arXiv subjects

Zhao Zhang

Publications and source records attributed to Zhao Zhang.

At least 91 records · Page 5Linked to original sources

arXiv2025

Seeing is Believing? Mitigating OCR Hallucinations in Multimodal Large Language Models

Recent advancements in multimodal large language models have enhanced document understanding by integrating textual and visual information. However, existing models exhibit incompleteness within their paradigm in real-world scenarios, particularly under visual degradation. In such conditions, the current response paradigm often fails to adequately perceive visual degradation and ambiguity, leading to overreliance on linguistic priors or misaligned visual-textual reasoning. This difficulty in recognizing uncertainty frequently results in the generation of hallucinatory content, especially when a precise answer is not feasible. To better demonstrate and analyze this phenomenon and problem, we propose KIE-HVQA, the first benchmark dedicated to evaluating OCR hallucination in degraded document understanding. This dataset includes test samples spanning identity cards and invoices, with simulated real-world degradations for OCR reliability. This setup allows for evaluating models' capacity, under degraded input, to distinguish reliable visual information and answer accordingly, thereby highlighting the challenge of avoiding hallucination on uncertain data. To achieve vision-faithful reasoning and thereby avoid the aforementioned issues, we further introduce a GRPO-based framework featuring a novel reward mechanism. By incorporating a self-awareness of visual uncertainty and an analysis method that initiates refusal to answer to increase task difficulty within our supervised fine-tuning and reinforcement learning framework, we successfully mitigated hallucinations in ambiguous regions. Experiments on Qwen2.5-VL demonstrate that our 7B-parameter model achieves a 22\% absolute improvement in hallucination-free accuracy over GPT-4o on KIE-HVQA and there is no significant performance drop in standard tasks, highlighting both effectiveness and robustness.

cs.CV↗

arXiv2025

Understanding the Landscape of Ampere GPU Memory Errors

Graphics Processing Units (GPUs) have become a de facto solution for accelerating high-performance computing (HPC) applications. Understanding their memory error behavior is an essential step toward achieving efficient and reliable HPC systems. In this work, we present a large-scale cross-supercomputer study to characterize GPU memory reliability, covering three supercomputers - Delta, Polaris, and Perlmutter - all equipped with NVIDIA A100 GPUs. We examine error logs spanning 67.77 million GPU device-hours across 10,693 GPUs. We compare error rates and mean-time-between-errors (MTBE) and highlight both shared and distinct error characteristics among these three systems. Based on these observations and analyses, we discuss the implications and lessons learned, focusing on the reliable operation of supercomputers, the choice of checkpointing interval, and the comparison of reliability characteristics with those of previous-generation GPUs. Our characterization study provides valuable insights into fault-tolerant HPC system design and operation, enabling more efficient execution of HPC applications.

cs.DC↗

arXiv2025

DPF-CM: A Data Processing Framework with Privacy-Preserving Vector Databases for Chinese Medical LLMs Training and Deployment

Current open-source training pipelines for Chinese medical language models predominantly emphasize optimizing training methodologies to enhance the performance of large language models (LLMs), yet lack comprehensive exploration into training data processing. To address this gap, we propose DPF-CM, a holistic Data Processing Framework for Chinese Medical LLMs training and deployment. DPF-CM comprises two core modules. The first module is a data processing pipeline tailored for model training. Beyond standard data processing operations, we (1) introduce a chained examples context-learning strategy to generate question-oriented instructions to mitigate the lack of instruction content, and (2) implement an ensemble-based filtering mechanism for preference data curation that averages multiple reward models to suppress noisy samples. The second module focuses on privacy preservation during model deployment. To prevent privacy risks from the inadvertent exposure of training data, we propose a Privacy Preserving Vector Database (PPVD) approach, which involves model memory search, high-risk database construction, secure database construction, and match-and-replace, four key stages to minimize privacy leakage during inference collectively. Experimental results show that DPF-CM significantly improves model accuracy, enabling our trained Chinese medical LLM to achieve state-of-the-art performance among open-source counterparts. Moreover, the framework reduces training data privacy leakage by 27%.

cs.LG↗

arXiv2025

Trae Agent: An LLM-based Agent for Software Engineering with Test-time Scaling

Software issue resolution is a critical challenge in software engineering and has garnered increasing attention in recent years. With the rapid advancement of large language models (LLMs), substantial progress has been made in addressing real-world software engineering tasks. Recent studies have introduced ensemble reasoning techniques to enhance the performance of LLM-based issue resolution. However, existing prompting-based methods still face limitations in effectively exploring large ensemble spaces and lack the capacity for repository-level understanding, both of which constrain their overall effectiveness. In this paper, we propose Trae Agent, the first agent-based ensemble reasoning approach for repository-level issue resolution. Trae Agent formulates our goal as an optimal solution search problem and addresses two key challenges, i.e., large ensemble spaces and repository-level understanding, through modular agents for generation, pruning, and selection. We conduct extensive experiments using three leading LLMs on the widely-adopted SWE-bench benchmark, comparing Trae Agent against four state-of-the-art ensemble reasoning techniques. Experimental results demonstrate that Trae Agent consistently achieves superior performance, with an average improvement of 10.22% over all baselines in terms of Pass@1. Trae Agent has achieved first place on the SWE-bench Verified leaderboard, with a notable Pass@1 score of 75.20%. We are pleased to release Trae Agent as an open-source project to support the research community, with all resources available at https://github.com/bytedance/trae-agent.

cs.SE↗

arXiv2025

Entanglement blossom in a simplex matryoshka

Exotic entanglement entropy scaling properties usually come with interesting entanglement structures in real space and novel metrics of the spacetime lattice. One prominent example is the rainbow chain where lattice sites symmetric about the center form entangled Bell pairs due to an effective long-range coupling from the strong inhomogeneity of the coupling strength. This manuscript generalizes the rainbow chain to higher dimensional space on lattices with Hausdorff dimension one and enlarged local Hilbert space keeping the Hamiltonian frustration free. The effective Hamiltonian from the Schrieffer-Wolf transformation is given by a stacking of layers of $k$-simplices with $0$-dimensional (fully-connected) antiferromagnetic Hamiltonians, which can be diagonalized analytically with Young operators. The original lattice can be obtained from proliferating disinclination defects in a regular $k$-dimensional cubical lattice, which introduces curvature at the center of the lattice. The model interpolates between the SYK model and the free-fermionic XX spin chain, and hence might be potentially useful in understanding black hole physics and holography.

quant-ph↗

arXiv2025

Lande g-factor measurements for the 5d6s 3D2 hyperfine levels of 176Lu+

We report measurements of the Lande g-factors for the 5d6s $^3$D$_2$ hyperfine levels of $^{176}$Lu$^+$ to a fractional inaccuracy of $5\times 10^{-7}$. Combining these measurements with theoretical calculations allows us to estimate hyperfine-mediated modifications to the quadrupole moments for each state and infer a value of $δΘ= 1.59(34)\times 10^{-4} \,ea_0^2$ for the residual quadrupole moment of the $^1S_0\leftrightarrow{^3}D_2$ hyperfine-averaged clock transition.

physics.atom-ph↗

arXiv2025

Text Detoxification: Data Efficiency, Semantic Preservation and Model Generalization

The widespread dissemination of toxic content on social media poses a serious threat to both online environments and public discourse, highlighting the urgent need for detoxification methods that effectively remove toxicity while preserving the original semantics. However, existing approaches often struggle to simultaneously achieve strong detoxification performance, semantic preservation, and robustness to out-of-distribution data. Moreover, they typically rely on costly, manually annotated parallel corpora while showing poor data efficiency. To address these challenges, we propose a two-stage training framework that jointly optimizes for data efficiency, semantic preservation, and model generalization. We first perform supervised fine-tuning on a small set of high-quality, filtered parallel data to establish a strong initialization. Then, we leverage unlabeled toxic inputs and a custom-designed reward model to train the LLM using Group Relative Policy Optimization. Experimental results demonstrate that our method effectively mitigates the trade-offs faced by previous work, achieving state-of-the-art performance with improved generalization and significantly reduced dependence on annotated data. Our code is available at: https://github.com/allacnobug/Detoxification-of-Text.

cs.LG↗

arXiv2025

NTIRE 2025 Image Shadow Removal Challenge Report

This work examines the findings of the NTIRE 2025 Shadow Removal Challenge. A total of 306 participants have registered, with 17 teams successfully submitting their solutions during the final evaluation phase. Following the last two editions, this challenge had two evaluation tracks: one focusing on reconstruction fidelity and the other on visual perception through a user study. Both tracks were evaluated with images from the WSRD+ dataset, simulating interactions between self- and cast-shadows with a large number of diverse objects, textures, and materials.

cs.CV↗

arXiv2025

One-Shot Simulation of Static Disorder in Quantum Dynamics with Equilibrium Initial State via Matrix Product State Sampling

Static disorder plays a crucial role in the electronic dynamics and spectroscopy of complex molecular systems. Traditionally, obtaining observables averaged over static disorder requires thousands of realizations via direct sampling of the disorder distribution, leading to high computational costs. In this work, we extend the auxiliary degree-of-freedom based matrix product state (MPS) method to handle system-bath correlated thermal equilibrium initial states. We validate the effectiveness of the extended method by computing the dipole-dipole time correlation function of the Holstein model relevant to the emission spectrum of molecular aggregates. Our results show that the method accurately captures static disorder effects using a one-shot quantum dynamical simulation, with only a moderate increase in MPS bond dimension, thereby significantly reducing computational cost. Moreover, it enables the generation of a much larger number of samples than the conventional direct sampling method at negligible additional cost, thus reducing statistical errors. This method provides a broadly useful tool for calculating equilibrium time correlation functions in system-bath coupled models with static disorder.

physics.chem-ph↗

arXiv2025

DeformCL: Learning Deformable Centerline Representation for Vessel Extraction in 3D Medical Image

In the field of 3D medical imaging, accurately extracting and representing the blood vessels with curvilinear structures holds paramount importance for clinical diagnosis. Previous methods have commonly relied on discrete representation like mask, often resulting in local fractures or scattered fragments due to the inherent limitations of the per-pixel classification paradigm. In this work, we introduce DeformCL, a new continuous representation based on Deformable Centerlines, where centerline points act as nodes connected by edges that capture spatial relationships. Compared with previous representations, DeformCL offers three key advantages: natural connectivity, noise robustness, and interaction facility. We present a comprehensive training pipeline structured in a cascaded manner to fully exploit these favorable properties of DeformCL. Extensive experiments on four 3D vessel segmentation datasets demonstrate the effectiveness and superiority of our method. Furthermore, the visualization of curved planar reformation images validates the clinical significance of the proposed framework. We release the code in https://github.com/barry664/DeformCL

cs.CV↗

arXiv2025

RAW Image Reconstruction from RGB on Smartphones. NTIRE 2025 Challenge Report

Numerous low-level vision tasks operate in the RAW domain due to its linear properties, bit depth, and sensor designs. Despite this, RAW image datasets are scarce and more expensive to collect than the already large and public sRGB datasets. For this reason, many approaches try to generate realistic RAW images using sensor information and sRGB images. This paper covers the second challenge on RAW Reconstruction from sRGB (Reverse ISP). We aim to recover RAW sensor images from smartphones given the corresponding sRGB images without metadata and, by doing this, ``reverse" the ISP transformation. Over 150 participants joined this NTIRE 2025 challenge and submitted efficient models. The proposed methods and benchmark establish the state-of-the-art for generating realistic RAW data.

eess.IV↗

arXiv2025

FCKT: Fine-Grained Cross-Task Knowledge Transfer with Semantic Contrastive Learning for Targeted Sentiment Analysis

In this paper, we address the task of targeted sentiment analysis (TSA), which involves two sub-tasks, i.e., identifying specific aspects from reviews and determining their corresponding sentiments. Aspect extraction forms the foundation for sentiment prediction, highlighting the critical dependency between these two tasks for effective cross-task knowledge transfer. While most existing studies adopt a multi-task learning paradigm to align task-specific features in the latent space, they predominantly rely on coarse-grained knowledge transfer. Such approaches lack fine-grained control over aspect-sentiment relationships, often assuming uniform sentiment polarity within related aspects. This oversimplification neglects contextual cues that differentiate sentiments, leading to negative transfer. To overcome these limitations, we propose FCKT, a fine-grained cross-task knowledge transfer framework tailored for TSA. By explicitly incorporating aspect-level information into sentiment prediction, FCKT achieves fine-grained knowledge transfer, effectively mitigating negative transfer and enhancing task performance. Experiments on three datasets, including comparisons with various baselines and large language models (LLMs), demonstrate the effectiveness of FCKT. The source code is available on https://github.com/cwei01/FCKT.

cs.CL↗

arXiv2025

Absolute frequency measurement of a Lu$^+$ $(^{3}\rm D_1)$ optical frequency standard via link to international atomic time

We report on an absolute frequency measurement of the ${\rm Lu}^{+}\,(^{3}\rm D_1)$ standard frequency which is defined as the hyperfine-average of $^{1}\rm S_0$ to $^{3}\rm D_1$ optical clock transitions in $^{176}{\rm Lu}^{+}$. The measurement result of $353\,638\,794\,073\,800.35(33)$Hz with a fractional uncertainty of $9.2 \times 10^{-16}$ was obtained by operating a single-ion $^{176}{\rm Lu}^{+}$ frequency standard intermittently over 3 months with a total uptime of 162 hours. Traceability to the International System of Units (SI) is realized by remote link to International Atomic Time. This is the first reported absolute frequency value for a ${\rm Lu}^{+}\,(^{3}\rm D_1)$ optical frequency standard.

physics.atom-ph↗

arXiv2025

Nonlinear Expectation Inference for Efficient Uncertainty Quantification and History Matching of Transient Darcy Flows in Porous Media with Random Parameters Under Distribution Uncertainty

The uncertainty quantification of Darcy flows using history matching is important for the evaluation and prediction of subsurface reservoir performance. Conventional methods aim to obtain the maximum a posterior or maximum likelihood estimate (MLE) using gradient-based, heuristic or ensemble-based methods. These methods can be computationally expensive for high-dimensional problems since forward simulation needs to be run iteratively as physical parameters are updated. In the current study, we propose a nonlinear expectation inference (NEI) method for efficient history matching and uncertainty quantification accounting for distribution or Knightian uncertainty. Forward simulation runs are conducted on prior realisations once, and then a range of expectations are computed in the data space based on subsets of prior realisations with no repetitive forward runs required. In NEI, no prior probability distribution for data is assumed. Instead, the probability distribution is assumed to be uncertain with prior and posterior uncertainty quantified by nonlinear expectations. The inferred result of NEI is the posterior subsets on which the expected flow rates are consistent with observation. The accuracy and efficiency of the new method are validated using single- and two-phase Darcy flows in 2D and 3D heterogeneous reservoirs.

math.NA↗

arXiv2025

Transfer learning empowers material Z classification with muon tomography

Cosmic-ray muon sources exhibit distinct scattering angle distributions when interacting with materials of different atomic numbers (Z values), facilitating the identification of various Z-class materials, particularly those radioactive high-Z nuclear elements. Most of the traditional identification methods are based on complex muon event reconstruction and trajectory fitting processes. Supervised machine learning methods offer some improvement but rely heavily on prior knowledge of target materials, significantly limiting their practical applicability in detecting concealed materials. For the first time, transfer learning is introduced into the field of muon tomography in this work. We propose two lightweight neural network models for fine-tuning and adversarial transfer learning, utilizing muon tomography data of bare materials to predict the Z-class of coated materials. By employing the inverse cumulative distribution function method, more accurate scattering angle distributions could be obtained from limited data, leading to an improvement by nearly 4\% in prediction accuracy compared with the traditional random sampling based training. When applied to coated materials with limited labeled or even unlabeled muon tomography data, the proposed method achieves an overall prediction accuracy exceeding 96\%, with high-Z materials reaching nearly 99\%. Simulation results indicate that transfer learning improves prediction accuracy by approximately 10\% compared to direct prediction without transfer. This study demonstrates the effectiveness of transfer learning in overcoming the physical challenges associated with limited labeled/unlabeled data, highlights the promising potential of transfer learning in the field of muon tomography.

physics.ins-det↗

arXiv2025

Contrastive Learning Guided Latent Diffusion Model for Image-to-Image Translation

The diffusion model has demonstrated superior performance in synthesizing diverse and high-quality images for text-guided image translation. However, there remains room for improvement in both the formulation of text prompts and the preservation of reference image content. First, variations in target text prompts can significantly influence the quality of the generated images, and it is often challenging for users to craft an optimal prompt that fully captures the content of the input image. Second, while existing models can introduce desired modifications to specific regions of the reference image, they frequently induce unintended alterations in areas that should remain unchanged. To address these challenges, we propose pix2pix-zeroCon, a zero-shot diffusion-based method that eliminates the need for additional training by leveraging patch-wise contrastive loss. Specifically, we automatically determine the editing direction in the text embedding space based on the reference image and target prompts. Furthermore, to ensure precise content and structural preservation in the edited image, we introduce cross-attention guiding loss and patch-wise contrastive loss between the generated and original image embeddings within a pre-trained diffusion model. Notably, our approach requires no additional training and operates directly on a pre-trained text-to-image diffusion model. Extensive experiments demonstrate that our method surpasses existing models in image-to-image translation, achieving enhanced fidelity and controllability.

cs.CV↗

arXiv2025

Highly Entangled 2D Ground States: Tensor Network, Order Parameter and Correlation

In this article we present analytical results on the exact tensor network representations and correlation functions of the first examples of 2D ground states with quantum phase transitions between area law and extensive entanglement entropy. The tensor networks constructed are one dimension higher than the lattices of the physical systems, allowing entangled physical degrees of freedoms to be paired with one another arbitrarily far away. Contraction rules of the internal legs are specified by a simple translationally invariant set of rules in terms of the tesselation of cubes or prisms in 3D space. The networks directly generalize the previous holographic tensor networks for 1D Fredkin and Motzkin chains. We also analyze the correlation in the spin and color sectors from the scaling of the height function of random surfaces, revealing additional characterizations of the exotic phase transitions.

quant-ph↗

arXiv2025

Bridging Social Psychology and LLM Reasoning: Conflict-Aware Meta-Review Generation via Cognitive Alignment

The rapid growth of scholarly submissions has overwhelmed traditional peer review systems, driving the need for intelligent automation to preserve scientific rigor. While large language models (LLMs) show promise in automating manuscript critiques, their ability to synthesize high-stakes meta-reviews, which require conflict-aware reasoning and consensus derivation, remains underdeveloped. Existing methods fail to effectively handle conflicting viewpoints within differing opinions, and often introduce additional cognitive biases, such as anchoring effects and conformity bias.To overcome these limitations, we propose the Cognitive Alignment Framework (CAF), a dual-process architecture that transforms LLMs into adaptive scientific arbitrators. By operationalizing Kahneman's dual-process theory, CAF introduces a three-step cognitive pipeline: review initialization, incremental integration, and cognitive alignment.Empirical validation shows that CAF outperforms existing LLM-based methods, with sentiment consistency gains reaching up to 19.47\% and content consistency improving by as much as 12.95\%.

cs.AI↗