arXiv subjects

Mengxiang Li

Publications and source records attributed to Mengxiang Li.

5 recordsLinked to original sources

arXiv2026

Beyond Static Personas: Situational Personality Steering for Large Language Models

Personalized Large Language Models (LLMs) facilitate more natural, human-like interactions in human-centric applications. However, existing personalization methods are constrained by limited controllability and high resource demands. Furthermore, their reliance on static personality modeling restricts adaptability across varying situations. To address these limitations, we first demonstrate the existence of situation-dependency and consistent situation-behavior patterns within LLM personalities through a multi-perspective analysis of persona neurons. Building on these insights, we propose IRIS, a training-free, neuron-based Identify-Retrieve-Steer framework for advanced situational personality steering. Our approach comprises situational persona neuron identification, situation-aware neuron retrieval, and similarity-weighted steering. We empirically validate our framework on PersonalityBench and our newly introduced SPBench, a comprehensive situational personality benchmark. Experimental results show that our method surpasses best-performing baselines, demonstrating IRIS's generalization and robustness to complex, unseen situations and different models architecture.

cs.CL↗

arXiv2025

Morphology-optimized Multi-Scale Fusion: Combining Local Artifacts and Mesoscopic Semantics for Deepfake Detection and Localization

While the pursuit of higher accuracy in deepfake detection remains a central goal, there is an increasing demand for precise localization of manipulated regions. Despite the remarkable progress made in classification-based detection, accurately localizing forged areas remains a significant challenge. A common strategy is to incorporate forged region annotations during model training alongside manipulated images. However, such approaches often neglect the complementary nature of local detail and global semantic context, resulting in suboptimal localization performance. Moreover, an often-overlooked aspect is the fusion strategy between local and global predictions. Naively combining the outputs from both branches can amplify noise and errors, thereby undermining the effectiveness of the localization. To address these issues, we propose a novel approach that independently predicts manipulated regions using both local and global perspectives. We employ morphological operations to fuse the outputs, effectively suppressing noise while enhancing spatial coherence. Extensive experiments reveal the effectiveness of each module in improving the accuracy and robustness of forgery localization.

cs.CV↗

arXiv2024

RB-SQL: A Retrieval-based LLM Framework for Text-to-SQL

Large language models (LLMs) with in-context learning have significantly improved the performance of text-to-SQL task. Previous works generally focus on using exclusive SQL generation prompt to improve the LLMs' reasoning ability. However, they are mostly hard to handle large databases with numerous tables and columns, and usually ignore the significance of pre-processing database and extracting valuable information for more efficient prompt engineering. Based on above analysis, we propose RB-SQL, a novel retrieval-based LLM framework for in-context prompt engineering, which consists of three modules that retrieve concise tables and columns as schema, and targeted examples for in-context learning. Experiment results demonstrate that our model achieves better performance than several competitive baselines on public datasets BIRD and Spider.

cs.CL↗

arXiv2024

Towards Robustness and Diversity: Continual Learning in Dialog Generation with Text-Mixup and Batch Nuclear-Norm Maximization

In our dynamic world where data arrives in a continuous stream, continual learning enables us to incrementally add new tasks/domains without the need to retrain from scratch. A major challenge in continual learning of language model is catastrophic forgetting, the tendency of models to forget knowledge from previously trained tasks/domains when training on new ones. This paper studies dialog generation under the continual learning setting. We propose a novel method that 1) uses \textit{Text-Mixup} as data augmentation to avoid model overfitting on replay memory and 2) leverages Batch-Nuclear Norm Maximization (BNNM) to alleviate the problem of mode collapse. Experiments on a $37$-domain task-oriented dialog dataset and DailyDialog (a $10$-domain chitchat dataset) demonstrate that our proposed approach outperforms the state-of-the-art in continual learning.

cs.CL↗

arXiv2016

A Multi Perspective Approach for Understanding the Determinants of Cloud Computing Adoption among Australian SMEs

Cloud computing is proved to be an effective computing technology for organisations through the advantages that it offers such as cost-effectiveness, IT technical agility and scalability, enhancing businesses processes, and increasing enterprises competitiveness. In Australia, there is an emerging trend that small and medium-sized enterprises (SMEs) begin to adopt this technology in the conventional working practices. However, there is a dearth of prior studies on examining the factors that influence the cloud computing adoption among Australian SMEs. To fill the empirical vacuum, this research-in-progress proposes an integrated framework for examining the determinants of cloud computing service adoption with the consideration of the unique characteristics of Australian SMEs, such as relatively low adoption of cloud computing services, less innovative, and limited knowledge about cloud computing and its benefits and hindrances. To this end, we are conducting consecutive studies to investigate this research issue. An exploratory interview study will be undertaken to identify and verify the unique characteristics of Australian SMEs toward the cloud computing adoption. This is followed by an organisational level survey that examines the effects of those determinants on cloud computing adoption. Finally, a decision model for cloud computing adoption among Australian SMEs will be developed by using a Multi Criteria Decision Approach (MCDA) through rating, prioritising, and ranking of various criteria and alternatives available to the decision makers. Adopting the mixed-method research fashion, this research-in-progress intends to make significant implications to scholars and practitioners alike in the cloud computing research and applications areas.

cs.CY↗