arXiv subjects

Zijian Wang

Publications and source records attributed to Zijian Wang.

At least 91 records · Page 5Linked to original sources

arXiv2025

Eliciting Chain-of-Thought in Base LLMs via Gradient-Based Representation Optimization

Chain-of-Thought (CoT) reasoning is a critical capability for large language models (LLMs), enabling them to tackle com- plex multi-step tasks. While base LLMs, pre-trained on general text corpora, often struggle with reasoning due to a lack of specialized training, recent studies reveal their latent reason- ing potential tied to hidden states. However, existing hidden state manipulation methods, such as linear activation steering, suffer from limitations due to their rigid and unconstrained nature, often leading to distribution shifts and degraded text quality. In this work, we propose a novel approach for elic- iting CoT reasoning from base LLMs through hidden state manipulation grounded in probabilistic conditional generation. By reformulating the challenge as an optimization problem with a balanced likelihood and prior regularization framework, our method guides hidden states toward reasoning-oriented trajectories while preserving linguistic coherence. Extensive evaluations across mathematical, commonsense, and logical reasoning benchmarks demonstrate that our approach con- sistently outperforms existing steering methods, offering a theoretically principled and effective solution for enhancing reasoning capabilities in base LLMs.

cs.CL↗

arXiv2025

UT-OSANet: A Multimodal Deep Learning model for Evaluating and Classifying Obstructive Sleep Apnea

Obstructive sleep apnea (OSA) is a highly prevalent sleep disorder that is associated with increased risks of cardiovascular morbidity and all-cause mortality. While existing diagnostic approaches can roughly classify OSA severity or detect isolated respiratory events, they lack the precision and comprehensiveness required for high resolution, event level diagnosis. Here, we present UT OSANet, a deep learning based model designed as a event level, multi scenario diagnostic tool for OSA. This model facilitates detailed identification of events associated with OSA, including apnea, hypopnea, oxygen desaturation, and arousal. Moreover, the model employs flexibly adjustable input modalities such as electroencephalography (EEG), airflow, and SpO 2. It utilizes a random masked modality combination training strategy, allowing it to comprehend cross-modal relationships while sustaining consistent performance across varying modality conditions. This model was trained and evaluated utilizing 9,021 polysomnography (PSG) recordings from five independent datasets. achieving sensitivities up to 0.93 and macro F1 scores of 0.84, 0.85 across home, clinical, and research scenarios. This model serves as an event-level, multi-scenario diagnostic instrument for real-world applications of OSA, while also establishing itself as a means to deepen the mechanistic comprehension of respiratory processes in sleep disorders and their extensive health implications.

eess.SP↗

arXiv2025

Planning-Aware Code Infilling via Horizon-Length Prediction

Fill-in-the-Middle (FIM), or infilling, has become integral to code language models, enabling generation of missing code given both left and right contexts. However, the current FIM training paradigm which performs next-token prediction (NTP) over reordered sequence often leads to models struggling to generate content that aligns well with the surrounding context. We hypothesize that NTP alone is insufficient for models to learn effective planning conditioned on the distant right context, a critical factor for successful code infilling. To overcome this, we propose Horizon-Length Prediction (HLP), a novel training objective that teaches models to predict the number of remaining middle tokens at each step. HLP advances FIM with lookahead planning, enabling models to inherently learn infilling boundaries for arbitrary left and right contexts without relying on dataset-specific post-processing. Our evaluation across different model families and sizes shows that HLP significantly improves FIM performance by up to 24% relatively on diverse benchmarks, across file-level and repository-level. Furthermore, the enhanced planning capability gained through HLP boosts model performance on code reasoning. Importantly, HLP incurs negligible training overhead and no additional inference cost, ensuring its practicality for real-world scenarios.

cs.LG↗

arXiv2025

Search for light Dark Sectors with GeV Muon Beams

Sub-GeV light dark matter often requires new light mediators, such as a dark $Z$ boson in the $L_μ- L_τ$ gauge theory. We study the search potential for such a $Z^\prime$ boson via the process $μe^- \to μe^- X$, with $X$ decaying invisibly, in a muon on-target experiment using a high-intensity 1-10 GeV muon beam from facilities such as HIAF-HIRIBL. Events are identified by the scattered muon and electron from the target using silicon strip detectors in a single-station telescope system. Backgrounds are suppressed through a trained boosted decision tree (BDT) classifier, and activity in downstream subdetectors remains low. This approach can probe a $Z^\prime$ boson in the 10 MeV mass range with improved sensitivity. Nearly three orders of magnitude improvement is achievable with a full multi-telescope station system employing a 160 GeV muon beam at CERN, such as in the MUonE experiment.

hep-ph↗

arXiv2025

A DeepONet joint Neural Tangent Kernel Hybrid Framework for Physics-Informed Inverse Source Problems and Robust Image Reconstruction

This work presents a novel hybrid approach that integrates Deep Operator Networks (DeepONet) with the Neural Tangent Kernel (NTK) to solve complex inverse problem. The method effectively addresses tasks such as source localization governed by the Navier-Stokes equations and image reconstruction, overcoming challenges related to nonlinearity, sparsity, and noisy data. By incorporating physics-informed constraints and task-specific regularization into the loss function, the framework ensures solutions that are both physically consistent and accurate. Validation on diverse synthetic and real datasets demonstrates its robustness, scalability, and precision, showcasing its broad potential applications in computational physics and imaging sciences.

cs.CV↗

arXiv2025

ThoughtProbe: Classifier-Guided LLM Thought Space Exploration via Probing Representations

This paper introduces ThoughtProbe, a novel inference time framework that leverages the hidden reasoning features of Large Language Models (LLMs) to improve their reasoning performance. Unlike previous works that manipulate the hidden representations to steer LLM generation, we harness them as discriminative signals to guide the tree structured response space exploration. In each node expansion, a classifier serves as a scoring and ranking mechanism that efficiently allocates computational resources by prioritizing higher score candidates for continuation. After completing the tree expansion, we collect answers from all branches to form a candidate answer pool. We then propose a branch aggregation method that marginalizes over all supporting branches by aggregating their CoT scores, thereby identifying the optimal answer from the pool. Experimental results show that our framework's comprehensive exploration not only covers valid reasoning chains but also effectively identifies them, achieving significant improvements across multiple arithmetic reasoning benchmarks.

cs.CL↗

arXiv2025

Balancing Knowledge Updates: Toward Unified Modular Editing in LLMs

Knowledge editing has emerged as an efficient approach for updating factual knowledge in large language models (LLMs). It typically locates knowledge storage modules and then modifies their parameters. However, most existing methods focus on the weights of multilayer perceptron (MLP) modules, which are often identified as the main repositories of factual information. Other components, such as attention (Attn) modules, are often ignored during editing. This imbalance can leave residual outdated knowledge and limit editing effectiveness. We perform comprehensive knowledge localization experiments on advanced LLMs and find that Attn modules play a substantial role in factual knowledge storage and retrieval, especially in earlier layers. Based on these insights, we propose IntAttn-Edit, a method that extends the associative memory paradigm to jointly update both MLP and Attn modules. Our approach uses a knowledge balancing strategy that allocates update magnitudes in proportion to each module's measured contribution to knowledge storage. Experiments on standard benchmarks show that IntAttn-Edit achieves higher edit success, better generalization, and stronger knowledge preservation than prior methods. Further analysis shows that the balancing strategy keeps editing performance within an optimal range across diverse settings.

cs.CL↗

arXiv2025

CFVBench: A Comprehensive Video Benchmark for Fine-grained Multimodal Retrieval-Augmented Generation

Multimodal Retrieval-Augmented Generation (MRAG) enables Multimodal Large Language Models (MLLMs) to generate responses with external multimodal evidence, and numerous video-based MRAG benchmarks have been proposed to evaluate model capabilities across retrieval and generation stages. However, existing benchmarks remain limited in modality coverage and format diversity, often focusing on single- or limited-modality tasks, or coarse-grained scene understanding. To address these gaps, we introduce CFVBench, a large-scale, manually verified benchmark constructed from 599 publicly available videos, yielding 5,360 open-ended QA pairs. CFVBench spans high-density formats and domains such as chart-heavy reports, news broadcasts, and software tutorials, requiring models to retrieve and reason over long temporal video spans while maintaining fine-grained multimodal information. Using CFVBench, we systematically evaluate 7 retrieval methods and 14 widely-used MLLMs, revealing a critical bottleneck: current models (even GPT5 or Gemini) struggle to capture transient yet essential fine-grained multimodal details. To mitigate this, we propose Adaptive Visual Refinement (AVR), a simple yet effective framework that adaptively increases frame sampling density and selectively invokes external tools when necessary. Experiments show that AVR consistently enhances fine-grained multimodal comprehension and improves performance across all evaluated MLLMs

cs.CL↗

arXiv2025

Fractional quantum Hall states under density decoherence

Fractional quantum Hall states are promising platforms for topological quantum computation due to their capacity to encode quantum information in topologically degenerate ground states and in the fusion space of non-abelian anyons. We investigate how the information encoded in two paradigmatic states, the Laughlin and Moore-Read states, is affected by density decoherence -- coupling of local charge density to non-thermal noise. We identify a critical filling factor $ν_c$, above which the quantum information remains fully recoverable for arbitrarily strong decoherence. The $ν=1/3$ Laughlin state and $ν= 1/2$ Moore-Read state both lie within this range. Below $ν_c$ both classes of states undergo a decoherence induced Berezinskii-Kosterlitz-Thousless (BKT) transition into a critical decohered phase. For Laughlin states, information encoded in the topological ground state manifold degrades continuously with decoherence strength inside this critical phase, vanishing only in the limit of infinite decoherence strength. On the other hand, quantum information encoded in the fusion space of non-abelian anyons of the Moore-Read states remains fully recoverable for arbitrary strong decoherence even beyond the BKT transition. These results lend further support to the promise of non-Abelian FQH states as platforms for topological quantum computation and raises the question of how errors in such states can be corrected.

cond-mat.str-el↗

arXiv2025

Learning the detector in optical tomography

We propose a method to reconstruct the optical absorption of a highly-scattering medium probed by diffuse light. The method consists of learning the optical detection system and then using this result to reconstruct the absorption. Our results are illustrated by numerical simulations.

physics.optics↗

arXiv2025

Diffusion Low Rank Hybrid Reconstruction for Sparse View Medical Imaging

This work presents TV-LoRA, a novel method for low-dose sparse-view CT reconstruction that combines a diffusion generative prior (NCSN++ with SDE modeling) and multi-regularization constraints, including anisotropic TV and nuclear norm (LoRA), within an ADMM framework. To address ill-posedness and texture loss under extremely sparse views, TV-LoRA integrates generative and physical constraints, and utilizes a 2D slice-based strategy with FFT acceleration and tensor-parallel optimization for efficient inference. Experiments on AAPM-2016, CTHD, and LIDC datasets with $N_{\mathrm{view}}=8,4,2$ show that TV-LoRA consistently surpasses benchmarks in SSIM, texture recovery, edge clarity, and artifact suppression, demonstrating strong robustness and generalizability. Ablation studies confirm the complementary effects of LoRA regularization and diffusion priors, while the FFT-PCG module provides a speedup. Overall, Diffusion + TV-LoRA achieves high-fidelity, efficient 3D CT reconstruction and broad clinical applicability in low-dose, sparse-sampling scenarios.

cs.CV↗

arXiv2025

Training Language Model Agents to Find Vulnerabilities with CTF-Dojo

Large language models (LLMs) have demonstrated exceptional capabilities when trained within executable runtime environments, notably excelling at software engineering tasks through verified feedback loops. Yet, scalable and generalizable execution-grounded environments remain scarce, limiting progress in training more capable ML agents. We introduce CTF-Dojo, the first large-scale executable runtime tailored for training LLMs with verifiable feedback, featuring 658 fully functional Capture-The-Flag (CTF)-style challenges containerized in Docker with guaranteed reproducibility. To enable rapid scaling without manual intervention, we develop CTF-Forge, an automated pipeline that transforms publicly available artifacts into ready-to-use execution environments in minutes, eliminating weeks of expert configuration traditionally required. We trained LLM-based agents on just 486 high-quality, execution-verified trajectories from CTF-Dojo, achieving up to 11.6% absolute gains over strong baselines across three competitive benchmarks: InterCode-CTF, NYU CTF Bench, and Cybench. Our best-performing 32B model reaches 31.9% Pass@1, establishing a new open-weight state-of-the-art that rivals frontier models like DeepSeek-V3-0324 and Gemini-2.5-Flash. By framing CTF-style tasks as a benchmark for executable-agent learning, CTF-Dojo demonstrates that execution-grounded training signals are not only effective but pivotal in advancing high-performance ML agents without dependence on costly proprietary systems.

cs.SE↗

arXiv2025

A Time-Series Foundation Model by Universal Delay Embedding

This study introduces Universal Delay Embedding (UDE), a pretrained foundation model designed to revolutionize time-series forecasting through principled integration of delay embedding representation and Koopman operator prediction. Leveraging Takens' embedding theorem, UDE as a dynamical representation of observed data constructs two-dimensional subspace patches from Hankel matrices, theoretically preserving dynamical and topological properties of underlying dynamical systems. Such patches are viewed as images, which can be efficiently processed by exploiting advanced deep learning technologies. Computationally, these patches further serve as tokens for learning a self-attention encoder, thus enabling accurate prediction of nonlinear time-series by a finite-dimensional Koopman operator in a linear manner in a latent space. Extensive evaluations across various benchmarks and real-world climate datasets demonstrate over 20% average reduction in mean squared error versus state-of-the-art foundation models, alongside superior generalization in fine-tuning scenarios. In particular, the learned dynamical representations and Koopman operator prediction forms from the patches exhibit exceptional interpretability, with consistent identification of topologically informative subspaces and robust encoding of domain-invariant dynamics, establishing UDE as a scalable, interpretable framework for universal time-series modeling and forecasting with broad scientific and industrial applicability.

cs.LG↗

arXiv2025

FoMo4Wheat: Toward reliable crop vision foundation models with globally curated data

Vision-driven field monitoring is central to digital agriculture, yet models built on general-domain pretrained backbones often fail to generalize across tasks, owing to the interaction of fine, variable canopy structures with fluctuating field conditions. We present FoMo4Wheat, one of the first crop-domain vision foundation model pretrained with self-supervision on ImAg4Wheat, the largest and most diverse wheat image dataset to date (2.5 million high-resolution images collected over a decade at 30 global sites, spanning >2,000 genotypes and >500 environmental conditions). This wheat-specific pretraining yields representations that are robust for wheat and transferable to other crops and weeds. Across ten in-field vision tasks at canopy and organ levels, FoMo4Wheat models consistently outperform state-of-the-art models pretrained on general-domain dataset. These results demonstrate the value of crop-specific foundation models for reliable in-field perception and chart a path toward a universal crop foundation model with cross-species and cross-task capabilities. FoMo4Wheat models and the ImAg4Wheat dataset are publicly available online: https://github.com/PheniX-Lab/FoMo4Wheat and https://huggingface.co/PheniX-Lab/FoMo4Wheat. The demonstration website is: https://fomo4wheat.phenix-lab.com/.

cs.CV↗

arXiv2025

On Evaluating the Poisoning Robustness of Federated Learning under Local Differential Privacy

Federated learning (FL) combined with local differential privacy (LDP) enables privacy-preserving model training across decentralized data sources. However, the decentralized data-management paradigm leaves LDPFL vulnerable to participants with malicious intent. The robustness of LDPFL protocols, particularly against model poisoning attacks (MPA), where adversaries inject malicious updates to disrupt global model convergence, remains insufficiently studied. In this paper, we propose a novel and extensible model poisoning attack framework tailored for LDPFL settings. Our approach is driven by the objective of maximizing the global training loss while adhering to local privacy constraints. To counter robust aggregation mechanisms such as Multi-Krum and trimmed mean, we develop adaptive attacks that embed carefully crafted constraints into a reverse training process, enabling evasion of these defenses. We evaluate our framework across three representative LDPFL protocols, three benchmark datasets, and two types of deep neural networks. Additionally, we investigate the influence of data heterogeneity and privacy budgets on attack effectiveness. Experimental results demonstrate that our adaptive attacks can significantly degrade the performance of the global model, revealing critical vulnerabilities and highlighting the need for more robust LDPFL defense strategies against MPA. Our code is available at https://github.com/ZiJW/LDPFL-Attack

cs.CR↗

arXiv2025

Cyber-Zero: Training Cybersecurity Agents without Runtime

Large Language Models (LLMs) have achieved remarkable success in software engineering tasks when trained with executable runtime environments, particularly in resolving GitHub issues. However, such runtime environments are often unavailable in other domains, especially cybersecurity, where challenge configurations and execution contexts are ephemeral or restricted. We present Cyber-Zero, the first runtime-free framework for synthesizing high-quality agent trajectories to train cybersecurity LLMs. Cyber-Zero leverages publicly available CTF writeups and employs persona-driven LLM simulation to reverse-engineer runtime behaviors and generate realistic, long-horizon interaction sequences without actual environments. Using trajectories synthesized by Cyber-Zero, we train LLM-based agents that achieve up to 13.1% absolute performance gains over baseline models on three prominent CTF benchmarks: InterCode-CTF, NYU CTF Bench, and Cybench. Our best model, Cyber-Zero-32B, establishes new state-of-the-art performance among open-weight models, matching the capabilities of proprietary systems like DeepSeek-V3-0324 and Claude-3.5-Sonnet while offering superior cost-effectiveness, and demonstrating that runtime-free trajectory synthesis can effectively democratize the development of state-of-the-art cybersecurity agents.

cs.CR↗

arXiv2025

Perceive-Sample-Compress: Towards Real-Time 3D Gaussian Splatting

Recent advances in 3D Gaussian Splatting (3DGS) have demonstrated remarkable capabilities in real-time and photorealistic novel view synthesis. However, traditional 3DGS representations often struggle with large-scale scene management and efficient storage, particularly when dealing with complex environments or limited computational resources. To address these limitations, we introduce a novel perceive-sample-compress framework for 3D Gaussian Splatting. Specifically, we propose a scene perception compensation algorithm that intelligently refines Gaussian parameters at each level. This algorithm intelligently prioritizes visual importance for higher fidelity rendering in critical areas, while optimizing resource usage and improving overall visible quality. Furthermore, we propose a pyramid sampling representation to manage Gaussian primitives across hierarchical levels. Finally, to facilitate efficient storage of proposed hierarchical pyramid representations, we develop a Generalized Gaussian Mixed model compression algorithm to achieve significant compression ratios without sacrificing visual fidelity. The extensive experiments demonstrate that our method significantly improves memory efficiency and high visual quality while maintaining real-time rendering speed.

cs.GR↗

arXiv2025

Outdoor Monocular SLAM with Global Scale-Consistent 3D Gaussian Pointmaps

3D Gaussian Splatting (3DGS) has become a popular solution in SLAM due to its high-fidelity and real-time novel view synthesis performance. However, some previous 3DGS SLAM methods employ a differentiable rendering pipeline for tracking, lack geometric priors in outdoor scenes. Other approaches introduce separate tracking modules, but they accumulate errors with significant camera movement, leading to scale drift. To address these challenges, we propose a robust RGB-only outdoor 3DGS SLAM method: S3PO-GS. Technically, we establish a self-consistent tracking module anchored in the 3DGS pointmap, which avoids cumulative scale drift and achieves more precise and robust tracking with fewer iterations. Additionally, we design a patch-based pointmap dynamic mapping module, which introduces geometric priors while avoiding scale ambiguity. This significantly enhances tracking accuracy and the quality of scene reconstruction, making it particularly suitable for complex outdoor environments. Our experiments on the Waymo, KITTI, and DL3DV datasets demonstrate that S3PO-GS achieves state-of-the-art results in novel view synthesis and outperforms other 3DGS SLAM methods in tracking accuracy. Project page: https://3dagentworld.github.io/S3PO-GS/.

cs.CV↗