I am a Ph.D. Candidate in Computer Science at Hanyang University, working in the Data Intelligence Lab with Prof. Dong-Kyu Chae. I am also co-advised by Prof. Junegak Joung from the Department of Industrial Data Engineering (IDE), Hanyang University. My Ph.D. is supported by Brain Korea 21 and National Research Foundation of Korea (NRF). Previously, my Ph.D. was partially supported by Samsung Electronics. I was a research collaborator with Oracle, USA. Before joining Hanyang, I obtained a bachelor's degree in Electrical Engineering from AUST in 2021.

I am interested in building adaptive and trustworthy machine learning systems that remain reliable under real-world distribution shifts. My work focuses on enabling models to generalize beyond static training settings and operate effectively in dynamic, open-world environments.
Test-Time Adaptation: How models can adapt at inference time without access to source data, particularly under long-tailed, open-set, and continually shifting distributions. I am interested in designing confidence-aware adaptation, memory-driven learning, and energy-based objectives that improve robustness while maintaining stability. My work also explores pseudo-label reliability, out-of-distribution detection, and instance-level adaptation mechanisms.
Multi-Context Evaluation: I investigate bias, robustness, and safety, particularly in multilingual and culturally diverse settings. I am interested in building evaluation benchmarks and datasets that capture real-world complexity, including commonsense reasoning, fairness, and cultural alignment.
Privacy-Preserving Systems: How to handle heterogeneous data, client imbalance, and unreliable participation in large-scale distributed settings. My goal is to develop adaptive aggregation, trust-aware optimization, and personalized learning strategies that scale to real-world deployments.
Graph-Based Learning: I am additionally interested in graph-based learning and structured representation learning, especially for problems involving relational reasoning and complex dependencies. My work explores how graph representations can improve generalization in domains such as knowledge modeling and interaction prediction.

Contact: takihr@hanyang.ac.kr (Official) | takihasanrafi@gmail.com (Permanent)
Office: 709, R&D Building-I, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 04763

Selected Papers

Full List is here!

	Continual Test-Time Adaptation: A Comprehensive Survey Sarthak Kumar Maharana, Shambhavi Mishra, Yunbei Zhang, Shuaicheng Niu, Taki Hasan Rafi, Jihun Hamm, Marco Pedersoli, Jose Dolz, Yunhui Guo ArXiv 2026b Test-Time [paper]
	Family Matters: A Systematic Study of Spatial vs. Frequency Masking for Continual Test-Time Adaptation Chandler Timm C. Doloriel, Yunbei Zhang, Yeonguk Yu, Taki Hasan Rafi, Muhammad Salman Siddiqui, Tor Kristian Stevik, Habib Ullah, Fadi Al Machot, Kristian Hovde Liland ArXiv 2026a Test-Time [paper]
	Learning from Unknown for Open-Set Test-Time Adaptation Taki Hasan Rafi, Amit Agarwal, Hitesh L. Patel, Dong-Kyu Chae 🏆 WACV 2026 (Oral, 85/2550 ~ A/R: 3.3%) Test-Time [paper]
	Towards Robust Continual Test-Time Adaptation via Neighbor Filtration Taki Hasan Rafi, Amit Agarwal, Hitesh L. Patel, Dong-Kyu Chae CIKM 2025 (185/604 ~ A/R: 30.6%) Test-Time [paper]
	Crowdsource, Crawl, or Generate? Creating SEA-VL, a Multicultural Vision-Language Dataset for Southeast Asia Samuel Cahyawijaya et al. (with Taki Hasan Rafi) 🏆 ACL 2025 (Main, Oral) Evaluation [paper]
	MVTamperBench: Evaluating Robustness of Vision-Language Models Amit Agarwal et al. (with Taki Hasan Rafi) ACL 2025 (Findings) Evaluation [paper]
	SweEval: Do LLMs Really Swear? A Safety Benchmark for Testing Limits for Enterprise Use Hitesh Laxmichand Patel et al. (with Taki Hasan Rafi) NAACL 2025 (Industry) Evaluation [paper]
	Instance-Aware Test-Time Adaptation for Domain Generalization Taki Hasan Rafi, Karlo Serbetar, Amit Agarwal, Hitesh L. Patel, Bhargava Kumar, Dong-Kyu Chae DASFAA 2025 (236/731 ~ A/R: 32.3%) Test-Time [paper]