Hashmat Shadab Malik

About Me

I am currently a PhD Candidate at Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI) and before that have obtained my Masters degree in computer vision from the same institution. At MBZUAI, I am affliated with the Intelligent Visual Analytics Lab (IVAL) where I am primarily advised by Dr. Salman Khan and have a close collaboration with Dr. Muzammal Naseer. My secondary advisor is Dr. Fahad Khan.

I received my undergraduate degree in Electronics and Communication Engineering with a minor in Computer Science from the National Institute of Technology (NIT) Srinagar, India. After graduation, I worked as a researcher at Indian Institute of Science IISc, before transitioning into academia in 2020 and joining MBZUAI as a Masters Student under Dr. Salman Khan.

MBZUAI
(2021-Present)

NIT Srinagar
(2014-2018)

Research Interests

My research focuses on the Safety and Reliability of AI, with a particular emphasis on understanding, evaluating, and enhancing the robustness of vision-based models.

News

• [June. 2026] Our work HistoVL got accepted at MIUA 2026! It will be presented as an oral presentation + poster. 🎉
• [Dec. 2025] Our work FaceGuardian got accepted at SaTML 2026! It will be presented as an oral presentation + poster. 🎉
• [Aug. 2025] Our work Robust-LLaVA will be presented as an oral presentation in Trustworthy FMs Workshop at ICCV 2025! 🎉
• [Jun. 2025] Our work HSAT has been accepted at MICCAI 2025! 🎉
• [Mar. 2025] Our work titled “Towards Evaluating the Robustness of Visual State Space Models” got accepted at CVPRW 2025! 🎉
• [Mar. 2025] Our work HSAT, enhances robustness of histopathology vision models by leveraging patient-slide-patch relationships to construct heirarchy-wise attacks and integrate the generated adversarial examples in the model training. HSAT achieves a 54.31% improvement against white-box attacks and reduces performance drop to 3-4% against black-box attacks.
• [Feb. 2025] Our work Robust-LLaVA, enhances the adversarial robustness of Multi-modal Large Language Models (MLLMs) by leveraging large-scale robust classification models. It achieves 2× and 1.5× average robustness gains in captioning and VQA tasks, and improves resistance to jailbreaking attacks by over 10% compared to state-of-the-art methods.
• [Dec. 2024] Our work titled “ObjectCompose: Evaluating Resilience of Vision-Based Models on Object-to-Background Compositional Changes” secured Best Student Paper Runner Up Award at ACCV 2024! 🎉.
• [Sep. 2024] Our work titled “ObjectCompose: Evaluating Resilience of Vision-Based Models on Object-to-Background Compositional Changes” got accepted at ACCV 2024 for Oral Presentation (Top 5%)! 🎉.
• [Jun. 2024] Our paper titled “Evaluating Robustness of Volumetric Medical Segmentation Models” got accepted at BMVC 2024! 🎉
• [Jan. 2023] I joined MBZUAI as a PhD student in Computer Vision with full scholarship.
• [Sep. 2022] Our work titled “Adversarial Pixel Restoration as a Pretext Task for Transferable Perturbations” got accepted at BMVC 2022 for Oral Presentation (Top 9%)! 🎉.
• [Jan. 2021] I joined MBZUAI as a Masters student in Computer Vision with full scholarship.

Publications

Academic Services

Conference Reviewing

• Served as a Reviewer at CVPR’25, BMVC'2025, ECCV’24, BMVC’24, WACV’24

Teaching Assistant

• Advanced Computer Vision (CV801) course - Fall 2024 with Prof. Hisham Cholakkal
• Human and Computer Vision (CV701) course - Fall 2023 with Prof. Mohammad Yaqub

Awards and Merits

• Full scholarship for PhD in Computer Vision at the Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI).
• Secured Best Student Paper Runner Up Award at ACCV 2024.
• Full scholarship for Masters in Computer Vision at the Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI).
• Secured 140th state rank in IIT-JEE Mains 2014 (among top 2% of 1.5 million students across the country).

Powered by Jekyll and Minimal Light theme.