Our work
Filter by year to explore research from the group on edge intelligence, collaborative inference and quantum-guided models.
NLLP Workshop @ EMNLP 2026
Tasmiad Hasan · Arafat Zaman Ratul · Sarker Sadman Saalim · S.M. Shah Nawaz Hossain · Khan Raiyan Ibne Reza · Sumaiya Tabassum Nimi
A human-verified benchmark examining whether multimodal LLMs can read and reason over Bengali land records, from handwritten symbols and ownership fractions to legal document questions.
arXiv preprint
Khan Raiyan Ibne Reza · Sumaiya Tabassum Nimi · Omar Ibne Shahid
KrishokChat is a citation-grounded Bengali agricultural benchmark with traceable sources, real farmer queries, and explicit evaluation of safety in chemical advisory.
Quantum Machine Intelligence
Shahreen Sultana · Mutasim Fuad Sarker · Md Faiyaz Bin Younus · Md Adnan Arefeen · Sumaiya Tabassum Nimi
Explores quantum-state geometry and a fidelity-based scoring strategy to detect out-of-distribution inputs in hybrid classical-quantum neural network classifiers.
arXiv preprint
Mutasim Fuad Sarker · Adiba Rahman Namira · Wafa Binte Alam · Md Adnan Arefeen · Mahzabeen Emu · Sumaiya Tabassum Nimi
Combines agentic AI-guided clinical feature selection with a compact quantum-enhanced time-series model for in-hospital cardiac arrest mortality prediction on MIMIC-IV.
arXiv preprint
Khan Raiyan Ibne Reza · Sanjana Aktar Maria · Mohammad Tushar Abdullah · Asfee Bhuiyan Leen · Sumaiya Tabassum Nimi
TRACE-BN transfers structured Bangla-English tutoring behavior, including grammar explanations and learner-error diagnosis, to a compact offline language model for resource-constrained deployment.
arXiv preprint
Khan Raiyan Ibne Reza · Sanjana Aktar Maria · Sumaiya Tabassum Nimi
Evaluates RAG architectures for Bengali agricultural advisory, revealing how retrieval performance varies sharply across colloquial farmer queries, formal terminology, and language conditions.
IEEE SMARTCOMP 2023
Nimi, S. T. · Arefeen, A. · Uddin, Y. S. · Debnath, B. · Chakradhar, S.
A context-driven collaborative vision transformer framework that enables efficient edge intelligence through faction-based model partitioning and collaborative inference.
IEEE SMARTCOMP 2022
Nimi, S. T. · Arefeen, A. · Uddin, Y. S. · Debnath, B. · Chakradhar, S.
Chimera introduces context-aware splittable deep multitasking models that balance accuracy and resource usage for edge intelligence applications.
IEEE SMARTCOMP 2022
Nimi, S. T.
A study of deep learning model composition strategies that enable multiple inference tasks to share computation efficiently on resource-constrained edge devices.
ECML-PKDD 2021
Nimi, S. T. · Arefeen, A. · Uddin, Y. S. · Lee, Y.
EARLIN performs early out-of-distribution detection to make collaborative inference between edge devices and the cloud more resource-efficient.
Full texts of our publications: Papers folder ↗