Searcharxiv⌕ Search

arXiv subjects

Zheng Zhang

Publications and source records attributed to Zheng Zhang.

At least 181 records · Page 10Linked to original sources
arXiv2025

NetEcho: From Real-World Streaming Side-Channels to Full LLM Conversation Recovery

In the rapidly expanding landscape of Large Language Model (LLM) applications, real-time output streaming has become the dominant interaction paradigm. While this enhances user experience, recent research reveals that it exposes a non-trivial attack surface through network side-channels. Adversaries can exploit patterns in encrypted traffic to infer sensitive information and reconstruct private conversations. In response, LLM providers and third-party services are deploying defenses such as traffic padding and obfuscation to mitigate these vulnerabilities. This paper starts by presenting a systematic analysis of contemporary side-channel defenses in mainstream LLM applications, with a focus on services from vendors like OpenAI and DeepSeek. We identify and examine seven representative deployment scenarios, each incorporating active/passive mitigation techniques. Despite these enhanced security measures, our investigation uncovers significant residual information that remains vulnerable to leakage within the network traffic. Building on this discovery, we introduce NetEcho, a novel, LLM-based framework that comprehensively unleashes the network side-channel risks of today's LLM applications. NetEcho is designed to recover entire conversations -- including both user prompts and LLM responses -- directly from encrypted network traffic. It features a deliberate design that ensures high-fidelity text recovery, transferability across different deployment scenarios, and moderate operational cost. In our evaluations on medical and legal applications built upon leading models like DeepSeek-v3 and GPT-4o, NetEcho can recover avg $\sim$70\% information of each conversation, demonstrating a critical limitation in current defense mechanisms. We conclude by discussing the implications of our findings and proposing future directions for augmenting network traffic security.

cs.CR↗
arXiv2025

Anomalous enhancement of magnetism by nonmagnetic doping in the honeycomb-lattice antiferromagnet ErOCl

Tuning magnetic anisotropy through chemical doping is a powerful strategy for designing functional materials with enhanced magnetic properties. Here, we report an enhanced Er^3+ magnetic moment resulting from nonmagnetic Lu^3+ substitution in the honeycomb-lattice antiferromagnet ErOCl. Unlike the Curie-Weiss type divergence typically observed in diluted magnetic systems, our findings reveal a distinct enhancement of magnetization per Er^3+ ion under high magnetic fields, suggesting an unconventional mechanism. Structural analysis reveals that Lu^3+ doping leads to a pronounced contraction of the c axis, which is attributed to chemical pressure effects, while preserving the layered SmSI-type crystal structure with space group R-3m. High-resolution Raman spectroscopy reveals a systematic blueshift of the first and seventh crystalline electric field (CEF) excitations, indicating an increase in the axial CEF parameter B_2^0. This modification enhances the magnetic anisotropy along the c axis, leading to a significant increase in magnetization at low temperatures and under high magnetic fields, contrary to conventional expectations for magnetic dilution. Our work not only clarifies the intimate connection between magnetism and CEF in rare-earth compounds, but more importantly, it reveals a physical pathway to effectively tune magnetic anisotropy via anisotropic lattice distortion induced by chemical pressure.

cond-mat.mtrl-sci↗
arXiv2025

LaX: Boosting Low-Rank Training of Foundation Models via Latent Crossing

Training foundation models such as ViTs and LLMs requires tremendous computing cost. Low-rank matrix or tensor factorization offers a parameter-efficient alternative, but often downgrades performance due to the restricted parameter space. In this work, we introduce {\textbf{Latent Crossing (LaX)}} -- a simple yet effective plug-and-play module that enhances the capacity of low-rank models by enabling information flow across low-rank subspaces. We extensively validate the benefits of LaX on pre-training tasks with ViT-Base/Large and LLaMA-like models ranging from 60M to 1B parameters. LaX boosts low-rank model performance to match or exceed the full-rank baselines while using 2-3\(\times\) fewer parameters. When equipped with low-rank adapters (i.e., LoRA) for fine-tuning LLaMA-7/13B, LaX consistently improves performance on arithmetic and common sense reasoning tasks with negligible cost.

cs.LG↗
arXiv2025

SharpZO: Hybrid Sharpness-Aware Vision Language Model Prompt Tuning via Forward-Only Passes

Fine-tuning vision language models (VLMs) has achieved remarkable performance across various downstream tasks; yet, it requires access to model gradients through backpropagation (BP), making them unsuitable for memory-constrained, inference-only edge devices. To address this limitation, previous work has explored various BP-free fine-tuning methods. However, these approaches often rely on high-variance evolutionary strategies (ES) or zeroth-order (ZO) optimization, and often fail to achieve satisfactory performance. In this paper, we propose a hybrid Sharpness-aware Zeroth-order optimization (SharpZO) approach, specifically designed to enhance the performance of ZO VLM fine-tuning via a sharpness-aware warm-up training. SharpZO features a two-stage optimization process: a sharpness-aware ES stage that globally explores and smooths the loss landscape to construct a strong initialization, followed by a fine-grained local search via sparse ZO optimization. The entire optimization relies solely on forward passes. Detailed theoretical analysis and extensive experiments on CLIP models demonstrate that SharpZO significantly improves accuracy and convergence speed, achieving up to 7% average gain over state-of-the-art forward-only methods.

cs.LG↗
arXiv2025

Saten: Sparse Augmented Tensor Networks for Post-Training Compression of Large Language Models

The efficient implementation of large language models (LLMs) is crucial for deployment on resource-constrained devices. Low-rank tensor compression techniques, such as tensor-train (TT) networks, have been widely studied for over-parameterized neural networks. However, their applications to compress pre-trained large language models (LLMs) for downstream tasks (post-training) remains challenging due to the high-rank nature of pre-trained LLMs and the lack of access to pretraining data. In this study, we investigate low-rank tensorized LLMs during fine-tuning and propose sparse augmented tensor networks (Saten) to enhance their performance. The proposed Saten framework enables full model compression. Experimental results demonstrate that Saten enhances both accuracy and compression efficiency in tensorized language models, achieving state-of-the-art performance.

cs.CL↗
arXiv2025

Bolster Hallucination Detection via Prompt-Guided Data Augmentation

Large language models (LLMs) have garnered significant interest in AI community. Despite their impressive generation capabilities, they have been found to produce misleading or fabricated information, a phenomenon known as hallucinations. Consequently, hallucination detection has become critical to ensure the reliability of LLM-generated content. One primary challenge in hallucination detection is the scarcity of well-labeled datasets containing both truthful and hallucinated outputs. To address this issue, we introduce Prompt-guided data Augmented haLlucination dEtection (PALE), a novel framework that leverages prompt-guided responses from LLMs as data augmentation for hallucination detection. This strategy can generate both truthful and hallucinated data under prompt guidance at a relatively low cost. To more effectively evaluate the truthfulness of the sparse intermediate embeddings produced by LLMs, we introduce an estimation metric called the Contrastive Mahalanobis Score (CM Score). This score is based on modeling the distributions of truthful and hallucinated data in the activation space. CM Score employs a matrix decomposition approach to more accurately capture the underlying structure of these distributions. Importantly, our framework does not require additional human annotations, offering strong generalizability and practicality for real-world applications. Extensive experiments demonstrate that PALE achieves superior hallucination detection performance, outperforming the competitive baseline by a significant margin of 6.55%.

cs.LG↗
arXiv2025

Jigsaw3D: Disentangled 3D Style Transfer via Patch Shuffling and Masking

Controllable 3D style transfer seeks to restyle a 3D asset so that its textures match a reference image while preserving the integrity and multi-view consistency. The prevalent methods either rely on direct reference style token injection or score-distillation from 2D diffusion models, which incurs heavy per-scene optimization and often entangles style with semantic content. We introduce Jigsaw3D, a multi-view diffusion based pipeline that decouples style from content and enables fast, view-consistent stylization. Our key idea is to leverage the jigsaw operation - spatial shuffling and random masking of reference patches - to suppress object semantics and isolate stylistic statistics (color palettes, strokes, textures). We integrate these style cues into a multi-view diffusion model via reference-to-view cross-attention, producing view-consistent stylized renderings conditioned on the input mesh. The renders are then style-baked onto the surface to yield seamless textures. Across standard 3D stylization benchmarks, Jigsaw3D achieves high style fidelity and multi-view consistency with substantially lower latency, and generalizes to masked partial reference stylization, multi-object scene styling, and tileable texture generation. Project page is available at: https://babahui.github.io/jigsaw3D.github.io/

cs.CV↗
arXiv2025

DeepOHeat-v1: Efficient Operator Learning for Fast and Trustworthy Thermal Simulation and Optimization in 3D-IC Design

Thermal analysis is crucial in 3D-IC design due to increased power density and complex heat dissipation paths. Although operator learning frameworks such as DeepOHeat~\cite{liu2023deepoheat} have demonstrated promising preliminary results in accelerating thermal simulation, they face critical limitations in prediction capability for multi-scale thermal patterns, training efficiency, and trustworthiness of results during design optimization. This paper presents DeepOHeat-v1, an enhanced physics-informed operator learning framework that addresses these challenges through three key innovations. First, we integrate Kolmogorov-Arnold Networks with learnable activation functions as trunk networks, enabling an adaptive representation of multi-scale thermal patterns. This approach achieves a 1.25x and 6.29x reduction in error in two representative test cases. Second, we introduce a separable training method that decomposes the basis function along the coordinate axes, achieving 62x training speedup and 31x GPU memory reduction in our baseline case, and enabling thermal analysis at resolutions previously infeasible due to GPU memory constraints. Third, we propose a confidence score to evaluate the trustworthiness of the predicted results, and further develop a hybrid optimization workflow that combines operator learning with finite difference (FD) using Generalized Minimal Residual (GMRES) method for incremental solution refinement, enabling efficient and trustworthy thermal optimization. Experimental results demonstrate that DeepOHeat-v1 achieves accuracy comparable to optimization using high-fidelity finite difference solvers, while speeding up the entire optimization process by $70.6\times$ in our test cases, effectively minimizing the peak temperature through optimal placement of heat-generating components. Open source code is available at https://github.com/xlyu0127/DeepOHeat-v1.

cs.LG↗
arXiv2025

CoLA: Compute-Efficient Pre-Training of LLMs via Low-Rank Activation

The full-size MLPs and the projection layers in attention introduce tremendous model sizes of large language models (LLMs), consuming extensive computational resources in pre-training. We empirically observe that the activations of pre-trained LLMs exhibit low-rank property. Motivated by such observations, we propose CoLA and its memory-efficient implementation, CoLA-M, to replace these full-size layers with compute-efficient auto-encoders that naturally enforce low-rank activations throughout training. This fundamental architectural change eliminates the activation redundancy and significantly boosts model capacity and training efficiency. Experiments on LLaMA models with 60 million to 7 billion parameters show that CoLA reduces the computing cost by $\bf 2\pmb{\times}$ and improves training throughput by $\bf 1.86\pmb{\times}$ while maintaining full-rank level performance. CoLA-M further squeezes memory cost without sacrificing throughput, offering a pre-training approach with collectively superior parameter, computing, and memory efficiency. The LLMs produced are also $\bf 2\pmb{\times}$ smaller, enabling faster inference with lower memory cost on resource-constrained platforms.

cs.LG↗
arXiv2025

Fair Uncertainty Quantification for Depression Prediction

Trustworthy depression prediction based on deep learning, incorporating both predictive reliability and algorithmic fairness across diverse demographic groups, is crucial for clinical application. Recently, achieving reliable depression predictions through uncertainty quantification has attracted increasing attention. However, few studies have focused on the fairness of uncertainty quantification (UQ) in depression prediction. In this work, we investigate the algorithmic fairness of UQ, namely Equal Opportunity Coverage (EOC) fairness, and propose Fair Uncertainty Quantification (FUQ) for depression prediction. FUQ pursues reliable and fair depression predictions through group-based analysis. Specifically, we first group all the participants by different sensitive attributes and leverage conformal prediction to quantify uncertainty within each demographic group, which provides a theoretically guaranteed and valid way to quantify uncertainty for depression prediction and facilitates the investigation of fairness across different demographic groups. Furthermore, we propose a fairness-aware optimization strategy that formulates fairness as a constrained optimization problem under EOC constraints. This enables the model to preserve predictive reliability while adapting to the heterogeneous uncertainty levels across demographic groups, thereby achieving optimal fairness. Through extensive evaluations on several visual and audio depression datasets, our approach demonstrates its effectiveness.

cs.LG↗
arXiv2025

End4: End-to-end Denoising Diffusion for Diffusion-Based Inpainting Detection

The powerful generative capabilities of diffusion models have significantly advanced the field of image synthesis, enhancing both full image generation and inpainting-based image editing. Despite their remarkable advancements, diffusion models also raise concerns about potential misuse for malicious purposes. However, existing approaches struggle to identify images generated by diffusion-based inpainting models, even when similar inpainted images are included in their training data. To address this challenge, we propose a novel detection method based on End-to-end denoising diffusion (End4). Specifically, End4 designs a denoising reconstruction model to improve the alignment degree between the latent spaces of the reconstruction and detection processes, thus reconstructing features that are more conducive to detection. Meanwhile, it leverages a Scale-aware Pyramid-like Fusion Module (SPFM) that refines local image features under the guidance of attention pyramid layers at different scales, enhancing feature discriminability. Additionally, to evaluate detection performance on inpainted images, we establish a comprehensive benchmark comprising images generated from five distinct masked regions. Extensive experiments demonstrate that our End4 effectively generalizes to unseen masking patterns and remains robust under various perturbations. Our code and dataset will be released soon.

cs.CV↗
arXiv2025

MultiMind: Enhancing Werewolf Agents with Multimodal Reasoning and Theory of Mind

Large Language Model (LLM) agents have demonstrated impressive capabilities in social deduction games (SDGs) like Werewolf, where strategic reasoning and social deception are essential. However, current approaches remain limited to textual information, ignoring crucial multimodal cues such as facial expressions and tone of voice that humans naturally use to communicate. Moreover, existing SDG agents primarily focus on inferring other players' identities without modeling how others perceive themselves or fellow players. To address these limitations, we use One Night Ultimate Werewolf (ONUW) as a testbed and present MultiMind, the first framework integrating multimodal information into SDG agents. MultiMind processes facial expressions and vocal tones alongside verbal content, while employing a Theory of Mind (ToM) model to represent each player's suspicion levels toward others. By combining this ToM model with Monte Carlo Tree Search (MCTS), our agent identifies communication strategies that minimize suspicion directed at itself. Through comprehensive evaluation in both agent-versus-agent simulations and studies with human players, we demonstrate MultiMind's superior performance in gameplay. Our work presents a significant advancement toward LLM agents capable of human-like social reasoning across multimodal domains.

cs.AI↗
arXiv2025

Few-Shot Query Intent Detection via Relation-Aware Prompt Learning

Intent detection is a crucial component of modern conversational systems, since accurately identifying user intent at the beginning of a conversation is essential for generating effective responses. Recent efforts have focused on studying this problem under a challenging few-shot scenario. These approaches primarily leverage large-scale unlabeled dialogue text corpora to pretrain language models through various pretext tasks, followed by fine-tuning for intent detection with very limited annotations. Despite the improvements achieved, existing methods have predominantly focused on textual data, neglecting to effectively capture the crucial structural information inherent in conversational systems, such as the query-query relation and query-answer relation. To address this gap, we propose SAID, a novel framework that integrates both textual and relational structure information in a unified manner for model pretraining for the first time. Building on this framework, we further propose a novel mechanism, the query-adaptive attention network (QueryAdapt), which operates at the relation token level by generating intent-specific relation tokens from well-learned query-query and query-answer relations explicitly, enabling more fine-grained knowledge transfer. Extensive experimental results on two real-world datasets demonstrate that SAID significantly outperforms state-of-the-art methods.

cs.CL↗
arXiv2025

SAC-MIL: Spatial-Aware Correlated Multiple Instance Learning for Histopathology Whole Slide Image Classification

We propose Spatial-Aware Correlated Multiple Instance Learning (SAC-MIL) for performing WSI classification. SAC-MIL consists of a positional encoding module to encode position information and a SAC block to perform full instance correlations. The positional encoding module utilizes the instance coordinates within the slide to encode the spatial relationships instead of the instance index in the input WSI sequence. The positional encoding module can also handle the length extrapolation issue where the training and testing sequences have different lengths. The SAC block is an MLP-based method that performs full instance correlation in linear time complexity with respect to the sequence length. Due to the simple structure of MLP, it is easy to deploy since it does not require custom CUDA kernels, compared to Transformer-based methods for WSI classification. SAC-MIL has achieved state-of-the-art performance on the CAMELYON-16, TCGA-LUNG, and TCGA-BRAC datasets. The code will be released upon acceptance.

cs.CV↗
arXiv2025

Outer Contour-driven Ruled Surface Generation for Linear Hot-wire Rough Machining

We propose a novel method to generate a small set of ruled surfaces that do not collide with the input shape for linear hot-wire rough machining. Central to our technique is a new observation: the ruled surfaces constructed by vertical extrusion from planar smooth curves, which approach the input shape's outer contour lines while having no collisions, are capable of removing materials effectively during rough machining. Accordingly, we develop an iterative algorithm that alternates in each iteration between computing a viewpoint to determine an outer contour line and optimizing a smooth curve to approximate that contour line under the collision-free constraint. Specifically, a view selection approach based on genetic algorithm is used to optimize the viewpoint for removing materials as much as possible, and present an adaptive algorithm to find the constrained curves. The feasibility and practicability of our method are demonstrated through 10 physical examples. Compared with manual designs, our method obtains lower errors with the same number of cuts.

cs.CG↗
arXiv2025

GLITTER: An AI-assisted Platform for Material-Grounded Asynchronous Discussion in Flipped Learning

Flipped classrooms promote active learning by having students engage with materials independently before class, allowing in-class time for collaborative problem-solving. During this pre-class phase, asynchronous online discussions help students build knowledge and clarify concepts with peers. However, it remains difficult to engage with temporally dispersed peer contributions, connect discussions with static learning materials, and prepare for in-class sessions based on their self-learning outcome. Our formative study identified cognitive challenges students encounter, including navigation barriers, reflection gaps, and contribution difficulty and anxiety. We present GLITTER, an AI-assisted discussion platform for pre-class learning in flipped classrooms. GLITTER helps students identify posts with shared conceptual dimensions, scaffold knowledge integration through conceptual blending, and enhance metacognition via personalized reflection reports. A lab study within subjects (n = 12) demonstrates that GLITTER improves discussion engagement, sparks new ideas, supports reflection, and increases preparedness for in-class activities.

cs.HC↗
arXiv2025

An Experimental Framework to Study Turbulence Induced Particle Mobilization

An experimental framework was developed to study the initiation of particle mobilization in a laboratory setting. Large and heavy particles mobilized by a turbulent, gaseous carrier-phase were considered. An airfoil oscillated in the free-stream, generating a tonal free-stream disturbance that perturbed a turbulent boundary layer. The flow developing behind this forced flow was characterized using hot-wire anemometry and particle image velocimetry (PIV). Downstream of the oscillating airfoil mechanism, the turbulent boundary layer responded to the forcing in the form of excess energy at the forcing scale. The signature of this forcing scale was observed to span the entire wall-normal extent of the flow, extending all the way down to the wall. The size of this flow scale was shown to be controlled by changing the frequency of oscillation, while the energy in this flow scale was controlled via the amplitude of oscillation. Demonstrative measurements are presented in which this forced carrier-phase flow mobilized a particle-phase on a particle bed. A PIV-based approach was used to measure the initiation of particle motion as well as the incoming carrier-phase velocity field. The particles on the bed were mobilized ``on-demand'' by the deflection of the airfoil. Consistent with prior work, it was observed that particle mobilization was correlated with the large-scale flow structures of the carrier-phase.

physics.flu-dyn↗
arXiv2025

ShortListing Model: A Streamlined SimplexDiffusion for Discrete Variable Generation

Generative modeling of discrete variables is challenging yet crucial for applications in natural language processing and biological sequence design. We introduce the Shortlisting Model (SLM), a novel simplex-based diffusion model inspired by progressive candidate pruning. SLM operates on simplex centroids, reducing generation complexity and enhancing scalability. Additionally, SLM incorporates a flexible implementation of classifier-free guidance, enhancing unconditional generation performance. Extensive experiments on DNA promoter and enhancer design, protein design, character-level and large-vocabulary language modeling demonstrate the competitive performance and strong potential of SLM. Our code can be found at https://github.com/GenSI-THUAIR/SLM

cs.LG↗