Chen, Cheng, Li, and Liu in 2020 and give an answer of a question proposed by Chang, Gerbner, Li, Methuku, Nagy, Patkós, and Vizer in 2022.

math.CO↗

arXiv2026

Entropy-Guided k-Guard Sampling for Long-Horizon Autoregressive Video Generation

Autoregressive (AR) architectures have achieved significant successes in LLMs, inspiring explorations for video generation. In LLMs, top-p/top-k sampling strategies work exceptionally well: language tokens have high semantic density and low redundancy, so a fixed size of token candidates already strikes a balance between semantic accuracy and generation diversity. In contrast, video tokens have low semantic density and high spatio-temporal redundancy. This mismatch makes static top-k/top-p strategies ineffective for video decoders: they either introduce unnecessary randomness for low-uncertainty regions (static backgrounds) or get stuck in early errors for high-uncertainty regions (foreground objects). Prediction errors will accumulate as more frames are generated and eventually severely degrade long-horizon quality. To address this, we propose Entropy-Guided k-Guard (ENkG) sampling, a simple yet effective strategy that adapts sampling to token-wise dispersion, quantified by the entropy of each token's predicted distribution. ENkG uses adaptive token candidate sizes: for low-entropy regions, it employs fewer candidates to suppress redundant noise and preserve structural integrity; for high-entropy regions, it uses more candidates to mitigate error compounding. ENkG is model-agnostic, training-free, and adds negligible overhead. Experiments demonstrate consistent improvements in perceptual quality and structural stability compared to static top-k/top-p strategies.

cs.CV↗

arXiv2026

ProRAG: Process-Supervised Reinforcement Learning for Retrieval-Augmented Generation

Reinforcement learning (RL) has become a promising paradigm for optimizing Retrieval-Augmented Generation (RAG) in complex reasoning tasks. However, traditional outcome-based RL approaches often suffer from reward sparsity and inefficient credit assignment, as coarse-grained scalar rewards fail to identify specific erroneous steps within long-horizon trajectories. This ambiguity frequently leads to "process hallucinations", where models reach correct answers through flawed logic or redundant retrieval steps. Although recent process-aware approaches attempt to mitigate this via static preference learning or heuristic reward shaping, they often lack the on-policy exploration capabilities required to decouple step-level credit from global outcomes. To address these challenges, we propose ProRAG, a process-supervised reinforcement learning framework designed to integrate learned step-level supervision into the online optimization loop. Our framework consists of four stages: (1) Supervised Policy Warmup to initialize the model with a structured reasoning format; (2) construction of an MCTS-based Process Reward Model (PRM) to quantify intermediate reasoning quality; (3) PRM-Guided Reasoning Refinement to align the policy with fine-grained process preferences; and (4) Process-Supervised Reinforcement Learning with a dual-granularity advantage mechanism. By aggregating step-level process rewards with global outcome signals, ProRAG provides precise feedback for every action. Extensive experiments on five multi-hop reasoning benchmarks demonstrate that ProRAG achieves superior overall performance compared to strong outcome-based and process-aware RL baselines, particularly on complex long-horizon tasks, validating the effectiveness of fine-grained process supervision. The code and model are available at https://github.com/lilinwz/ProRAG.

cs.AI↗

arXiv2026

Gallai-Ramsey Numbers for $\ell$-Connected Graphs

Given a nonempty graph $G$, a collection of nonempty graphs $\cal{H}$, and a positive integer $k$, the Gallai-Ramsey number $\mathrm{gr}_k(G:\mathcal{H})$ is defined to be the minimum positive integer $n$ such that every exact $k$-edge-coloring of a complete graph $K_n$ contains either a rainbow copy of $G$ or a monochromatic copy of some element in $\mathcal{H}$. In this paper, we obtain some exact values and general lower and upper bounds for $\mathrm{gr}_k(G:\mathcal{F}^\ell)$, where $\mathcal{F}^\ell$ is the set of $\ell$-connected graphs and $G\in\{P_5, K_{1,3}\}$.

math.CO↗

arXiv2026

NextFlow: Unified Sequential Modeling Activates Multimodal Understanding and Generation

We present NextFlow, a unified decoder-only autoregressive transformer trained on 6 trillion interleaved text-image discrete tokens. By leveraging a unified vision representation within a unified autoregressive architecture, NextFlow natively activates multimodal understanding and generation capabilities, unlocking abilities of image editing, interleaved content and video generation. Motivated by the distinct nature of modalities - where text is strictly sequential and images are inherently hierarchical - we retain next-token prediction for text but adopt next-scale prediction for visual generation. This departs from traditional raster-scan methods, enabling the generation of 1024x1024 images in just 5 seconds - orders of magnitude faster than comparable AR models. We address the instabilities of multi-scale generation through a robust training recipe. Furthermore, we introduce a prefix-tuning strategy for reinforcement learning. Experiments demonstrate that NextFlow achieves state-of-the-art performance among unified models and rivals specialized diffusion baselines in visual quality.

cs.CV↗