Selected work

Representative Papers

Latent Stability Analysis of Malware Representations Under Feature-Space Perturbations

Jul 27, 2026

This study addresses the lack of systematic evaluation of latent-space stability and decision boundary sensitivity in existing static malware detectors under feature perturbations. The authors propose a latent-space stability analysis framework that integrates adversarial feature perturbations with multiple representation schemes—including raw EMBER features, PCA-compressed embeddings, variational autoencoders, Mandelbrot escape-time descriptors, and a PINN-based latent flow module. To characterize dynamic evolution under perturbation, they introduce a novel metric, Latent Escape Divergence (LED), alongside PINNFlow-derived residuals, velocity, risk, and gradient shift measures. Experimental results show that while the composite representations do not significantly improve classification performance—achieving ROC AUCs of 0.9962 on clean EMBER samples and 0.9846 with PCA-64 compression—they offer distinctive analytical value for perturbation diagnostics.

0 citationsRead paper

A Multi-Agent Framework for Medical AI: Leveraging Fine-Tuned GPT, LLaMA, and DeepSeek R1 for Evidence-Based and Bias-Aware Clinical Query Processing

Feb 15, 2026

This work addresses the challenges of weak validation, insufficient evidence, and unreliable confidence in clinical large language models for medical question answering. To enhance reliability and safety, the authors propose a modular multi-agent framework that orchestrates specialized agents for clinical reasoning, evidence retrieval, and answer refinement. The system integrates uncertainty quantification via Monte Carlo Dropout and perplexity estimation, interpretability analysis using LIME and SHAP, and a lexical-sentiment bias detection mechanism. Built upon fine-tuned GPT, LLaMA, and DeepSeek R1 models with PubMed-augmented evidence, the framework achieves an accuracy of 87%, a relevance score of 0.80, and a reduced perplexity of 4.13—all within an end-to-end latency of 36.5 seconds—significantly outperforming the BioGPT baseline.

0 citationsRead paper

Recent publications

Latest Papers

Latent Stability Analysis of Malware Representations Under Feature-Space Perturbations

Jul 27, 2026

This study addresses the lack of systematic evaluation of latent-space stability and decision boundary sensitivity in existing static malware detectors under feature perturbations. The authors propose a latent-space stability analysis framework that integrates adversarial feature perturbations with multiple representation schemes—including raw EMBER features, PCA-compressed embeddings, variational autoencoders, Mandelbrot escape-time descriptors, and a PINN-based latent flow module. To characterize dynamic evolution under perturbation, they introduce a novel metric, Latent Escape Divergence (LED), alongside PINNFlow-derived residuals, velocity, risk, and gradient shift measures. Experimental results show that while the composite representations do not significantly improve classification performance—achieving ROC AUCs of 0.9962 on clean EMBER samples and 0.9846 with PCA-64 compression—they offer distinctive analytical value for perturbation diagnostics.

0 citationsRead paper

A Multi-Agent Framework for Medical AI: Leveraging Fine-Tuned GPT, LLaMA, and DeepSeek R1 for Evidence-Based and Bias-Aware Clinical Query Processing

Feb 15, 2026

This work addresses the challenges of weak validation, insufficient evidence, and unreliable confidence in clinical large language models for medical question answering. To enhance reliability and safety, the authors propose a modular multi-agent framework that orchestrates specialized agents for clinical reasoning, evidence retrieval, and answer refinement. The system integrates uncertainty quantification via Monte Carlo Dropout and perplexity estimation, interpretability analysis using LIME and SHAP, and a lexical-sentiment bias detection mechanism. Built upon fine-tuned GPT, LLaMA, and DeepSeek R1 models with PubMed-augmented evidence, the framework achieves an accuracy of 87%, a relevance score of 0.80, and a reduced perplexity of 4.13—all within an end-to-end latency of 36.5 seconds—significantly outperforming the BioGPT baseline.

0 citationsRead paper