Publications
Journal Papers
Biotwinmine: Digital Twin-Based Optimization of Biomining for Sustainable Rare Earth Element Production
SSRN | Status: Under Review
Authors: Mahfuz Ahmed Anik, Abdur Rahman, Md Iqramul Hoque, Azmine Toushik Wasi, et al.
Proposes a Digital Twin-Driven framework that integrates real-time process monitoring and predictive simulations to enhance biomining efficiency and sustainability for Rare Earth Elements.
A Theoretical Framework for Graph-based Digital Twins for Supply Chain Management
ArXiv | Status: In Review
Authors: Azmine Toushik Wasi, Mahfuz Ahmed Anik, Abdur Rahman, Md Iqramul Hoque, et al.
Combines graph modeling with Digital Twin architecture to create a dynamic representation of supply networks. Integrates sustainability metrics (carbon footprints) into operational dashboards.
Springback Prediction in V-Bending: An XAI-Enhanced Framework
Status: Ongoing Research
Authors: Md. Iqramul Hoque, Mahfuz Ahmed Anik, Engr Mohammed Abdul Karim
Conducted V-bending experiments on Aluminum, Mild Steel, and Copper. Used Polynomial Ridge Regression (R² 0.8373) and XAI tools (LIME, SHAP) to analyze process parameter influence on springback.
A Review on Prescriptive Analytics with Machine Learning
Status: Ongoing Research
Authors: Md. Iqramul Hoque, Abdur Rahman, Wahid Faisal, et al.
A comprehensive review of prescriptive analytics methodologies leveraging machine learning techniques.
Conference Papers
A Hybrid Approach to Climate Prediction: Physics-Informed Neural Networks for Accurate Temperature Forecasting
IEEE ICCIT 2025
Authors: Md. Iqramul Hoque, Mahfuz Ahmed Anik, Azmine Toushik Wasi, Dr. Abul Mukid Md. Mukaddes
Proposed a hybrid PINN model embedding physical constraints (seasonal cycles) to predict daily temperatures in Bangladesh. Achieved R² of 0.87, outperforming traditional ML models.
Physics-Informed Neural Networks for Clinical Time-Series Forecasting with Clinical Utility
IEEE ICCIT 2025
Authors: Azmine Toushik Wasi, Md. Iqramul Hoque, Mahfuz Ahmed Anik, Dr. Abul Mukid Md. Mukaddes
Implemented a PINN integrating physiological constraints (fluid balance) to forecast patient deterioration. The model enhances interpretability and consistency with medical principles.
Workshop Papers
Position: Adjacent Technologies Are the Key Enablers of Scalable and Safe Clinical MLLM Deployment
NeurIPS 2025 Workshop
Authors: Azmine Toushik Wasi, Md Iqramul Hoque
Argues that the clinical impact of MLLMs depends on an ecosystem of enabling technologies (data lakes, monitoring, API infrastructure). Emphasizes the need for strategic investment in "adjacent technologies" for scalable deployment.
CIOL at SemEval-2025 Task 11: Multilingual Pre-trained Model Fusion for Text-based Emotion Recognition
ACL’25W | SemEval-2025
Authors: Md. Iqramul Hoque, Mahfuz Ahmed Anik, Abdur Rahman, Azmine Toushik Wasi
Addresses challenges in multilingual emotion detection by leveraging language-specific transformer models for multi-label classification and intensity prediction. Achieved strong performance in Russian (0.848 F1).
CIOL at CLPsych 2025: Using Large Language Models for Understanding and Summarizing Clinical Texts
NAACL’25W | CLPsych 2025
Authors: Md Iqramul Hoque, Mahfuz Ahmed Anik, Azmine Toushik Wasi
Proposed a framework for evidence extraction, well-being scoring, and summary generation using LLMs. Achieved high consistency scores in summary generation (0.801 timeline-level).
Akatsuki-CIOL@ DravidianLangTech 2025: Ensemble-Based Approach for Fake News Detection
NAACL’25W | DravidianLangTech
Authors: Mahfuz Ahmed Anik, Md Iqramul Hoque, Wahid Faisal, Azmine Toushik Wasi, Md Manjurul Ahsan
Developed a fine-tuned transformer model for fake news detection in Malayalam. The binary classifier achieved a macro F1 score of 0.814, ranking 14th in the shared task.
Who (or What) is Responsible? Moral Agency and Accountability in Hybrid Creation
NeurIPS 2025 | CreativeAI Workshop
Authors: Azmine Toushik Wasi, Md. Iqramul Hoque, Mahfuz Ahmed Anik
Examines moral agency in human-AI co-creation. Proposes a distributed responsibility framework to align hybrid AI ecosystems with human values and accountability.
Position: Without Integrated Infrastructure, Clinical MLLMs Will Remain Technically Impressive but Clinically Marginal
ACM MM MCHM Workshop
Authors: Azmine Toushik Wasi, Md. Iqramul Hoque, Mahfuz Ahmed Anik
Highlights that MLLMs require high-fidelity data pipelines and secure API infrastructures to be clinically viable.
A Theoretical Framework for Governing Adaptive Generative AI Systems in Safety-Critical Industries
TAAS | Adaptive GenAI Governance
Authors: Azmine Toushik Wasi, Md. Iqramul Hoque, Mahfuz Ahmed Anik
Introduces a "regulate-to-learn" framework for governing adaptive GenAI, utilizing dynamic certification and regulatory sandboxes for high-stakes domains.