Publications

2026

1. ECV’26

   Firebolt-VL: Efficient Vision-Language Understanding with Cross-Modality Modulation

   Quoc-Huy Trinh, Mustapha Abdullahi (co-first author) , Bo Zhao, and 1 more author

   In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, Jun 2026

   Bib HTML PDF

   @inproceedings{Trinh_2026_CVPR,
     author = {Trinh, Quoc-Huy and Abdullahi, Mustapha and Zhao, Bo and Jha, Debesh},
     title = {Firebolt-VL: Efficient Vision-Language Understanding with Cross-Modality Modulation},
     booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
     journal = {9th Workshop on Efficient Deep Learning for Computer Vision (ECV) at CVPR},
     month = jun,
     year = {2026},
     pages = {3654-3663},
   }

2. PMLR (DLI’25)

   Evaluating Deep Learning Models for African Wildlife Image Classification: From DenseNet to Vision Transformers

   Lukman Jibril Aliyu, Umar Sani Muhammad, Bilqisu Ismail, and 5 more authors

   Proceedings of Machine Learning Research, Feb 2026

   Oral 🎤 Bib HTML PDF

   Oral presentation at Deep Learning Indaba 2025, selected as one of the top 7 out of 19 papers accepted from over 100 submissions

   @article{aliyu2025evaluating,
     title = {Evaluating Deep Learning Models for African Wildlife Image Classification: From DenseNet to Vision Transformers},
     author = {Aliyu, Lukman Jibril and Muhammad, Umar Sani and Ismail, Bilqisu and Muhammad, Nasiru and Wakili, Almustapha A and Yimam, Seid Muhie and Muhammad, Shamsuddeen Hassan and Abdullahi, Mustapha},
     journal = {Proceedings of Machine Learning Research},
     volume = {302},
     booktitle = {Deep Learning Indaba 2025 Research Track},
     month = feb,
     year = {2026},
   }

2025

1. SAA’25

   EARL: Efficient Agentic Reinforcement Learning Systems for Large Language Models

   Zheyue Tan, Mustapha Abdullahi , Tuo Shi, and 5 more authors

   Workshop on Systems for Agentic AI (SAA), in conjunction with the ACM Symposium on Operating Systems Principles (SOSP), Oct 2025

   Bib HTML

   @article{tan2025earl,
     title = {EARL: Efficient Agentic Reinforcement Learning Systems for Large Language Models},
     author = {Tan, Zheyue and Abdullahi, Mustapha and Shi, Tuo and Yuan, Huining and Xu, Zelai and Yu, Chao and Li, Boxun and Zhao, Bo},
     journal = {Workshop on Systems for Agentic AI (SAA), in conjunction with the ACM Symposium on Operating Systems Principles (SOSP)},
     month = oct,
     year = {2025},
     address = {Seoul, Korea},
   }

2024

1. JAIMLD’24

   Beyond traditional assessment: Exploring the impact of large language models on grading practices

   Oluwole Fagbohun, Nwaamaka Pearl Iduwe, Mustapha Abdullahi , and 2 more authors

   Journal of Artificial Intelligence, Machine Learning and Data Science, Feb 2024

   Bib HTML

   @article{fagbohun2024beyond,
     title = {Beyond traditional assessment: Exploring the impact of large language models on grading practices},
     author = {Fagbohun, Oluwole and Iduwe, Nwaamaka Pearl and Abdullahi, Mustapha and Ifaturoti, Adeseye and Nwanna, OM},
     journal = {Journal of Artificial Intelligence, Machine Learning and Data Science},
     volume = {2},
     number = {1},
     pages = {1--8},
     month = feb,
     year = {2024},
   }

2021

1. ABUCOMP’21

   An Artificial Intelligence of Things (AIOT) based Assistive Smart Glass for the Visually Impaired [POSTER]

   Mustapha Abdullahi

   Undergraduate Project Exhibition, Department of Computer Engineering, Ahmadu Bello University, Zaria., Oct 2021

   Best Project 🏆 Abs Bib PDF

   2021 best undergraduate project in the department of Computer Engineering, ABU, Zaria

   The blind and visually impaired constantly face challenges interacting with the society and performing common tasks. This project aims to develop a functional and cost-effective assistive device to aid the blind and visually impaired improve their quality of life as well as independence. The AIOT Smart Glass (SG) consists of a camera module to capture images and feeds them to the raspberry pi for analysis such as screen description, object recognition, reading printed text, etc. The AIOT smart glass is controlled easily using the Respeaker 2-Mics Pi Hat for voice commands or button taps which is directly integrated with a conversational chatbot for seamless usage. It has an additional feature of facial recognition; assisting users to recognize the faces of their families and other close members. The output of the system is passed as audio response through an earphone to the user. Results shows that the AIOT smart glass take an average of about 5 secs to responds to query in the presence of good internet availability.

   @article{mustapha_iglass,
     title = {An Artificial Intelligence of Things (AIOT) based Assistive Smart Glass for the Visually Impaired [POSTER]},
     author = {Abdullahi, Mustapha},
     journal = {Undergraduate Project Exhibition, Department of Computer Engineering, Ahmadu Bello University, Zaria.},
     numpages = {1},
     year = {2021},
     month = oct,
     url = {},
     dimensions = {true},
   }