Does AI Assistance Leave a Temporal Fingerprint? Detecting Overreliance in AI-Assisted Writing and Programming
研究通过分析写作和编程过程数据，使用时间特征检测AI辅助的过度依赖问题，区分正常协作与完全委托，以维护学术诚信。
0 citationsRead paper
Selected work
Representative Papers
An 84-Format Numeric Catalog with Bit-Exact Conformance Vectors: A Vendor-Neutral Reference for FP8, BF16, MXFP4, and Microscaling Formats
This work addresses the proliferation of numeric formats in machine learning hardware, which introduces elusive silent discrepancies during cross-platform model migration, necessitating a vendor-neutral, bit-level precise reference standard. The authors construct a comprehensive catalog encompassing 84 numeric formats and introduce six bit-level consistency suites. They propose a novel verification anchor based on the φ=3.0 identity, combined with SHA-256 fingerprints and unified row patterns, to achieve semantic alignment across platforms for formats such as FP8, BF16, and MXFP4, while explicitly documenting permissible deviations from the ml_dtypes specification. All resources are packaged in JSON and integrated with IEEE P3109 v3.2.0 standard mappings, and have been open-sourced on GitHub to support reproducibility and auditability of cross-platform numeric behavior.
0 citationsRead paper
Cyber-Physical Anomaly Detection in IoT-Enabled Smart Grids Using Machine Learning and Metaheuristic Feature Optimization
This study addresses the challenge of distinguishing between physical faults and false data injection attacks in IoT-enabled smart grids, which often manifest as indistinguishable anomalies. To tackle this issue, the authors propose a lightweight detection framework that integrates genetic algorithms with tree-based ensemble models—specifically Extra Trees, XGBoost, and Random Forest—to enable metaheuristic-driven feature selection for dimensionality reduction of PMU/IED measurements. Evaluated on the MSU/ORNL power system attack dataset, the proposed GA+Extra Trees approach reduces the feature dimensionality from 112 to an average of 27.4 while achieving a macro-F1 score of 0.9212 and a ROC-AUC of 0.9837. The results demonstrate that the method not only significantly mitigates feature redundancy but also enhances both detection performance and model interpretability.
0 citationsRead paper
AI Misuse in Education Is a Measurement Problem: Toward a Learning Visibility Framework
This work addresses the risks posed by the misuse of artificial intelligence in education, which obscures the visibility of the learning process and threatens academic integrity, equity, and cognitive development. The authors propose a “learning visibility” framework that reconceptualizes AI misuse as a measurement challenge rather than a detection problem. Integrating cognitive offloading theory, learning analytics, and multimodal timeline reconstruction techniques, the framework establishes an assessment system centered on process transparency, normative AI use, and shared evidentiary practices. Moving beyond the limitations of conventional AI-detection tools, this approach offers educators a principled pathway for AI integration that upholds educational values, fosters trust, and enhances transparency, thereby effectively mitigating the “black box” effect induced by AI-mediated learning environments.
0 citationsRead paper
Allee Synaptic Plasticity and Memory
Existing neural plasticity models suffer from high sensitivity to noise and unbounded synaptic weight growth. To address these issues, this work proposes a biologically inspired nonlinear synaptic plasticity model grounded in the Allee effect, incorporating a time-dependent dynamic threshold mechanism that couples eligibility traces with oscillatory inputs. This design enables self-limiting synaptic updates and robust memory storage/retrieval under noisy conditions. The model employs biologically constrained nonlinear dynamics, balancing stability with temporal adaptability. Experiments demonstrate substantial improvements over classical Hebbian and Oja models: memory capacity and retrieval reliability increase significantly, classification accuracy rises by 12.6% in dynamic noise environments, and the risk of synaptic weight divergence decreases by 83%. Collectively, the proposed framework establishes a novel paradigm for brain-inspired memory systems that simultaneously satisfies biological plausibility and engineering robustness.
0 citationsRead paper
Recent publications
Latest Papers
Does AI Assistance Leave a Temporal Fingerprint? Detecting Overreliance in AI-Assisted Writing and Programming
研究通过分析写作和编程过程数据，使用时间特征检测AI辅助的过度依赖问题，区分正常协作与完全委托，以维护学术诚信。
0 citationsRead paper
An 84-Format Numeric Catalog with Bit-Exact Conformance Vectors: A Vendor-Neutral Reference for FP8, BF16, MXFP4, and Microscaling Formats
This work addresses the proliferation of numeric formats in machine learning hardware, which introduces elusive silent discrepancies during cross-platform model migration, necessitating a vendor-neutral, bit-level precise reference standard. The authors construct a comprehensive catalog encompassing 84 numeric formats and introduce six bit-level consistency suites. They propose a novel verification anchor based on the φ=3.0 identity, combined with SHA-256 fingerprints and unified row patterns, to achieve semantic alignment across platforms for formats such as FP8, BF16, and MXFP4, while explicitly documenting permissible deviations from the ml_dtypes specification. All resources are packaged in JSON and integrated with IEEE P3109 v3.2.0 standard mappings, and have been open-sourced on GitHub to support reproducibility and auditability of cross-platform numeric behavior.
0 citationsRead paper
Cyber-Physical Anomaly Detection in IoT-Enabled Smart Grids Using Machine Learning and Metaheuristic Feature Optimization
This study addresses the challenge of distinguishing between physical faults and false data injection attacks in IoT-enabled smart grids, which often manifest as indistinguishable anomalies. To tackle this issue, the authors propose a lightweight detection framework that integrates genetic algorithms with tree-based ensemble models—specifically Extra Trees, XGBoost, and Random Forest—to enable metaheuristic-driven feature selection for dimensionality reduction of PMU/IED measurements. Evaluated on the MSU/ORNL power system attack dataset, the proposed GA+Extra Trees approach reduces the feature dimensionality from 112 to an average of 27.4 while achieving a macro-F1 score of 0.9212 and a ROC-AUC of 0.9837. The results demonstrate that the method not only significantly mitigates feature redundancy but also enhances both detection performance and model interpretability.
0 citationsRead paper
AI Misuse in Education Is a Measurement Problem: Toward a Learning Visibility Framework
This work addresses the risks posed by the misuse of artificial intelligence in education, which obscures the visibility of the learning process and threatens academic integrity, equity, and cognitive development. The authors propose a “learning visibility” framework that reconceptualizes AI misuse as a measurement challenge rather than a detection problem. Integrating cognitive offloading theory, learning analytics, and multimodal timeline reconstruction techniques, the framework establishes an assessment system centered on process transparency, normative AI use, and shared evidentiary practices. Moving beyond the limitations of conventional AI-detection tools, this approach offers educators a principled pathway for AI integration that upholds educational values, fosters trust, and enhances transparency, thereby effectively mitigating the “black box” effect induced by AI-mediated learning environments.
0 citationsRead paper
Allee Synaptic Plasticity and Memory
Existing neural plasticity models suffer from high sensitivity to noise and unbounded synaptic weight growth. To address these issues, this work proposes a biologically inspired nonlinear synaptic plasticity model grounded in the Allee effect, incorporating a time-dependent dynamic threshold mechanism that couples eligibility traces with oscillatory inputs. This design enables self-limiting synaptic updates and robust memory storage/retrieval under noisy conditions. The model employs biologically constrained nonlinear dynamics, balancing stability with temporal adaptability. Experiments demonstrate substantial improvements over classical Hebbian and Oja models: memory capacity and retrieval reliability increase significantly, classification accuracy rises by 12.6% in dynamic noise environments, and the risk of synaptic weight divergence decreases by 83%. Collectively, the proposed framework establishes a novel paradigm for brain-inspired memory systems that simultaneously satisfies biological plausibility and engineering robustness.
0 citationsRead paper