Koushik Ahamed Tonmoy

Dhaka, Bangladesh

koushik.tonmoy [at] northsouth [dot] edu

Hi! I am an undergraduate AI researcher and Computer Science and Engineering student at North South University. I study when language and vision-language models genuinely reason and when they rely on shortcuts, superficial correlations, or memorized patterns. My work focuses on benchmark development and rigorous model evaluation across multimodal reasoning, culturally grounded NLP, trustworthy AI, and formal computational reasoning.

My recent research includes TimeSpot (ICML 2026), BengaliMoralBench (ACM FAccT 2026), an evaluation of executable Turing-machine synthesis by LLMs (Findings of EMNLP 2026), and a review of geo-spatial and geo-temporal reasoning (AACL-IJCNLP 2026 Main Conference). I have also worked on transformer-based source-code vulnerability detection.

I enjoy collaborating across institutions to turn broad questions about model intelligence into measurable and reproducible evaluations. Alongside research, I serve as a Marker at North South University, supporting academic assessment and examination.

news

Sep 08, 2026	Our review paper, Geo-spatial and Geo-temporal Reasoning in Vision-Language and Large Language Models: A Review, has been accepted to the Main conference at AACL-IJCNLP 2026!
Aug 21, 2026	Our paper, From Language Specifications to Executable Turing Machines: Evaluating LLMs as Computational Machine Designers, has been accepted to Findings of EMNLP 2026!
Jul 02, 2026	The Qatar Computing Research Institute (QCRI) highlighted our collaboration and ICML 2026 paper, TimeSpot, in a LinkedIn feature.
Jun 16, 2026	One paper has been accepted to the AI for Math Workshop (AI4Math) at ICML 2026!
Jun 11, 2026	One paper has been accepted to the Muslims in ML Workshop (MusiML) at ICML 2026!
Apr 30, 2026	*TimeSpot* has been accepted as a regular paper at ICML 2026!
Apr 16, 2026	*BengaliMoralBench* has been accepted to ACM FAccT 2026!

selected publications

C

BengaliMoralBench: A Benchmark for Auditing Moral Reasoning in Large Language Models within Bengali Language and Culture

SZ Ridoy^*, AT Wasi^*, Koushik Ahamed Tonmoy, TH Rafi, and DK Chae

In FAccT ’26: Proceedings of the 2026 ACM Conference on Fairness, Accountability, and Transparency, 2026

DOI HTML PDF Website
C

TimeSpot: Benchmarking Geo-Temporal Understanding in Vision–Language Models in Real-World Settings

AT Wasi^*, SZ Ridoy^*, Koushik Ahamed Tonmoy, K Tshering, SM Hasan, W Faisal, T Mohiuddin, and MR Parvez

In Proceedings of the 43rd International Conference on Machine Learning (ICML), 2026

HTML PDF Website
C

From Language Specifications to Executable Turing Machines: Evaluating LLMs as Computational Machine Designers

SZ Ridoy, SM Hasan, AT Wasi, K Tshering, Koushik Ahamed Tonmoy, and MM Khan

In Findings of the Association for Computational Linguistics: EMNLP 2026, 2026

DOI HTML PDF Website
C

Geo-Spatial and Geo-Temporal Reasoning in Vision–Language and Large Language Models: A Review

SZ Ridoy^*, AT Wasi^*, Koushik Ahamed Tonmoy, and MR Parvez

In Proceedings of the 5th Asia-Pacific Chapter of the Association for Computational Linguistics and the 15th International Joint Conference on Natural Language Processing (AACL-IJCNLP), 2026