A highly motivated and aspiring researcher with a strong background in deep learning, with recent work centered on multimodal learning. Over the past few years, I have actively contributed to the field, publishing research articles in top-tier venues such as ACL and WACV. My primary research interests now lie in preference optimization and reinforcement learning for large language models. I have also worked on vision-language models, domain adaptation, and federated learning, and continue to explore these areas with strong interest.
Rafid Ahmed, Intesar Tahmid, Mir Sazzat Hossain, Tasnimul Hossain Tomal, Md Mahir Jawad, Anam Borhan Uddin, Md Fahim, Md Farhad Alam Bhuiyan
In Findings of the Association for Computational Linguistics: ACL 2026, pages 37362–37378, San Diego, California, United States. Association for Computational Linguistics.
Md Fahim, Md Farhan Ishmam, Mir Sazzat Hossain, M Ashraful Amin, Amin Ahsan Ali, AKM Rahman
Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2026, pp. 6496-6506.
Rafid Ahmed, Intesar Tahmid, Mir Sazzat Hossain, Tasnimul Hossain Tomal, Md Fahim, Md Farhad Alam Bhuiyan
Proceedings of The Second AAAI Bridge Program on AI for Medicine and Healthcare, PMLR 317:1-14, 2026.
Newaz Ben Alam, Akm Moshiur Rahman Mazumder, Mir Sazzat Hossain, Mysha Samiha, Md Alvi Noor Hossain, Md Fahim, Amin Ahsan Ali, Ashraful Islam, M Ashraful Amin, Akmmahbubur Rahman
In Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics, pages 2178–2194, Mumbai, India. The Asian Federation of Natural Language Processing and The Association for Computational Linguistics.
How GSPO fixes the instability of GRPO by moving the importance ratio from tokens to whole responses: one length-normalized sequence ratio, sequence-level clipping, and equal weights for every token, which stabilizes large Mixture-of-Experts training and powered Qwen3.
How CISPO, from MiniMax-M1, keeps every token in the gradient: instead of clipping token updates like PPO and GRPO, which silently drops rare reasoning tokens such as 'Wait' and 'However', it clips the importance sampling weight and lets every token keep learning.
