Research Interests

I’m broadly interested in natural language processing and deep learning with a focus on developing Trustworthy AI. I am passionate about exploring and building reliable AI based systems. Due to the rapid advancements in Large Language Models (LLMs), their application has become ubiquitous across all domains. Consequently, if the generated content lacks controllability, identifiability, and reliability, it can pose potential harm. I am passionate to investigate on below domains,

Human-AI Interaction and Collaboration :

Building intelligent , interactive and reliable NLP systems capable of producing factually accurate outputs to assist users in responding to their requests. These systems should be able to achieve this by understanding user queries, offering reasoning for each generated response, and supporting the information with authentic sources.

NLP for Social Good :

Combining NLP techniques for interdisciplinary research, including healthcare, social science and mental health.

Trustworthy and Responsible NLP :

T Identifying and addressing misinformation, racial attacks, and bias in contents of social media. Creating solutions that enable the model to interpret underlying factors of detection.

Reasoning and problem solving :

LLMs should be able to reason their generation. It should have the capability to request additional information from users in order to better understand their preferences and be equipped to fulfill their requests effectively. Based on that it should gradually be able to generate coherent responses . Thus the hallucination in LLMs can be reduced to a great extent.

Linguistic aspects of Prompt :

Can LLMs understand the linguistic properties of the prompt? Can this information help the LLM in generating coherent response.

About me

Hi there! I am S.M Towhidul Islam, an Electrical & Electronic Engineering Graduate student from Islamic University Of Technology, Gazipur, Bangladesh.

I am a proud graduate of 2022 with a CGPA of 3.68. Presently, I am working as an Applied ML Researcher at Robi Axiata Limited, Bangladesh. Additionally, I am contributing as a Research Affiliate at The Artificial Intelligence Institute of UofSC (AIISC) under the guidance of Professor Dr. Amitava Das.

My recent publications at ACL2023 & EMNLP 2023 main are a testament to my passion and dedication towards research. My research interest lies in fact verification, reducing the hallucination of LLMs and focusing on trustworthy AI. I am constantly exploring new ideas and opportunities to contribute to the field of NLP.

I am open to collaborate with any research opportunity related to my area of interest. If you have any queries, please feel free to reach out to me via email.

Undergraduate Projects, Journals & Papers

News Updates

• October 11, 2023 - Happy to share, my shared task paper in The “First Bangla Language Processing (BLP) Workshop @EMNLP 2023” also got accepted.

• October 7, 2023 - I am happy to share that two of my co-author papers got accepted in EMNLP Main 2023.

1. The Troubling Emergence of Hallucination in Large Language Models - An Extensive Definition, Quantification, and Prescriptive Remediations

2. Counter Turing Test (CT^2): Al-Generated Text Detection is Not as Easy as You May Think - Introducing Al Detectability Index (ADI)

• August 21, 2023 - I have stood #4th and #9th in the shared tasks of the first Bangla Language Processing (BLP) Workshop that will be co-located with EMNLP on 7th December, 2023 in Singapore.

• May 2, 2023 - I have one paper , “FACTIFY-5WQA: 5W Aspect-based Fact Verification through Question Answering” accepted at ACL2023. Huge thanks and congratulations to all collaborators!

• March 5, 2023 - My team has ranked 5th in the first round of ‘Apurba presents Bhashabhrom: EEE DAY 2023 Datathon’ Kaggle Competition where we excelled in detecting grammatical errors in Bengali sentences using a sequence labeling approach.

• December 20, 2022 - Our paper titled “OOG - Optuna Optimized GAN Sampling Technique for Tabular Imbalanced Malware Data” was presented today at the 2022 IEEE International Conference on Big Data in Osaka, Japan.

• November 9, 2022 - Our team, IUT_Reset has achieved the first place in IEEE Big Data Cup 2022 and our paper has been accepted for presentation in the conference.

Skills

• Programming Languages: Python, SQL, R
• Big Data Technologies: Hive, Hue, Cloudera , AWS
• Database Management: Oracle
• Data Visualization: Tableau, Power BI
• Machine Learning: NLP, State-of-the-art Models
• Web Development: Streamlit, Gradio