arXiv subjects

Le-Minh Nguyen

Publications and source records attributed to Le-Minh Nguyen.

At least 19 recordsLinked to original sources

arXiv2026

L-MAD: A Systematic Evaluation of Multi-Agent Debate Structures in Legal Reasoning

While multi-agent debate (MAD) frameworks have shown significant potential in general reasoning, their effectiveness in highly structured, knowledge-heavy legal domains remains under-explored. In this work, we introduce the Legal Multi-Agent Debate (L-MAD) framework to systematically evaluate different debate structures and aggregation methods within Legal Textual Entailment. By assigning distinct expert personas to multiple agents, L-MAD improves upon strong single-agent baselines by up to 8\%. Furthermore, analyzing how debate scales reveals a clear trade-off: increasing the agent population reduces inconsistency and improves accuracy, whereas extending discussion rounds induces a detrimental \textit{over-deliberation drift} where agents reinforce each other's mistakes. Ultimately, our findings outline the practical boundaries and safety margins of deploying collaborative multi-agent systems in high-stakes legal reasoning environments.

cs.AI↗

arXiv2026

Beyond Forgetting: Machine Unlearning Elicits Controllable Side Behaviors and Capabilities

We consider Representation Misdirection (RM), a class of large language model (LLM) unlearning methods that achieve forgetting by redirecting the forget-representations, that is, latent representations of forget-samples, toward a target vector. Despite being important, the roles of the target vector used in RM, however, remain underexplored. Here, we approach and revisit RM through the lens of the Linear Representation Hypothesis. Specifically, if one can identify a one-dimensional representation corresponding to a high-level concept, the Linear Representation Hypothesis enables linear operations on this concept vector within the forget-representation space. Under this view, we hypothesize that, beyond forgetting, machine unlearning via RM elicits controllable emergent side behaviors and stronger side capabilities corresponding to the high-level concept. Our hypothesis is empirically validated across a wide range of tasks, including behavioral control (e.g., controlling unlearned models' truthfulness, sentiment, refusal, and language) and capability enhancement (e.g., improving unlearned models' in-context learning (ICL) capability). Our findings reveal that this phenomenon could be either a hidden risk if misused or a mechanism that can be harnessed for developing unlearned models that require stronger capabilities and controllable behaviors.

cs.LG↗

arXiv2026

RegMean++: Enhancing Effectiveness and Generalization of Regression Mean for Model Merging

Regression Mean (RegMean), an approach that formulates model merging as a linear regression problem, aims to find the optimal weights for each linear layer in the merged model by minimizing the discrepancy in predictions between the merged and candidate models. RegMean provides a precise closed-form solution for the merging problem; therefore, it offers explainability and computational efficiency. However, RegMean merges each linear layer independently, overlooking how the features and information in earlier layers propagate through deeper layers and influence the final predictions of the merged model. Here, we introduce RegMean++, a simple yet effective alternative to RegMean, that explicitly incorporates both intra-layer and cross-layer dependencies between merged models' layers into RegMean's objective. By accounting for these dependencies, RegMean++ better captures the behaviors of the merged model. Extensive experiments demonstrate that RegMean++ consistently outperforms RegMean across diverse settings, including in-domain (ID) and out-of-domain (OOD) generalization, sequential merging, large-scale tasks, and robustness under several types of distribution shifts. Furthermore, RegMean++ achieves competitive performance across diverse settings compared to various advanced model merging methods.

cs.LG↗

arXiv2026

Improving LLM Unlearning Robustness via Random Perturbations

Here, we show that current LLM unlearning methods inherently reduce models' robustness, causing them to misbehave even when a single non-adversarial forget-token is present in the retain-query. Toward understanding underlying causes, we propose a novel theoretical framework that reframes the unlearning process as a backdoor attack and defense problem: we formulate how the forgetting process inadvertently learns to align forget-tokens (backdoor triggers) with the target-representations (target labels). As a result, forget-tokens act as backdoor triggers that, when activated in retain-queries, cause disruptions in unlearned models' behaviors, similar to successful backdoor attacks. The sense that, LLM unlearning methods themselves poison the model, make it more vulnerable to forget-tokens, and hide rather than erase target knowledge, describes their true mechanism. To mitigate the vulnerability caused by the forgetting process, we reinterpret the retaining process as a backdoor defense and propose Random Noise Augmentation (RNA), a lightweight, model and method-agnostic approach with theoretical guarantees for improving the robustness of unlearned models. Extensive experiments demonstrate that RNA significantly improves the robustness of unlearned models while preserving forget and retain performances. This backdoor attack-defense framework offers insights into the mechanism of unlearning that can shed light on future research directions for improving unlearning robustness.

cs.CL↗

arXiv2026

Fact4ac at the Financial Misinformation Detection Challenge Task: Reference-Free Financial Misinformation Detection via Fine-Tuning and Few-Shot Prompting of Large Language Models

The proliferation of financial misinformation poses a severe threat to market stability and investor trust, misleading market behavior and creating critical information asymmetry. Detecting such misleading narratives is inherently challenging, particularly in real-world scenarios where external evidence or supplementary references for cross-verification are strictly unavailable. This paper presents our winning methodology for the "Reference-Free Financial Misinformation Detection" shared task. Built upon the recently proposed RFC-BENCH framework (Jiang et al. 2026), this task challenges models to determine the veracity of financial claims by relying solely on internal semantic understanding and contextual consistency, rather than external fact-checking. To address this formidable evaluation setup, we propose a comprehensive framework that capitalizes on the reasoning capabilities of state-of-the-art Large Language Models (LLMs). Our approach systematically integrates in-context learning, specifically zero-shot and few-shot prompting strategies, with Parameter-Efficient Fine-Tuning (PEFT) via Low-Rank Adaptation (LoRA) to optimally align the models with the subtle linguistic cues of financial manipulation. Our proposed system demonstrated superior efficacy, successfully securing the first-place ranking on both official leaderboards. Specifically, we achieved an accuracy of 95.4% on the public test set and 96.3% on the private test set, highlighting the robustness of our method and contributing to the acceleration of context-aware misinformation detection in financial Natural Language Processing. Our models (14B and 32B) are available at https://huggingface.co/KaiNKaiho.

cs.CL↗

arXiv2026

Can Legislation Be Made Machine-Readable in PROLEG?

The anticipated positive social impact of regulatory processes requires both the accuracy and efficiency of their application. Modern artificial intelligence technologies, including natural language processing and machine-assisted reasoning, hold great promise for addressing this challenge. We present a framework to address the challenge of tools for regulatory application, based on current state-of-the-art (SOTA) methods for natural language processing (large language models or LLMs) and formalization of legal reasoning (the legal representation system PROLEG). As an example, we focus on Article 6 of the European General Data Protection Regulation (GDPR). In our framework, a single LLM prompt simultaneously transforms legal text into if-then rules and a corresponding PROLEG encoding, which are then validated and refined by legal domain experts. The final output is an executable PROLEG program that can produce human-readable explanations for instances of GDPR decisions. We describe processes to support the end-to-end transformation of a segment of a regulatory document (Article 6 from GDPR), including the prompting frame to guide an LLM to "compile" natural language text to if-then rules, then to further "compile" the vetted if-then rules to PROLEG. Finally, we produce an instance that shows the PROLEG execution. We conclude by summarizing the value of this approach and note observed limitations with suggestions to further develop such technologies for capturing and deploying regulatory frameworks.

cs.CL↗

arXiv2025

Finetuning LLMs for Automatic Form Interaction on Web-Browser in Selenium Testing Framework

Automated web application testing is a critical component of modern software development, with frameworks like Selenium widely adopted for validating functionality through browser automation. Among the essential aspects of such testing is the ability to interact with and validate web forms, a task that requires syntactically correct, executable scripts with high coverage of input fields. Despite its importance, this task remains underexplored in the context of large language models (LLMs), and no public benchmark or dataset exists to evaluate LLMs on form interaction generation systematically. This paper introduces a novel method for training LLMs to generate high-quality test cases in Selenium, specifically targeting form interaction testing. We curate both synthetic and human-annotated datasets for training and evaluation, covering diverse real-world forms and testing scenarios. We define clear metrics for syntax correctness, script executability, and input field coverage. Our empirical study demonstrates that our approach significantly outperforms strong baselines, including GPT-4o and other popular LLMs, across all evaluation metrics. Our work lays the groundwork for future research on LLM-based web testing and provides resources to support ongoing progress in this area.

cs.SE↗

arXiv2025

VLSP 2025 MLQA-TSR Challenge: Vietnamese Multimodal Legal Question Answering on Traffic Sign Regulation

This paper presents the VLSP 2025 MLQA-TSR - the multimodal legal question answering on traffic sign regulation shared task at VLSP 2025. VLSP 2025 MLQA-TSR comprises two subtasks: multimodal legal retrieval and multimodal question answering. The goal is to advance research on Vietnamese multimodal legal text processing and to provide a benchmark dataset for building and evaluating intelligent systems in multimodal legal domains, with a focus on traffic sign regulation in Vietnam. The best-reported results on VLSP 2025 MLQA-TSR are an F2 score of 64.55% for multimodal legal retrieval and an accuracy of 86.30% for multimodal question answering.

cs.CL↗

arXiv2025

Overview of SCIDOCA 2025 Shared Task on Citation Prediction, Discovery, and Placement

We present an overview of the SCIDOCA 2025 Shared Task, which focuses on citation discovery and prediction in scientific documents. The task is divided into three subtasks: (1) Citation Discovery, where systems must identify relevant references for a given paragraph; (2) Masked Citation Prediction, which requires selecting the correct citation for masked citation slots; and (3) Citation Sentence Prediction, where systems must determine the correct reference for each cited sentence. We release a large-scale dataset constructed from the Semantic Scholar Open Research Corpus (S2ORC), containing over 60,000 annotated paragraphs and a curated reference set. The test set consists of 1,000 paragraphs from distinct papers, each annotated with ground-truth citations and distractor candidates. A total of seven teams registered, with three submitting results. We report performance metrics across all subtasks and analyze the effectiveness of submitted systems. This shared task provides a new benchmark for evaluating citation modeling and encourages future research in scientific document understanding. The dataset and task materials are publicly available at https://github.com/daotuanan/scidoca2025-shared-task.

cs.DL↗

arXiv2025

NOWJ@COLIEE 2025: A Multi-stage Framework Integrating Embedding Models and Large Language Models for Legal Retrieval and Entailment

This paper presents the methodologies and results of the NOWJ team's participation across all five tasks at the COLIEE 2025 competition, emphasizing advancements in the Legal Case Entailment task (Task 2). Our comprehensive approach systematically integrates pre-ranking models (BM25, BERT, monoT5), embedding-based semantic representations (BGE-m3, LLM2Vec), and advanced Large Language Models (Qwen-2, QwQ-32B, DeepSeek-V3) for summarization, relevance scoring, and contextual re-ranking. Specifically, in Task 2, our two-stage retrieval system combined lexical-semantic filtering with contextualized LLM analysis, achieving first place with an F1 score of 0.3195. Additionally, in other tasks--including Legal Case Retrieval, Statute Law Retrieval, Legal Textual Entailment, and Legal Judgment Prediction--we demonstrated robust performance through carefully engineered ensembles and effective prompt-based reasoning strategies. Our findings highlight the potential of hybrid models integrating traditional IR techniques with contemporary generative models, providing a valuable reference for future advancements in legal information processing.

cs.CL↗

arXiv2024

ZeFaV: Boosting Large Language Models for Zero-shot Fact Verification

In this paper, we propose ZeFaV - a zero-shot based fact-checking verification framework to enhance the performance on fact verification task of large language models by leveraging the in-context learning ability of large language models to extract the relations among the entities within a claim, re-organized the information from the evidence in a relationally logical form, and combine the above information with the original evidence to generate the context from which our fact-checking model provide verdicts for the input claims. We conducted empirical experiments to evaluate our approach on two multi-hop fact-checking datasets including HoVer and FEVEROUS, and achieved potential results results comparable to other state-of-the-art fact verification task methods.

cs.CL↗

arXiv2024

A Mutual Inclusion Mechanism for Precise Boundary Segmentation in Medical Images

In medical imaging, accurate image segmentation is crucial for quantifying diseases, assessing prognosis, and evaluating treatment outcomes. However, existing methods lack an in-depth integration of global and local features, failing to pay special attention to abnormal regions and boundary details in medical images. To this end, we present a novel deep learning-based approach, MIPC-Net, for precise boundary segmentation in medical images. Our approach, inspired by radiologists' working patterns, features two distinct modules: (i) \textbf{Mutual Inclusion of Position and Channel Attention (MIPC) module}: To enhance the precision of boundary segmentation in medical images, we introduce the MIPC module, which enhances the focus on channel information when extracting position features and vice versa; (ii) \textbf{GL-MIPC-Residue}: To improve the restoration of medical images, we propose the GL-MIPC-Residue, a global residual connection that enhances the integration of the encoder and decoder by filtering out invalid information and restoring the most effective information lost during the feature extraction process. We evaluate the performance of the proposed model using metrics such as Dice coefficient (DSC) and Hausdorff Distance (HD) on three publicly accessible datasets: Synapse, ISIC2018-Task, and Segpc. Our ablation study shows that each module contributes to improving the quality of segmentation results. Furthermore, with the assistance of both modules, our approach outperforms state-of-the-art methods across all metrics on the benchmark datasets, notably achieving a 2.23mm reduction in HD on the Synapse dataset, strongly evidencing our model's enhanced capability for precise image boundary segmentation. Codes will be available at https://github.com/SUN-1024/MIPC-Net.

eess.IV↗

arXiv2024

VLSP 2023 -- LTER: A Summary of the Challenge on Legal Textual Entailment Recognition

In this new era of rapid AI development, especially in language processing, the demand for AI in the legal domain is increasingly critical. In the context where research in other languages such as English, Japanese, and Chinese has been well-established, we introduce the first fundamental research for the Vietnamese language in the legal domain: legal textual entailment recognition through the Vietnamese Language and Speech Processing workshop. In analyzing participants' results, we discuss certain linguistic aspects critical in the legal domain that pose challenges that need to be addressed.

cs.CL↗

arXiv2024

Employing Label Models on ChatGPT Answers Improves Legal Text Entailment Performance

The objective of legal text entailment is to ascertain whether the assertions in a legal query logically follow from the information provided in one or multiple legal articles. ChatGPT, a large language model, is robust in many natural language processing tasks, including legal text entailment: when we set the temperature = 0 (the ChatGPT answers are deterministic) and prompt the model, it achieves 70.64% accuracy on COLIEE 2022 dataset, which outperforms the previous SOTA of 67.89%. On the other hand, if the temperature is larger than zero, ChatGPT answers are not deterministic, leading to inconsistent answers and fluctuating results. We propose to leverage label models (a fundamental component of weak supervision techniques) to integrate the provisional answers by ChatGPT into consolidated labels. By that way, we treat ChatGPT provisional answers as noisy predictions which can be consolidated by label models. The experimental results demonstrate that this approach can attain an accuracy of 76.15%, marking a significant improvement of 8.26% over the prior state-of-the-art benchmark. Additionally, we perform an analysis of the instances where ChatGPT produces incorrect answers, then we classify the errors, offering insights that could guide potential enhancements for future research endeavors.

cs.CL↗

arXiv2024

CAPTAIN at COLIEE 2023: Efficient Methods for Legal Information Retrieval and Entailment Tasks

The Competition on Legal Information Extraction/Entailment (COLIEE) is held annually to encourage advancements in the automatic processing of legal texts. Processing legal documents is challenging due to the intricate structure and meaning of legal language. In this paper, we outline our strategies for tackling Task 2, Task 3, and Task 4 in the COLIEE 2023 competition. Our approach involved utilizing appropriate state-of-the-art deep learning methods, designing methods based on domain characteristics observation, and applying meticulous engineering practices and methodologies to the competition. As a result, our performance in these tasks has been outstanding, with first places in Task 2 and Task 3, and promising results in Task 4. Our source code is available at https://github.com/Nguyen2015/CAPTAIN-COLIEE2023/tree/coliee2023.

cs.CL↗

arXiv2023

DA-TransUNet: Integrating Spatial and Channel Dual Attention with Transformer U-Net for Medical Image Segmentation

Accurate medical image segmentation is critical for disease quantification and treatment evaluation. While traditional Unet architectures and their transformer-integrated variants excel in automated segmentation tasks. However, they lack the ability to harness the intrinsic position and channel features of image. Existing models also struggle with parameter efficiency and computational complexity, often due to the extensive use of Transformers. To address these issues, this study proposes a novel deep medical image segmentation framework, called DA-TransUNet, aiming to integrate the Transformer and dual attention block(DA-Block) into the traditional U-shaped architecture. Unlike earlier transformer-based U-net models, DA-TransUNet utilizes Transformers and DA-Block to integrate not only global and local features, but also image-specific positional and channel features, improving the performance of medical image segmentation. By incorporating a DA-Block at the embedding layer and within each skip connection layer, we substantially enhance feature extraction capabilities and improve the efficiency of the encoder-decoder structure. DA-TransUNet demonstrates superior performance in medical image segmentation tasks, consistently outperforming state-of-the-art techniques across multiple datasets. In summary, DA-TransUNet offers a significant advancement in medical image segmentation, providing an effective and powerful alternative to existing techniques. Our architecture stands out for its ability to improve segmentation accuracy, thereby advancing the field of automated medical image diagnostics. The codes and parameters of our model will be publicly available at https://github.com/SUN-1024/DA-TransUnet.

eess.IV↗

arXiv2023

Improving Vietnamese Legal Question--Answering System based on Automatic Data Enrichment

Question answering (QA) in law is a challenging problem because legal documents are much more complicated than normal texts in terms of terminology, structure, and temporal and logical relationships. It is even more difficult to perform legal QA for low-resource languages like Vietnamese where labeled data are rare and pre-trained language models are still limited. In this paper, we try to overcome these limitations by implementing a Vietnamese article-level retrieval-based legal QA system and introduce a novel method to improve the performance of language models by improving data quality through weak labeling. Our hypothesis is that in contexts where labeled data are limited, efficient data enrichment can help increase overall performance. Our experiments are designed to test multiple aspects, which demonstrate the effectiveness of the proposed technique.

cs.CL↗

arXiv2022

Law to Binary Tree -- An Formal Interpretation of Legal Natural Language

Knowledge representation and reasoning in law are essential to facilitate the automation of legal analysis and decision-making tasks. In this paper, we propose a new approach based on legal science, specifically legal taxonomy, for representing and reasoning with legal documents. Our approach interprets the regulations in legal documents as binary trees, which facilitates legal reasoning systems to make decisions and resolve logical contradictions. The advantages of this approach are twofold. First, legal reasoning can be performed on the basis of the binary tree representation of the regulations. Second, the binary tree representation of the regulations is more understandable than the existing sentence-based representations. We provide an example of how our approach can be used to interpret the regulations in a legal document.

cs.CL↗