Institution profile

Newcastle University

Academic institutioneurope · gb
Official website
Research library256linked papers
Opportunities0open roles
Selected work

Representative Papers

Anti-bullying Adaptive Cruise Control: A proactive right-of-way protection approach

Dec 14, 2024arXiv.org

Existing adaptive cruise control (ACC) systems exhibit weak right-of-way protection under close-range cut-in maneuvers. This paper proposes Bullying-Resistant Adaptive Cruise Control (AACC), the first framework integrating online inverse optimal control (IOC)-driven driving style identification with Stackelberg game-theoretic interactive motion planning to enable real-time, personalized right-of-way preservation in right-hand traffic. Methodologically, AACC employs IOC to estimate the leading vehicle’s driving style online and formulates a Stackelberg game wherein the ego vehicle acts as leader and the cut-in vehicle as follower, yielding robust defensive trajectory plans. Experimental results demonstrate a substantial improvement in cut-in defense success rate, with safety and ride comfort enhanced by 79.8% and 20.4%, respectively, and traffic flow efficiency increased by 19.33%. Each planning step completes in under 50 ms, confirming feasibility for embedded real-time deployment.

3 citationsRead paper

NCL-BU at SemEval-2026 Task 3: Fine-tuning XLM-RoBERTa for Multilingual Dimensional Sentiment Regression

Apr 10, 2026

This study addresses multilingual dimensional aspect-based sentiment analysis, aiming to predict real-valued scores for each aspect in a text along two continuous affective dimensions—valence and arousal—within the range [1, 9]. Building upon XLM-RoBERTa-base, the authors employ a task-specific fine-tuning strategy that structures inputs as [CLS] text [SEP] aspect [SEP] and introduces a dual regression head with Sigmoid scaling to map outputs to the target range. Separate models are trained for English and Chinese across three domains: restaurants, laptops, and finance, with training and development sets merged for final evaluation. The work provides the first systematic demonstration that fine-tuning consistently and significantly outperforms few-shot prompting with prominent large language models such as GPT-5.2 and LLaMA-3/4, thereby establishing the superiority of fine-tuned approaches in multilingual dimensional sentiment regression.

1 citationsRead paper

RealFin: How Well Do LLMs Reason About Finance When Users Leave Things Unsaid?

Feb 06, 2026

This work addresses the overconfidence of large language models (LLMs) in financial reasoning, where they often produce incorrect answers due to failure in recognizing missing critical premises. To systematically evaluate this limitation, the authors introduce REALFIN, a bilingual benchmark that generates fluent yet premise-deficient questions by deliberately removing key information from original financial exam items. A multi-task evaluation framework is proposed to assess models’ capabilities across three dimensions: answering questions, identifying missing information, and proactively abstaining from responding when necessary. This study presents the first systematic assessment of LLMs’ awareness of implicit assumption gaps in financial contexts, revealing significant reliability shortcomings in both general-purpose and finance-specific models. The findings underscore that trustworthy financial reasoning requires models to possess the metacognitive judgment to distinguish known from unknown—embodying the principle of “knowing what one knows.”

1 citationsRead paper

Kernel-Based Learning of Safety Barriers

Jan 17, 2026

This work proposes a data-driven framework for safety verification and synthesis tailored to black-box AI systems operating in safety-critical settings with discrete-time stochastic dynamics. By constructing ambiguity sets in a reproducing kernel Hilbert space (RKHS) based on observed system trajectories, the approach leverages conditional mean embeddings to characterize uncertainty without requiring explicit knowledge of the system dynamics or noise distributions. A finite Fourier expansion is employed to transform the resulting semi-infinite optimization problem into a tractable linear program. The framework accommodates general temporal logic specifications and incorporates a distributionally robust mechanism to handle out-of-distribution behaviors. Empirical evaluations on black-box systems—including those with neural network controllers—demonstrate that the method ensures safety while maintaining strong scalability and robustness.

1 citationsRead paper

SoK: Decentralized AI (DeAI)

Nov 26, 2024arXiv.org

This paper addresses critical limitations of centralized AI—particularly proprietary large language models—including single-point failure, data bias, privacy leakage, and poor scalability. To this end, it proposes a blockchain-based Decentralized Artificial Intelligence (DeAI) framework. Methodologically, the study introduces the first taxonomy of DeAI protocols spanning the entire AI model lifecycle, systematically analyzing blockchain’s functional mappings in data contribution, model training, inference services, and governance. It integrates core technologies including consensus mechanisms, smart contracts, zero-knowledge proofs, decentralized storage, and incentive design. The contribution is a comprehensive DeAI analytical framework—the first of its kind—that comparatively characterizes existing approaches, identifies key research gaps, and outlines evolutionary pathways. This work provides both theoretical foundations and practical guidelines for building next-generation AI infrastructure that is transparent, secure, trustworthy, and equitably incentivized.

1 citationsRead paper
Recent publications

Latest Papers

The Quantum Plumber's Problem

Sep 10, 2026

研究通过扩展Elitzur-Vaidman炸弹测试场景，利用Kirkwood-Dirac准概率分布的负性，探讨了在‘量子水管工问题’中识别被阻路径的最佳策略。

0 citationsRead paper