Searcharxiv⌕ Search

arXiv subjects

Jing Zhang

Publications and source records attributed to Jing Zhang.

At least 145 records · Page 8Linked to original sources
arXiv2026

Sparse-RL: Breaking the Memory Wall in LLM Reinforcement Learning via Stable Sparse Rollouts

Reinforcement Learning (RL) has become essential for eliciting complex reasoning capabilities in Large Language Models (LLMs). However, the substantial memory overhead of storing Key-Value (KV) caches during long-horizon rollouts acts as a critical bottleneck, often prohibiting efficient training on limited hardware. While existing KV compression techniques offer a remedy for inference, directly applying them to RL training induces a severe policy mismatch, leading to catastrophic performance collapse. To address this, we introduce Sparse-RL empowers stable RL training under sparse rollouts. We show that instability arises from a fundamental policy mismatch among the dense old policy, the sparse sampler policy, and the learner policy. To mitigate this issue, Sparse-RL incorporates Sparsity-Aware Rejection Sampling and Importance-based Reweighting to correct the off-policy bias introduced by compression-induced information loss. Experimental results show that Sparse-RL reduces rollout overhead compared to dense baselines while preserving the performance. Furthermore, Sparse-RL inherently implements sparsity-aware training, significantly enhancing model robustness during sparse inference deployment. The corresponding training data and code are publicly available on the repository.

cs.LG↗
arXiv2026

Wanderland: Geometrically Grounded Simulation for Open-World Embodied AI

Reproducible closed-loop evaluation remains a major bottleneck in Embodied AI such as visual navigation. A promising path forward is high-fidelity simulation that combines photorealistic sensor rendering with geometrically grounded interaction in complex, open-world urban environments. Although recent video-3DGS methods ease open-world scene capturing, they are still unsuitable for benchmarking due to large visual and geometric sim-to-real gaps. To address these challenges, we introduce Wanderland, a real-to-sim framework that features multi-sensor capture, reliable reconstruction, accurate geometry, and robust view synthesis. Using this pipeline, we curate a diverse dataset of indoor-outdoor urban scenes and systematically demonstrate how image-only pipelines scale poorly, how geometry quality impacts novel view synthesis, and how all of these adversely affect navigation policy learning and evaluation reliability. Beyond serving as a trusted testbed for embodied navigation, Wanderland's rich raw sensor data further allows benchmarking of 3D reconstruction and novel view synthesis models. Our work establishes a new foundation for reproducible research in open-world embodied AI. Project website is at https://ai4ce.github.io/wanderland/.

cs.CV↗
arXiv2026

Progressive Learning with Anatomical Priors for Reliable Left Atrial Scar Segmentation from Late Gadolinium Enhancement MRI

Cardiac MRI late gadolinium enhancement (LGE) enables non-invasive identification of left atrial (LA) scar, whose spatial distribution is strongly associated with atrial fibrillation (AF) severity and recurrence. However, automatic LA scar segmentation remains challenging due to low contrast, annotation variability, and the lack of anatomical constraints, often leading to non-reliable predictions. Accordingly, our aim was to propose a progressive learning strategy to segment LA scar from LGE images inspired from a clinical workflow. A 3-stage framework based on SwinUNETR was implemented, comprising: 1) a first LA cavity pre-learning model, 2) dual-task model which further learns spatial relationship between LA geometry and scar patterns, and 3) fine-tuning on precise segmentation of the scar. Furthermore, we introduced an anatomy-aware spatially weighted loss that incorporates prior clinical knowledge by constraining scar predictions to anatomically plausible LA wall regions while mitigating annotation bias. Our preliminary results obtained on validation LGE volumes from LASCARQS public dataset after 5-fold cross validation, LA segmentation had Dice score of 0.94, LA scar segmentation achieved Dice score of 0.50, Hausdorff Distance of 11.84 mm, Average Surface Distance of 1.80 mm, outperforming only a one-stage scar segmentation with 0.49, 13.02 mm, 1.96 mm, repectively. By explicitly embedding clinical anatomical priors and diagnostic reasoning into deep learning, the proposed approach improved the accuracy and reliability of LA scar segmentation from LGE, revealing the importance of clinically informed model design.

cs.CV↗
arXiv2026

Heuristic-inspired Reasoning Priors Facilitate Data-Efficient Referring Object Detection

Most referring object detection (ROD) models, especially the modern grounding detectors, are designed for data-rich conditions, yet many practical deployments, such as robotics, augmented reality, and other specialized domains, would face severe label scarcity. In such regimes, end-to-end grounding detectors need to learn spatial and semantic structure from scratch, wasting precious samples. We ask a simple question: Can explicit reasoning priors help models learn more efficiently when data is scarce? To explore this, we first introduce a Data-efficient Referring Object Detection (De-ROD) task, which is a benchmark protocol for measuring ROD performance in low-data and few-shot settings. We then propose the HeROD (Heuristic-inspired ROD), a lightweight, model-agnostic framework that injects explicit, heuristic-inspired spatial and semantic reasoning priors, which are interpretable signals derived based on the referring phrase, into 3 stages of a modern DETR-style pipeline: proposal ranking, prediction fusion, and Hungarian matching. By biasing both training and inference toward plausible candidates, these priors promise to improve label efficiency and convergence performance. On RefCOCO, RefCOCO+, and RefCOCOg, HeROD consistently outperforms strong grounding baselines in scarce-label regimes. More broadly, our results suggest that integrating simple, interpretable reasoning priors provides a practical and extensible path toward better data-efficient vision-language understanding.

cs.CV↗
arXiv2026

PP-OCRv5: A Specialized 5M-Parameter Model Rivaling Billion-Parameter Vision-Language Models on OCR Tasks

The advent of "OCR 2.0" and large-scale vision-language models (VLMs) has set new benchmarks in text recognition. However, these unified architectures often come with significant computational demands, challenges in precise text localization within complex layouts, and a propensity for textual hallucinations. Revisiting the prevailing notion that model scale is the sole path to high accuracy, this paper introduces PP-OCRv5, a meticulously optimized, lightweight OCR system with merely 5 million parameters. We demonstrate that PP-OCRv5 achieves performance competitive with many billion-parameter VLMs on standard OCR benchmarks, while offering superior localization precision and reduced hallucinations. The cornerstone of our success lies not in architectural expansion but in a data-centric investigation. We systematically dissect the role of training data by quantifying three critical dimensions: data difficulty, data accuracy, and data diversity. Our extensive experiments reveal that with a sufficient volume of high-quality, accurately labeled, and diverse data, the performance ceiling for traditional, efficient two-stage OCR pipelines is far higher than commonly assumed. This work provides compelling evidence for the viability of lightweight, specialized models in the large-model era and offers practical insights into data curation for OCR. The source code and models are publicly available at https://github.com/PaddlePaddle/PaddleOCR.

cs.CV↗
arXiv2026

Feature Incremental Clustering with Generalization Bounds

In many learning systems, such as activity recognition systems, as new data collection methods continue to emerge in various dynamic environmental applications, the attributes of instances accumulate incrementally, with data being stored in gradually expanding feature spaces. How to design theoretically guaranteed algorithms to effectively cluster this special type of data stream, commonly referred to as activity recognition, remains unexplored. Compared to traditional scenarios, we will face at least two fundamental questions in this feature incremental scenario. (i) How to design preliminary and effective algorithms to address the feature incremental clustering problem? (ii) How to analyze the generalization bounds for the proposed algorithms and under what conditions do these algorithms provide a strong generalization guarantee? To address these problems, by tailoring the most common clustering algorithm, i.e., $k$-means, as an example, we propose four types of Feature Incremental Clustering (FIC) algorithms corresponding to different situations of data access: Feature Tailoring (FT), Data Reconstruction (DR), Data Adaptation (DA), and Model Reuse (MR), abbreviated as FIC-FT, FIC-DR, FIC-DA, and FIC-MR. Subsequently, we offer a detailed analysis of the generalization error bounds for these four algorithms and highlight the critical factors influencing these bounds, such as the amounts of training data, the complexity of the hypothesis space, the quality of pre-trained models, and the discrepancy of the reconstruction feature distribution. The numerical experiments show the effectiveness of the proposed algorithms, particularly in their application to activity recognition clustering tasks.

math.ST↗
arXiv2026

Omni-I2C: A Holistic Benchmark for High-Fidelity Image-to-Code Generation

We present Omni-I2C, a comprehensive benchmark designed to evaluate the capability of Large Multimodal Models (LMMs) in converting complex, structured digital graphics into executable code. We argue that this task represents a non-trivial challenge for the current generation of LMMs: it demands an unprecedented synergy between high-fidelity visual perception -- to parse intricate spatial hierarchies and symbolic details -- and precise generative expression -- to synthesize syntactically sound and logically consistent code. Unlike traditional descriptive tasks, Omni-I2C requires a holistic understanding where any minor perceptual hallucination or coding error leads to a complete failure in visual reconstruction. Omni-I2C features 1080 meticulously curated samples, defined by its breadth across subjects, image modalities, and programming languages. By incorporating authentic user-sourced cases, the benchmark spans a vast spectrum of digital content -- from scientific visualizations to complex symbolic notations -- each paired with executable reference code. To complement this diversity, our evaluation framework provides necessary depth; by decoupling performance into perceptual fidelity and symbolic precision, it transcends surface-level accuracy to expose the granular structural failures and reasoning bottlenecks of current LMMs. Our evaluation reveals a substantial performance gap among leading LMMs; even state-of-the-art models struggle to preserve structural integrity in complex scenarios, underscoring that multimodal code generation remains a formidable challenge. Data and code are available at https://github.com/MiliLab/Omni-I2C.

cs.CV↗
arXiv2026

LLMIA: An Out-of-the-Box Index Advisor via In-Context Learning with LLMs

Index recommendation is crucial for optimizing database performance. However, existing heuristic- and learning-based methods often rely on inefficient exhaustive search and estimated costs, leading to low efficiency (due to the vast search space) and unsatisfactory actual latency (due to inaccurate estimations). Inspired by the refinement strategies of experienced DBAs-who efficiently identify and iteratively refine indexes with database feedback-we present LLMIA, an out-of-the-box, tuning-free index advisor leveraging large language models (LLMs) through in-context learning for index recommendation. LLMIA injects database expertise into the LLM using a high-quality demonstration pool and comprehensive workload feature extraction, while iteratively incorporating database feedback to guide the index refinement. This design enables LLMIA to emulate the decision-making process of expert DBAs: efficiently recommending and refining indexes for various workloads within just a few interactions with the DBMS. We validate LLMIA with extensive experiments on five standard OLAP benchmarks (TPC-H with different scales, JOB, TPC-DS, SSB), where it consistently outperforms or matches 12 baselines by producing superior index recommendations with minimal database interactions. Additionally, LLMIA demonstrates robust generalization on two real-world commercial workloads, delivering high-quality recommendations without the need for additional adaptation or retraining, highlighting its out-of-the-box capability.

cs.DB↗
arXiv2026

A Flexible GKP-State-Embedded Fault-Tolerant Quantum Computation Configuration Based on a Three-Dimensional Cluster State

The integration of diverse quantum resources and the exploitation of more degrees of freedom provide key operational flexibility for universal fault-tolerant quantum computation. In this work, we propose a flexible Gottesman-Kitaev-Preskill-state-embedded fault-tolerant quantum computation architecture based on a three-dimensional cluster state constructed in polarization, frequency, and orbital angular momentum domains. Specifically, we design optical entanglement generators to produce three diverse entangled pairs, and subsequently construct a three-dimensional cluster state via a beam-splitter network with several time delays. Furthermore, we present a partially squeezed surface-GKP code to achieve fault-tolerant quantum computation and ultimately find the optimal choice of implementing the squeezing gate to give the best fault-tolerant performance (the fault-tolerant squeezing threshold is 11.5 dB). Our scheme is flexible, scalable, and experimentally feasible, providing versatile options for future optical fault-tolerant quantum computation architecture.

quant-ph↗
arXiv2026

TimeAPN: Adaptive Amplitude-Phase Non-Stationarity Normalization for Time Series Forecasting

Non-stationarity is a fundamental challenge in multivariate long-term time series forecasting, often manifested as rapid changes in amplitude and phase. These variations lead to severe distribution shifts and consequently degrade predictive performance. Existing normalization-based methods primarily rely on first- and second-order statistics, implicitly assuming that distributions evolve smoothly and overlooking fine-grained temporal dynamics. To address these limitations, we propose TimeAPN, an Adaptive Amplitude-Phase Non-Stationarity Normalization framework that explicitly models and predicts non-stationary factors from both the time and frequency domains. Specifically, TimeAPN first models the mean sequence jointly in the time and frequency domains, and then forecasts its evolution over future horizons. Meanwhile, phase information is extracted in the frequency domain, and the phase discrepancy between the predicted and ground-truth future sequences is explicitly modeled to capture temporal misalignment. Furthermore, TimeAPN incorporates amplitude information into an adaptive normalization mechanism, enabling the model to effectively account for abrupt fluctuations in signal energy. The predicted non-stationary factors are subsequently integrated with the backbone forecasting outputs through a collaborative de-normalization process to reconstruct the final non-stationary time series. The proposed framework is model-agnostic and can be seamlessly integrated with various forecasting backbones. Extensive experiments on seven real-world multivariate datasets demonstrate that TimeAPN consistently improves long-term forecasting accuracy across multiple prediction horizons and outperforms state-of-the-art reversible normalization methods.

cs.LG↗
arXiv2026

SARMAE: Masked Autoencoder for SAR Representation Learning

Synthetic Aperture Radar (SAR) imagery plays a critical role in all-weather, day-and-night remote sensing applications. However, existing SAR-oriented deep learning is constrained by data scarcity, while the physically grounded speckle noise in SAR imagery further hampers fine-grained semantic representation learning. To address these challenges, we propose SARMAE, a Noise-Aware Masked Autoencoder for self-supervised SAR representation learning. Specifically, we construct SAR-1M, the first million-scale SAR dataset, with additional paired optical images, to enable large-scale pre-training. Building upon this, we design Speckle-Aware Representation Enhancement (SARE), which injects SAR-specific speckle noise into masked autoencoders to facilitate noise-aware and robust representation learning. Furthermore, we introduce Semantic Anchor Representation Constraint (SARC), which leverages paired optical priors to align SAR features and ensure semantic consistency. Extensive experiments across multiple SAR datasets demonstrate that SARMAE achieves state-of-the-art performance on classification, detection, and segmentation tasks. Code and models will be available at https://github.com/MiliLab/SARMAE.

cs.CV↗
arXiv2026

Probing and Bridging Geometry-Interaction Cues for Affordance Reasoning in Vision Foundation Models

What does it mean for a visual system to truly understand affordance? We argue that this understanding hinges on two complementary capacities: geometric perception, which identifies the structural parts of objects that enable interaction, and interaction perception, which models how an agent's actions engage with those parts. To test this hypothesis, we conduct a systematic probing of Visual Foundation Models (VFMs). We find that models like DINO inherently encode part-level geometric structures, while generative models like Flux contain rich, verb-conditioned spatial attention maps that serve as implicit interaction priors. Crucially, we demonstrate that these two dimensions are not merely correlated but are composable elements of affordance. By simply fusing DINO's geometric prototypes with Flux's interaction maps in a training-free and zero-shot manner, we achieve affordance estimation competitive with weakly-supervised methods. This final fusion experiment confirms that geometric and interaction perception are the fundamental building blocks of affordance understanding in VFMs, providing a mechanistic account of how perception grounds action.

cs.CV↗
arXiv2026

Reducing Pilots in Channel Estimation with Predictive Foundation Models

Accurate channel state information (CSI) acquisition is essential for modern wireless systems, which becomes increasingly difficult under large antenna arrays, strict pilot overhead constraints, and diverse deployment environments. Existing artificial intelligence-based solutions often lack robustness and fail to generalize across scenarios. To address this limitation, this paper introduces a predictive-foundation-model-based channel estimation framework that enables accurate, low-overhead, and generalizable CSI acquisition. The proposed framework employs a predictive foundation model trained on large-scale cross-domain CSI data to extract universal channel representations and provide predictive priors with strong cross-scenario transferability. A pilot processing network based on a vision transformer architecture is further designed to capture spatial, temporal, and frequency correlations from pilot observations. An efficient fusion mechanism integrates predictive priors with real-time measurements, enabling reliable CSI reconstruction even under sparse or noisy conditions. Extensive evaluations across diverse configurations demonstrate that the proposed estimator significantly outperforms both classical and data-driven baselines in accuracy, robustness, and generalization capability.

cs.IT↗
arXiv2026

GeoZero: Incentivizing Reasoning from Scratch on Geospatial Scenes

Multimodal large language models (MLLMs) have undergone rapid development in advancing geospatial scene understanding. Recent studies have sought to enhance the reasoning capabilities of remote sensing MLLMs, typically through cold-start training with elaborately curated chain-of-thought (CoT) data. However, this approach not only incurs substantial annotation costs but also introduces human biases that may limit the diversity of model reasoning. To address these challenges, we propose GeoZero, a framework that enables MLLMs to perform geospatial reasoning without any predefined CoT supervision. Specifically, we construct two datasets, GeoZero-Instruct and GeoZero-Hard. GeoZero-Instruct allows the model to acquire preliminary geospatial knowledge through supervised fine-tuning, while GeoZero-Hard stimulates deep reasoning during the subsequent reinforcement learning stage. Furthermore, we introduce Answer-Anchored Group Relative Policy Optimization (A$^2$GRPO), where the reasoning process is regularized by the model's own answers, encouraging diverse yet accurate thinking. Extensive experiments on multiple remote sensing vision-language benchmarks demonstrate that GeoZero not only surpasses existing state-of-the-art methods but also fosters universal emergent reasoning capabilities across diverse geospatial tasks. Code, data, and models are available at https://github.com/MiliLab/GeoZero.

cs.CV↗
arXiv2026

Taming the Long Tail: Denoising Collaborative Information for Robust Semantic ID Generation

Item IDs form the backbone of industrial recommender systems, but suffer from representation instability and poor long-tail generalization in large, dynamic item corpora. Semantic IDs (SIDs) mitigate these issues by enabling knowledge sharing through quantization of item content features. Existing methods attempt to enhance SID expressiveness by incorporating collaborative information with content features; however, they often overlook a critical distinction: unlike relatively uniform content features, user-item interactions are highly skewed, resulting in a significant quality gap in collaborative information between popular and long-tail items. This mismatch leads to two critical limitations: (1) Collaborative Noise Corrupts Behavior-Content Alignment: Behavior-content alignment is a prevailing approach for modeling shared information. However, indiscriminate alignment allows collaborative noise from long-tail items to corrupt their content representations, leading to the loss of critical multimodal information. (2) Collaborative Noise Obscures Critical Behavioral SIDs: When modeling modality-specific information, prior works typically generate multiple behavioral SIDs with equal weights for each item. This equal-weight scheme fails to reflect the varying importance of different behavioral SIDs, making it difficult for downstream tasks to distinguish informative SIDs from noisy ones. To address these challenges, we propose ADC-SID, a framework that Adaptively Denoises Collaborative information for SID quantization. It comprises two key components: (i) Adaptive Behavior-Content Alignment, which adjusts alignment strength to mitigate corruption caused by collaborative noise; and (ii) Dynamic Behavioral Weighting Mechanism, which learns importance scores for behavioral SIDs to enable downstream models to suppress noise. Extensive experiments has demonstrated ADC-SID's superiority...

cs.IR↗
arXiv2026

Extrapolative Quantum Error Mitigation in Continuous-Variable Systems beyond the Training Horizon

Continuous-variable (CV) quantum systems provide a versatile platform for quantum information processing, in which quantum states can be represented in the quadrature phase space. In realistic implementations, environmental noise, primarily photon loss and dephasing, progressively degrades these states. Machine-learning-based quantum error mitigation (QEM) has recently emerged as a promising approach to suppress such noise; however, existing methods are typically limited to the training horizon and require training data that cover the entire evolution, which is experimentally demanding. Here we introduce a framework for extrapolative quantum error mitigation based on a time-conditioned Swin Transformer. By explicitly embedding the evolution time via adaptive layer normalization, the model learns a correction map that accounts for the continuous accumulation of noise while capturing nonlocal phase-space correlations. Numerical simulations under both Markovian and non-Markovian noise demonstrate accurate state recovery in the long-time regime, where existing approaches deteriorate. Our results establish extrapolative QEM as a practical route to mitigating noise in CV quantum systems without exhaustive training data.

quant-ph↗
arXiv2026

Generative Diffusion Models for High Dimensional Channel Estimation

Along with the prosperity of generative artificial intelligence (AI), its potential for solving conventional challenges in wireless communications has also surfaced. Inspired by this trend, we investigate the application of the advanced diffusion models (DMs), a representative class of generative AI models, to high dimensional wireless channel estimation. By capturing the structure of multiple-input multiple-output (MIMO) wireless channels via a deep generative prior encoded by DMs, we develop a novel posterior inference method for channel reconstruction. We further adapt the proposed method to recover channel information from low-resolution quantized measurements. Additionally, to enhance the over-the-air viability, we integrate the DM with the unsupervised Stein's unbiased risk estimator to enable learning from noisy observations and circumvent the requirements for ground truth channel data that is hardly available in practice. Results reveal that the proposed estimator achieves high-fidelity channel recovery while reducing estimation latency by a factor of 10 compared to state-of-the-art schemes, facilitating real-time implementation. Moreover, our method outperforms existing estimators while reducing the pilot overhead by half, showcasing its scalability to ultra-massive antenna arrays.

cs.IT↗
arXiv2026

SPEX: A Vision-Language Model for Land Cover Extraction on Spectral Remote Sensing Images

Spectral information has long been recognized as a critical cue in remote sensing observations. Although numerous vision-language models have been developed for pixel-level interpretation, spectral information remains underutilized, resulting in suboptimal performance, particularly in multispectral scenarios. To address this limitation, we construct a vision-language instruction-following dataset named SPIE, which encodes spectral priors of land-cover objects into textual attributes recognizable by large language models (LLMs), based on classical spectral index computations. Leveraging this dataset, we propose SPEX, a multimodal LLM designed for instruction-driven land cover extraction. To this end, we introduce several carefully designed components and training strategies, including multiscale feature aggregation, token context condensation, and multispectral visual pre-training, to achieve precise and flexible pixel-level interpretation. To the best of our knowledge, SPEX is the first multimodal vision-language model dedicated to land cover extraction in spectral remote sensing imagery. Extensive experiments on five public multispectral datasets demonstrate that SPEX consistently outperforms existing state-of-the-art methods in extracting typical land cover categories such as vegetation, buildings, and water bodies. Moreover, SPEX is capable of generating textual explanations for its predictions, thereby enhancing interpretability and user-friendliness. Code will be released at: https://github.com/MiliLab/SPEX.

cs.CV↗