arXiv subjects

Jing Li

Publications and source records attributed to Jing Li.

At least 289 records · Page 16Linked to original sources

arXiv2025

Adaptive Detoxification: Safeguarding General Capabilities of LLMs through Toxicity-Aware Knowledge Editing

Large language models (LLMs) exhibit impressive language capabilities but remain vulnerable to malicious prompts and jailbreaking attacks. Existing knowledge editing methods for LLM detoxification face two major challenges. First, they often rely on entity-specific localization, making them ineffective against adversarial inputs without explicit entities. Second, these methods suffer from over-editing, where detoxified models reject legitimate queries, compromising overall performance. In this paper, we propose ToxEdit, a toxicity-aware knowledge editing approach that dynamically detects toxic activation patterns during forward propagation. It then routes computations through adaptive inter-layer pathways to mitigate toxicity effectively. This design ensures precise toxicity mitigation while preserving LLMs' general capabilities. To more accurately assess over-editing, we also enhance the SafeEdit benchmark by incorporating instruction-following evaluation tasks. Experimental results on multiple LLMs demonstrate that our ToxEdit outperforms previous state-of-the-art methods in both detoxification performance and safeguarding general capabilities of LLMs.

cs.CL↗

arXiv2025

Self-Error-Instruct: Generalizing from Errors for LLMs Mathematical Reasoning

Although large language models demonstrate strong performance across various domains, they still struggle with numerous bad cases in mathematical reasoning. Previous approaches to learning from errors synthesize training data by solely extrapolating from isolated bad cases, thereby failing to generalize the extensive patterns inherent within these cases. This paper presents Self-Error-Instruct (SEI), a framework that addresses these model weaknesses and synthesizes more generalized targeted training data. Specifically, we explore a target model on two mathematical datasets, GSM8K and MATH, to pinpoint bad cases. Then, we generate error keyphrases for these cases based on the instructor model's (GPT-4o) analysis and identify error types by clustering these keyphrases. Next, we sample a few bad cases during each generation for each identified error type and input them into the instructor model, which synthesizes additional training data using a self-instruct approach. This new data is refined through a one-shot learning process to ensure that only the most effective examples are kept. Finally, we use these curated data to fine-tune the target model, iteratively repeating the process to enhance performance. We apply our framework to various models and observe improvements in their reasoning abilities across both in-domain and out-of-domain mathematics datasets. These results demonstrate the effectiveness of self-error instruction in improving LLMs' mathematical reasoning through error generalization.

cs.CL↗

arXiv2025

DRPruning: Efficient Large Language Model Pruning through Distributionally Robust Optimization

Large language models (LLMs) deliver impressive results but face challenges from increasing model sizes and computational costs. Structured pruning reduces model size and speeds up inference but often causes uneven degradation across domains, leading to biased performance. To address this, we propose DRPruning, a method that dynamically adjusts the data distribution during training to restore balanced performance across heterogeneous and multi-tasking data. Experiments in monolingual and multilingual settings show that DRPruning surpasses similarly sized models in both pruning and continued pretraining over perplexity, downstream tasks, and instruction tuning. Further analysis demonstrates the robustness of DRPruning towards various domains and distribution shifts. Furthermore, DRPruning can determine optimal reference losses and data ratios automatically, suggesting potential for broader applications. Code and scripts are available at https://github.com/hexuandeng/DRPruning.

cs.CL↗

arXiv2025

Multi-objective Large Language Model Alignment with Hierarchical Experts

Aligning large language models (LLMs) to simultaneously satisfy multiple objectives remains a significant challenge, especially given the diverse and often conflicting nature of human preferences. Existing alignment methods struggle to balance trade-offs effectively, often requiring costly retraining or yielding suboptimal results across the Pareto frontier of preferences. In this paper, we introduce \textit{HoE}(Hierarchical Mixture-of-Experts), a \textit{lightweight}, \textit{parameter-efficient}, and \textit{plug-and-play} approach that eliminates the need for model training, while enabling LLMs to adapt across the entire Pareto frontier and accommodate diverse user preferences. In particular, \textit{HoE} consists of three hierarchical components: LoRA Experts, Router Experts and Preference Routing, reaching optimal Pareto frontiers and achieving a trade-off between parameter size, training cost, and performance. We evaluate \textit{HoE} across various tasks on 14 objectives and 200 different preferences among 6 benchmarks, demonstrating superior performance over 15 recent baselines. Code is available in the supplementary materials.

cs.CL↗

arXiv2025

Neural Parameter Search for Slimmer Fine-Tuned Models and Better Transfer

Foundation models and their checkpoints have significantly advanced deep learning, boosting performance across various applications. However, fine-tuned models often struggle outside their specific domains and exhibit considerable redundancy. Recent studies suggest that combining a pruned fine-tuned model with the original pre-trained model can mitigate forgetting, reduce interference when merging model parameters across tasks, and improve compression efficiency. In this context, developing an effective pruning strategy for fine-tuned models is crucial. Leveraging the advantages of the task vector mechanism, we preprocess fine-tuned models by calculating the differences between them and the original model. Recognizing that different task vector subspaces contribute variably to model performance, we introduce a novel method called Neural Parameter Search (NPS-Pruning) for slimming down fine-tuned models. This method enhances pruning efficiency by searching through neural parameters of task vectors within low-rank subspaces. Our method has three key applications: enhancing knowledge transfer through pairwise model interpolation, facilitating effective knowledge fusion via model merging, and enabling the deployment of compressed models that retain near-original performance while significantly reducing storage costs. Extensive experiments across vision, NLP, and multi-modal benchmarks demonstrate the effectiveness and robustness of our approach, resulting in substantial performance gains. The code is publicly available at: https://github.com/duguodong7/NPS-Pruning.

cs.LG↗

arXiv2025

Quantum Multi-view Kernel Learning with Local Information

Kernel methods serve as powerful tools to capture nonlinear patterns behind data in machine learning. The quantum kernel, integrating kernel theory with quantum computing, has attracted widespread attention. However, existing studies encounter performance bottlenecks when processing complex data with localized structural patterns, stemming from the limitation in single-view feature representation and the exclusive reliance on global data structure. In this paper, we propose quantum multi-view kernel learning with local information, called L-QMVKL. Specifically, based on the multi-kernel learning, a representative method for multi-view data processing, we construct the quantum multi-kernel that combines view-specific quantum kernels to effectively fuse cross-view information. Further leveraging local information to capture intrinsic structural information, we design a sequential training strategy for the quantum circuit parameters and weight coefficients with the use of the hybrid global-local kernel alignment. We evaluate the effectiveness of L-QMVKL through comprehensive numerical simulations on the Mfeat dataset, demonstrating significant accuracy improvements achieved through leveraging multi-view methodology and local information. Meanwhile, the results show that L-QMVKL exhibits a higher accuracy than its classical counterpart. Our work holds promise for advancing the theoretical and practical understanding of quantum kernel methods.

quant-ph↗

arXiv2025

MTSA: Multi-turn Safety Alignment for LLMs through Multi-round Red-teaming

The proliferation of jailbreak attacks against large language models (LLMs) highlights the need for robust security measures. However, in multi-round dialogues, malicious intentions may be hidden in interactions, leading LLMs to be more prone to produce harmful responses. In this paper, we propose the \textbf{M}ulti-\textbf{T}urn \textbf{S}afety \textbf{A}lignment (\ourapproach) framework, to address the challenge of securing LLMs in multi-round interactions. It consists of two stages: In the thought-guided attack learning stage, the red-team model learns about thought-guided multi-round jailbreak attacks to generate adversarial prompts. In the adversarial iterative optimization stage, the red-team model and the target model continuously improve their respective capabilities in interaction. Furthermore, we introduce a multi-turn reinforcement learning algorithm based on future rewards to enhance the robustness of safety alignment. Experimental results show that the red-team model exhibits state-of-the-art attack capabilities, while the target model significantly improves its performance on safety benchmarks.

cs.CR↗

arXiv2025

Multi-Modality Expansion and Retention for LLMs through Parameter Merging and Decoupling

Fine-tuning Large Language Models (LLMs) with multimodal encoders on modality-specific data expands the modalities that LLMs can handle, leading to the formation of Multimodal LLMs (MLLMs). However, this paradigm heavily relies on resource-intensive and inflexible fine-tuning from scratch with new multimodal data. In this paper, we propose MMER (Multi-modality Expansion and Retention), a training-free approach that integrates existing MLLMs for effective multimodal expansion while retaining their original performance. Specifically, MMER reuses MLLMs' multimodal encoders while merging their LLM parameters. By comparing original and merged LLM parameters, MMER generates binary masks to approximately separate LLM parameters for each modality. These decoupled parameters can independently process modality-specific inputs, reducing parameter conflicts and preserving original MLLMs' fidelity. MMER can also mitigate catastrophic forgetting by applying a similar process to MLLMs fine-tuned on new tasks. Extensive experiments show significant improvements over baselines, proving that MMER effectively expands LLMs' multimodal capabilities while retaining 99% of the original performance, and also markedly mitigates catastrophic forgetting.

cs.CL↗

arXiv2025

Convergence Analysis of an Adaptive Nonconforming FEM for Phase-Field Dependent Topology Optimization in Stokes Flow

In this work, we develop an adaptive nonconforming finite element algorithm for the numerical approximation of phase-field parameterized topology optimization governed by the Stokes system. We employ the conforming linear finite element space to approximate the phase field, and the nonconforming linear finite elements (Crouzeix-Raviart elements) and piecewise constants to approximate the velocity field and the pressure field, respectively. We establish the convergence of the adaptive method, i.e., the sequence of minimizers contains a subsequence that converges to a solution of the first-order optimality system, and the associated subsequence of discrete pressure fields also converges. The analysis relies crucially on a new discrete compactness result of nonconforming linear finite elements over a sequence of adaptively generated meshes. We present numerical results for several examples to illustrate the performance of the algorithm, including a comparison with the uniform refinement strategy.

math.NA↗

arXiv2025

A Survey of Real-time Scheduling on Accelerator-based Heterogeneous Architecture for Time Critical Applications

Accelerator-based heterogeneous architectures, such as CPU-GPU, CPU-TPU, and CPU-FPGA systems, are widely adopted to support the popular artificial intelligence (AI) algorithms that demand intensive computation. When deployed in real-time applications, such as robotics and autonomous vehicles, these architectures must meet stringent timing constraints. To summarize these achievements, this article presents a comprehensive survey of real-time scheduling techniques for accelerator-based heterogeneous platforms. It highlights key advancements from the past ten years, showcasing how proposed solutions have evolved to address the distinct challenges and requirements of these systems. This survey begins with an overview of the hardware characteristics and common task execution models used in accelerator-based heterogeneous systems. It then categorizes the reviewed works based on soft and hard deadline constraints. For soft real-time approaches, we cover real-time scheduling methods supported by hardware vendors and strategies focusing on timing-critical scheduling, energy efficiency, and thermal-aware scheduling. For hard real-time approaches, we first examine support from processor vendors. We then discuss scheduling techniques that guarantee hard deadlines (with strict response time analysis). After reviewing general soft and hard real-time scheduling methods, we explore application- or scenario-driven real-time scheduling techniques for accelerator-enabled heterogeneous computing platforms. Finally, the article concludes with a discussion of open issues and challenges within this research area.

cs.DC↗

arXiv2025

Integrated Data Analysis of Plasma Electron Density Profile Tomography for HL-3 with Gaussian Process Regression

An integrated data analysis model based on Gaussian Process Regression is proposed for plasma electron density profile tomography in the HL-3 tokamak. The model combines line-integral measurements from the far-infrared laser interferometer with point measurements obtained via the frequency-modulated continuous wave reflectometry. By employing Gaussian Process Regression, the model effectively incorporates point measurements into 2D profile reconstructions, while coordinate mapping integrates magnetic equilibrium information. The average relative error of the reconstructed profile obtained by the integrated data analysis model with normalized magnetic flux is as low as 3.60*10^(-4). Additionally, sensitivity tests were conducted on the grid resolution, the standard deviation of diagnostic data, and noise levels, providing a robust foundation for the real application to experimental data.

cs.LG↗

arXiv2025

On the Crouzeix-Raviart Finite Element Approximation of Phase-Field Dependent Topology Optimization in Stokes Flow

In this work, we investigate a nonconforming finite element approximation of phase-field parameterized topology optimization governed by the Stokes flow. The phase field, the velocity field and the pressure field are approximated by conforming linear finite elements, nonconforming linear finite elements (Crouzeix-Raviart elements) and piecewise constants, respectively. When compared with the standard conforming counterpart, the nonconforming FEM can provide an approximation with fewer degrees of freedom, leading to improved computational efficiency. We establish the convergence of the resulting numerical scheme in the sense that the sequences of phase-field functions and discrete velocity fields contain subsequences that converge to a minimizing pair of the continuous problem in the $H^1$-norm and a mesh-dependent norm, respectively. We present extensive numerical results to illustrate the performance of the approach, including a comparison with the popular Taylor-Hood elements.

math.NA↗

arXiv2025

SmoothSegNet: A Global-Local Framework for Liver Tumor Segmentation with Clinical KnowledgeInformed Label Smoothing

Liver cancer is a leading cause of mortality worldwide, and accurate Computed Tomography (CT)-based tumor segmentation is essential for diagnosis and treatment. Manual delineation is time-intensive, prone to variability, and highlights the need for reliable automation. While deep learning has shown promise for automated liver segmentation, precise liver tumor segmentation remains challenging due to the heterogeneous nature of tumors, imprecise tumor margins, and limited labeled data. We present SmoothSegNet, a novel deep learning framework that addresses these challenges with the three key designs: (1) A novel knowledge-informed label smoothing technique that distills knowledge from clinical data to generate smooth labels, which are used to regularize model training, reducing the overfitting risk and enhancing model performance; (2) A global and local segmentation framework that breaks down the main task into two simpler sub-tasks, allowing optimized preprocessing and training for each; and (3) Pre- and post-processing pipelines customized to the challenges of each subtask aimed to enhance tumor visibility and refines tumor boundaries. We apply the proposed model on a challenging HCC-TACE-Seg dataset and show that SmoothSegNet outperformed various benchmarks in segmentation performance, particularly at smaller tumors (<10cm). Our ablation studies show that the three design components complementarily contribute to the model improved performance. Code for the proposed method are available at https://github.com/lingchm/medassist-liver-cancer.

cs.LG↗

arXiv2025

Human-In-The-Loop Software Development Agents: Challenges and Future Directions

Multi-agent LLM-driven systems for software development are rapidly gaining traction, offering new opportunities to enhance productivity. At Atlassian, we deployed Human-in-the-Loop Software Development Agents to resolve Jira work items and evaluated the generated code quality using functional correctness testing and GPT-based similarity scoring. This paper highlights two major challenges: the high computational costs of unit testing and the variability in LLM-based evaluations. We also propose future research directions to improve evaluation frameworks for Human-In-The-Loop software development tools.

cs.SE↗

arXiv2025

Hear the Scene: Audio-Enhanced Text Spotting

Recent advancements in scene text spotting have focused on end-to-end methodologies that heavily rely on precise location annotations, which are often costly and labor-intensive to procure. In this study, we introduce an innovative approach that leverages only transcription annotations for training text spotting models, substantially reducing the dependency on elaborate annotation processes. Our methodology employs a query-based paradigm that facilitates the learning of implicit location features through the interaction between text queries and image embeddings. These features are later refined during the text recognition phase using an attention activation map. Addressing the challenges associated with training a weakly-supervised model from scratch, we implement a circular curriculum learning strategy to enhance model convergence. Additionally, we introduce a coarse-to-fine cross-attention localization mechanism for more accurate text instance localization. Notably, our framework supports audio-based annotation, which significantly diminishes annotation time and provides an inclusive alternative for individuals with disabilities. Our approach achieves competitive performance against existing benchmarks, demonstrating that high accuracy in text spotting can be attained without extensive location annotations.

cs.CV↗

arXiv2025

S$^2$-MAD: Breaking the Token Barrier to Enhance Multi-Agent Debate Efficiency

Large language models (LLMs) have demonstrated remarkable capabilities across various natural language processing (NLP) scenarios, but they still face challenges when handling complex arithmetic and logical reasoning tasks. While Chain-Of-Thought (CoT) reasoning, self-consistency (SC) and self-correction strategies have attempted to guide models in sequential, multi-step reasoning, Multi-agent Debate (MAD) has emerged as a viable approach for enhancing the reasoning capabilities of LLMs. By increasing both the number of agents and the frequency of debates, the performance of LLMs improves significantly. However, this strategy results in a significant increase in token costs, presenting a barrier to scalability. To address this challenge, we introduce a novel sparsification strategy designed to reduce token costs within MAD. This approach minimizes ineffective exchanges of information and unproductive discussions among agents, thereby enhancing the overall efficiency of the debate process. We conduct comparative experiments on multiple datasets across various models, demonstrating that our approach significantly reduces the token costs in MAD to a considerable extent. Specifically, compared to MAD, our approach achieves an impressive reduction of up to 94.5\% in token costs while maintaining performance degradation below 2.0\%.

cs.CL↗

arXiv2025

TrajLearn: Trajectory Prediction Learning using Deep Generative Models

Trajectory prediction aims to estimate an entity's future path using its current position and historical movement data, benefiting fields like autonomous navigation, robotics, and human movement analytics. Deep learning approaches have become key in this area, utilizing large-scale trajectory datasets to model movement patterns, but face challenges in managing complex spatial dependencies and adapting to dynamic environments. To address these challenges, we introduce TrajLearn, a novel model for trajectory prediction that leverages generative modeling of higher-order mobility flows based on hexagonal spatial representation. TrajLearn predicts the next $k$ steps by integrating a customized beam search for exploring multiple potential paths while maintaining spatial continuity. We conducted a rigorous evaluation of TrajLearn, benchmarking it against leading state-of-the-art approaches and meaningful baselines. The results indicate that TrajLearn achieves significant performance gains, with improvements of up to ~40% across multiple real-world trajectory datasets. In addition, we evaluated different prediction horizons (i.e., various values of $k$), conducted resolution sensitivity analysis, and performed ablation studies to assess the impact of key model components. Furthermore, we developed a novel algorithm to generate mixed-resolution maps by hierarchically subdividing hexagonal regions into finer segments within a specified observation area. This approach supports selective detailing, applying finer resolution to areas of interest or high activity (e.g., urban centers) while using coarser resolution for less significant regions (e.g., rural areas), effectively reducing data storage requirements and computational overhead. We promote reproducibility and adaptability by offering complete code, data, and detailed documentation with flexible configuration options for various applications.

cs.LG↗

arXiv2025

Interference trapping of populations in a semi-infinite coupled-resonator waveguide

We study the energy structure and dynamics of a two-level emitter (2LE) locally coupled to a semi-infinite one-dimensional (1D) coupled-resonator array (CRA). The energy spectrum in the single-excitation subspace features a continuous band with scattering states, discrete levels with bound states, and a quantum phase transition characterized by the change of the number of bound states. The number of bound states is revealed by the behavior of the excited-state population at long times with quantum beat, residual oscillation, a constant with either non-zero or zero.

quant-ph↗