Spatial function-on-function quantile regression
本文提出了一种新的空间函数对函数分位数回归框架，通过考虑曲线间的空间相关性来分析空间索引的功能数据，并采用两阶段工具变量估计策略处理内生性问题。
0 citationsRead paper
Selected work
Representative Papers
Supply Chain Propagation of Textual Signals: LLM Embeddings and Cross-Sectional Return Predictability
This study investigates how integrating textual information from corporate annual reports with supply chain network structure can enhance stock return predictability. The authors propose a novel approach that combines FinBERT-derived embeddings from 10-K filings with a supply chain knowledge graph, leveraging network signal propagation to construct an augmented predictive factor. Within a multifactor asset pricing framework, this network-enhanced factor demonstrates significant out-of-sample predictive power for cross-sectional returns (t = −2.64). A long–short portfolio based on the factor achieves an annualized Sharpe ratio of 0.86, and it generates a risk-adjusted alpha of 7.27% per year (t = 2.30) after controlling for the Fama–French five factors. The results are robust across specifications, highlighting that inter-firm linkages embedded in supply chain networks contain valuable information for uncovering mispricing in equity markets.
0 citationsRead paper
Digital Image Forgery Detection Using Transfer Learning
This study addresses the growing prevalence of digital image manipulation by proposing a hybrid input representation that integrates compression-based difference features (FDIFF) with RGB channels, leveraging a transfer learning framework to enhance forgery detection performance. Built upon pretrained CNN architectures—DenseNet121, ResNet50, and VGG16—the method incorporates a Youden index–driven adaptive thresholding strategy to optimally balance true positive and false positive rates. Experimental results on the CASIA v2.0 dataset demonstrate that the proposed approach significantly improves detection robustness and practicality: DenseNet121 achieves the highest accuracy and AUC, while ResNet50 yields the best Matthews Correlation Coefficient (MCC), collectively validating the efficacy of combining compression-aware features with adaptive threshold optimization.
0 citationsRead paper
Edge Reliability Gap in Vision-Language Models: Quantifying Failure Modes of Compressed VLMs Under Visual Corruption
This study investigates whether compressed vision-language models (VLMs) exhibit systematic failure modes beyond mere increases in error rates when deployed on edge devices. Through categorizing errors into object blindness, semantic drift, and prior bias, and leveraging confidence calibration (ECE), structured negation reasoning probes, controlled ambiguity experiments, and a GPT-4o-based discriminator, the work reveals for the first time that compact VLMs undergo significant and non-uniform qualitative degradation on benchmarks like COCO—most notably a collapse in negation reasoning capabilities. For instance, SmolVLM2-500M achieves a 100% error rate on the false_yn template, underperforming Qwen2.5-VL-7B by 12.5 percentage points. The study further introduces a reproducible safety auditing pipeline, establishing a new paradigm for evaluating the reliability of compressed VLMs.
0 citationsRead paper
Psychological Factors Influencing University Students Trust in AI-Based Learning Assistants
This study investigates the psychological mechanisms underlying university students’ trust formation in AI-powered learning assistants. Moving beyond technology-centric perspectives, it integrates cognitive appraisal, affective response, social relational dynamics, and contextual moderators into the first psychology-driven, four-dimensional dynamic trust framework. The framework conceptualizes trust as a malleable psychological process co-shaped by individual differences and educational contexts, thereby bridging theoretical psychology with educational AI design. Employing a narrative literature review, the study synthesizes empirical and theoretical insights from mental health, human–AI interaction, and automation trust literatures to derive testable hypotheses and key research questions. The findings provide evidence-informed, mechanism-based intervention pathways for educators, educational policymakers, and AI designers—advancing the principled development and implementation of trustworthy, pedagogically grounded AI systems in higher education. (149 words)
0 citationsRead paper
Recent publications
Latest Papers
Spatial function-on-function quantile regression
本文提出了一种新的空间函数对函数分位数回归框架，通过考虑曲线间的空间相关性来分析空间索引的功能数据，并采用两阶段工具变量估计策略处理内生性问题。
0 citationsRead paper
Supply Chain Propagation of Textual Signals: LLM Embeddings and Cross-Sectional Return Predictability
This study investigates how integrating textual information from corporate annual reports with supply chain network structure can enhance stock return predictability. The authors propose a novel approach that combines FinBERT-derived embeddings from 10-K filings with a supply chain knowledge graph, leveraging network signal propagation to construct an augmented predictive factor. Within a multifactor asset pricing framework, this network-enhanced factor demonstrates significant out-of-sample predictive power for cross-sectional returns (t = −2.64). A long–short portfolio based on the factor achieves an annualized Sharpe ratio of 0.86, and it generates a risk-adjusted alpha of 7.27% per year (t = 2.30) after controlling for the Fama–French five factors. The results are robust across specifications, highlighting that inter-firm linkages embedded in supply chain networks contain valuable information for uncovering mispricing in equity markets.
0 citationsRead paper
Digital Image Forgery Detection Using Transfer Learning
This study addresses the growing prevalence of digital image manipulation by proposing a hybrid input representation that integrates compression-based difference features (FDIFF) with RGB channels, leveraging a transfer learning framework to enhance forgery detection performance. Built upon pretrained CNN architectures—DenseNet121, ResNet50, and VGG16—the method incorporates a Youden index–driven adaptive thresholding strategy to optimally balance true positive and false positive rates. Experimental results on the CASIA v2.0 dataset demonstrate that the proposed approach significantly improves detection robustness and practicality: DenseNet121 achieves the highest accuracy and AUC, while ResNet50 yields the best Matthews Correlation Coefficient (MCC), collectively validating the efficacy of combining compression-aware features with adaptive threshold optimization.
0 citationsRead paper
Edge Reliability Gap in Vision-Language Models: Quantifying Failure Modes of Compressed VLMs Under Visual Corruption
This study investigates whether compressed vision-language models (VLMs) exhibit systematic failure modes beyond mere increases in error rates when deployed on edge devices. Through categorizing errors into object blindness, semantic drift, and prior bias, and leveraging confidence calibration (ECE), structured negation reasoning probes, controlled ambiguity experiments, and a GPT-4o-based discriminator, the work reveals for the first time that compact VLMs undergo significant and non-uniform qualitative degradation on benchmarks like COCO—most notably a collapse in negation reasoning capabilities. For instance, SmolVLM2-500M achieves a 100% error rate on the false_yn template, underperforming Qwen2.5-VL-7B by 12.5 percentage points. The study further introduces a reproducible safety auditing pipeline, establishing a new paradigm for evaluating the reliability of compressed VLMs.
0 citationsRead paper
Psychological Factors Influencing University Students Trust in AI-Based Learning Assistants
This study investigates the psychological mechanisms underlying university students’ trust formation in AI-powered learning assistants. Moving beyond technology-centric perspectives, it integrates cognitive appraisal, affective response, social relational dynamics, and contextual moderators into the first psychology-driven, four-dimensional dynamic trust framework. The framework conceptualizes trust as a malleable psychological process co-shaped by individual differences and educational contexts, thereby bridging theoretical psychology with educational AI design. Employing a narrative literature review, the study synthesizes empirical and theoretical insights from mental health, human–AI interaction, and automation trust literatures to derive testable hypotheses and key research questions. The findings provide evidence-informed, mechanism-based intervention pathways for educators, educational policymakers, and AI designers—advancing the principled development and implementation of trustworthy, pedagogically grounded AI systems in higher education. (149 words)
0 citationsRead paper