Selected work

Representative Papers

Class-Specific Branch Attention for Mitigating Gradient Interference under Class Imbalance

Jun 04, 2026

This work addresses the suppression of minority-class learning in deep neural networks under severe class imbalance, which arises from shared feature representations. To mitigate this issue, the authors propose a Class-Specific Branch Attention (CSBA) mechanism. By analyzing inter-layer gradient flows and constructing a gradient conflict matrix based on class-specific gradient cosine similarity, they reveal—through the lens of optimization dynamics—for the first time how majority classes dominate and suppress gradients of minority classes. A lightweight channel reweighting module is integrated into a multi-branch convolutional architecture to implicitly decouple features and gradients in a class-aware manner. Experiments demonstrate that the proposed method significantly improves minority-class performance without compromising overall accuracy: the F1 score for the Physical-Damage class increases from 0.261 to 0.522, and Macro-F1 on CIFAR-10-LT rises from 0.595 to 0.655.

0 citationsRead paper

Multi-Stage Training for Abusive Comment Detection in Indic Languages

May 21, 2026

This study addresses the high false positive rates in detecting abusive comments in Indic languages on social media platforms, which often risk infringing on freedom of expression. To tackle this challenge, the authors propose a novel framework that integrates language-specific preprocessing, multi-model ensembling, and multi-stage training, carefully tailored to the linguistic characteristics of Indic languages. The approach significantly improves the accuracy of abusive content detection while markedly reducing misclassification of non-abusive texts. Experimental results across multiple Indic-language datasets demonstrate that the proposed framework effectively balances detection performance with false positive control, offering a new paradigm for multilingual content moderation that harmonizes accuracy with the preservation of free speech.

0 citationsRead paper

SG-CoT: An Ambiguity-Aware Robotic Planning Framework using Scene Graph Representations

Mar 18, 2026

This work addresses the challenge of planning failures in robotic systems caused by ambiguous natural language instructions when leveraging large language models (LLMs). To mitigate this issue, the authors propose SG-CoT, a novel framework that uniquely integrates scene graphs with chain-of-thought (CoT) reasoning. SG-CoT constructs a structured scene graph representation of the environment and guides the LLM to iteratively retrieve relevant information from the graph, enabling it to detect ambiguities and proactively generate clarification questions. This approach synergistically combines scene graph construction, LLM-based retrieval, CoT reasoning, and multi-agent interaction mechanisms. Empirical results demonstrate that SG-CoT improves task success rates by at least 4% in single-agent settings and by 15% in multi-agent environments, while also achieving over a 10% increase in ambiguity identification accuracy.

0 citationsRead paper

Melody or Machine: Detecting Synthetic Music with Dual-Stream Contrastive Learning

Nov 29, 2025

The rapid advancement of end-to-end AI music generation poses severe threats to artistic authenticity and copyright protection, while existing detection methods suffer from poor generalization to out-of-distribution (OOD) synthetic content. To address this, we propose CLAM, a dual-stream contrastive learning architecture that— for the first time—exploits subtle acoustic inconsistencies between vocal and instrumental representations to identify AI-generated music. CLAM incorporates a learnable cross-aggregation module and jointly optimizes binary cross-entropy loss with contrastive triplet loss. It employs parallel audio encoders—MERT and Wav2Vec 2.0—to extract complementary speech- and instrument-oriented acoustic features. Evaluated on MoM, a newly constructed large-scale, diverse benchmark comprising 130K tracks, CLAM achieves an F1 score of 0.925, substantially outperforming state-of-the-art methods. This demonstrates superior robustness and generalization capability for detecting AI-generated music across heterogeneous distributions.

0 citationsRead paper

AI Regulation in Telecommunications: A Cross-Jurisdictional Legal Study

Nov 27, 2025

This paper identifies a critical regulatory lag and fragmentation arising from the deep integration of AI into essential digital infrastructure—particularly telecommunications—where existing telecommunications, cybersecurity, and data protection laws largely overlook AI-specific risks such as model drift, algorithmic bias, and decision opacity; notably, AI systems in telecom remain virtually unregulated. Employing a comparative legal methodology, the study systematically analyzes policy documents from ten jurisdictions, synthesizing insights across AI governance, telecommunications law, and data protection theory. Its core contribution is the identification of “institutional coordination failure” as a fundamental governance gap, and the proposal of the first cross-sectoral, forward-looking regulatory framework tailored to AI-enabled digital infrastructure. This framework shifts governance from reactive compliance toward proactive, systemic integration, offering both theoretical foundations and actionable institutional design pathways for building unified, adaptive AI infrastructure regulation.

0 citationsRead paper

Recent publications

Latest Papers

Class-Specific Branch Attention for Mitigating Gradient Interference under Class Imbalance

Jun 04, 2026

This work addresses the suppression of minority-class learning in deep neural networks under severe class imbalance, which arises from shared feature representations. To mitigate this issue, the authors propose a Class-Specific Branch Attention (CSBA) mechanism. By analyzing inter-layer gradient flows and constructing a gradient conflict matrix based on class-specific gradient cosine similarity, they reveal—through the lens of optimization dynamics—for the first time how majority classes dominate and suppress gradients of minority classes. A lightweight channel reweighting module is integrated into a multi-branch convolutional architecture to implicitly decouple features and gradients in a class-aware manner. Experiments demonstrate that the proposed method significantly improves minority-class performance without compromising overall accuracy: the F1 score for the Physical-Damage class increases from 0.261 to 0.522, and Macro-F1 on CIFAR-10-LT rises from 0.595 to 0.655.

0 citationsRead paper

Multi-Stage Training for Abusive Comment Detection in Indic Languages

May 21, 2026

This study addresses the high false positive rates in detecting abusive comments in Indic languages on social media platforms, which often risk infringing on freedom of expression. To tackle this challenge, the authors propose a novel framework that integrates language-specific preprocessing, multi-model ensembling, and multi-stage training, carefully tailored to the linguistic characteristics of Indic languages. The approach significantly improves the accuracy of abusive content detection while markedly reducing misclassification of non-abusive texts. Experimental results across multiple Indic-language datasets demonstrate that the proposed framework effectively balances detection performance with false positive control, offering a new paradigm for multilingual content moderation that harmonizes accuracy with the preservation of free speech.

0 citationsRead paper

SG-CoT: An Ambiguity-Aware Robotic Planning Framework using Scene Graph Representations

Mar 18, 2026

This work addresses the challenge of planning failures in robotic systems caused by ambiguous natural language instructions when leveraging large language models (LLMs). To mitigate this issue, the authors propose SG-CoT, a novel framework that uniquely integrates scene graphs with chain-of-thought (CoT) reasoning. SG-CoT constructs a structured scene graph representation of the environment and guides the LLM to iteratively retrieve relevant information from the graph, enabling it to detect ambiguities and proactively generate clarification questions. This approach synergistically combines scene graph construction, LLM-based retrieval, CoT reasoning, and multi-agent interaction mechanisms. Empirical results demonstrate that SG-CoT improves task success rates by at least 4% in single-agent settings and by 15% in multi-agent environments, while also achieving over a 10% increase in ambiguity identification accuracy.

0 citationsRead paper

Melody or Machine: Detecting Synthetic Music with Dual-Stream Contrastive Learning

Nov 29, 2025

The rapid advancement of end-to-end AI music generation poses severe threats to artistic authenticity and copyright protection, while existing detection methods suffer from poor generalization to out-of-distribution (OOD) synthetic content. To address this, we propose CLAM, a dual-stream contrastive learning architecture that— for the first time—exploits subtle acoustic inconsistencies between vocal and instrumental representations to identify AI-generated music. CLAM incorporates a learnable cross-aggregation module and jointly optimizes binary cross-entropy loss with contrastive triplet loss. It employs parallel audio encoders—MERT and Wav2Vec 2.0—to extract complementary speech- and instrument-oriented acoustic features. Evaluated on MoM, a newly constructed large-scale, diverse benchmark comprising 130K tracks, CLAM achieves an F1 score of 0.925, substantially outperforming state-of-the-art methods. This demonstrates superior robustness and generalization capability for detecting AI-generated music across heterogeneous distributions.

0 citationsRead paper

AI Regulation in Telecommunications: A Cross-Jurisdictional Legal Study

Nov 27, 2025

This paper identifies a critical regulatory lag and fragmentation arising from the deep integration of AI into essential digital infrastructure—particularly telecommunications—where existing telecommunications, cybersecurity, and data protection laws largely overlook AI-specific risks such as model drift, algorithmic bias, and decision opacity; notably, AI systems in telecom remain virtually unregulated. Employing a comparative legal methodology, the study systematically analyzes policy documents from ten jurisdictions, synthesizing insights across AI governance, telecommunications law, and data protection theory. Its core contribution is the identification of “institutional coordination failure” as a fundamental governance gap, and the proposal of the first cross-sectoral, forward-looking regulatory framework tailored to AI-enabled digital infrastructure. This framework shifts governance from reactive compliance toward proactive, systemic integration, offering both theoretical foundations and actionable institutional design pathways for building unified, adaptive AI infrastructure regulation.

0 citationsRead paper