I am a Ph.D. Candidate in Computer Science at Florida State University, advised by Dr. Xiuwen Liu. My research focuses on the robustness and explainability of large multimodal models — studying how microscopic input perturbations cascade through neural representations to cause failures in vision-language understanding, embodied navigation, and reasoning.
I design mechanisms to detect and mitigate these vulnerabilities in Vision Transformers (ViTs), vision-language models (VLMs), and large language models (LLMs). My work has been published at IROS 2024, ISVC 2025, and WCCI-IJCNN 2026, and in Information Sciences (2026). Our paper Neural Expansion: A Unified Mechanism for How Deep Neural Networks Generalize has been accepted at NeurIPS 2026.
In Summer 2026, I interned at Chewy as a Machine Learning Engineer Intern II, working on LLM personalization and evaluation for an AI shopping assistant. In Summer 2025, I interned at Meta, working on vision features for large-scale feed ranking. I also hold an M.Sc. in Computer Science from FSU (GPA: 3.97/4.00) and a B.Sc. in CSE from Bangladesh University of Engineering and Technology (BUET).
Chashi Mahiul Islam, Samuel Jacob Chacko, Mao Nishino, Canlin Zhang, Xiuwen Liu
Conference on Neural Information Processing Systems (NeurIPS 2026) · Main Track 2026 Accepted
We proposed Neural Expansion, a geometric account of generalization based on growing directions along which predictions remain unchanged. Growing regions of stable predictions offer a common geometric explanation for several learning behaviors; the mechanism is supported by empirical evidence and partial theory.
Chashi Mahiul Islam, Samuel Jacob Chacko, Mao Nishino, Canlin Zhang, Xiuwen Liu
Conference on Neural Information Processing Systems (NeurIPS 2026) · Main Track 2026 Accepted
We proposed Neural Expansion, a geometric account of generalization based on growing directions along which predictions remain unchanged. Growing regions of stable predictions offer a common geometric explanation for several learning behaviors; the mechanism is supported by empirical evidence and partial theory.
Chashi Mahiul Islam, Alan Villarreal, Mao Nishino, Shaeke Salman, Xiuwen Liu
IEEE World Congress on Computational Intelligence (WCCI 2026), IJCNN 2026 Oral
We discover that microscopic floating-point input perturbations cause up to 10-million-fold amplification through LLM layers, collapsing logit margins and triggering unpredictable token flips — revealing a fundamental numerical chaos regime in large language models. We quantify regimes of stability and rounding-induced chaos to characterize when LLM predictions remain stable or become unpredictable.
Chashi Mahiul Islam, Alan Villarreal, Mao Nishino, Shaeke Salman, Xiuwen Liu
IEEE World Congress on Computational Intelligence (WCCI 2026), IJCNN 2026 Oral
We discover that microscopic floating-point input perturbations cause up to 10-million-fold amplification through LLM layers, collapsing logit margins and triggering unpredictable token flips — revealing a fundamental numerical chaos regime in large language models. We quantify regimes of stability and rounding-induced chaos to characterize when LLM predictions remain stable or become unpredictable.
Chashi Mahiul Islam, Samuel Jacob Chacko, Mao Nishino, Xiuwen Liu
International Symposium on Visual Computing (ISVC 2025) 2025 Oral
We examine how small input changes affect layer-wise representations in Vision Transformers (ViTs) and propose NeuroShield-ViT, which neutralizes vulnerable neurons without retraining. PGD-100 accuracy improves from 15.6% to 72.3%; classification accuracy also improves by 71.6% against IGO attacks on ImageNet1K.
Chashi Mahiul Islam, Samuel Jacob Chacko, Mao Nishino, Xiuwen Liu
International Symposium on Visual Computing (ISVC 2025) 2025 Oral
We examine how small input changes affect layer-wise representations in Vision Transformers (ViTs) and propose NeuroShield-ViT, which neutralizes vulnerable neurons without retraining. PGD-100 accuracy improves from 15.6% to 72.3%; classification accuracy also improves by 71.6% against IGO attacks on ImageNet1K.
Chashi Mahiul Islam*, Shaeke Salman*, Montasir Shams, Xiuwen Liu, Piyush Kumar (* equal contribution)
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024) 2024 Oral
We develop adversarial algorithms that exploit semantic representation gaps in vision-language navigation models to redirect robotic agents along malicious paths, achieving 91.7% landmark-conditioned path redirection and proposing a noise-sensitivity-based detection mechanism that improves navigation policy stability by 96.3%.
Chashi Mahiul Islam*, Shaeke Salman*, Montasir Shams, Xiuwen Liu, Piyush Kumar (* equal contribution)
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024) 2024 Oral
We develop adversarial algorithms that exploit semantic representation gaps in vision-language navigation models to redirect robotic agents along malicious paths, achieving 91.7% landmark-conditioned path redirection and proposing a noise-sensitivity-based detection mechanism that improves navigation policy stability by 96.3%.