Kotaro Yoshida

CV (Last Updated: Sep. 2025)

I am a first-year master’s student at the Institute of Science Tokyo, specializing in deep learning. My primary interests are improving generalization in deep neural networks and developing model merging methods for flexible model editing. During my undergraduate studies, I received the Outstanding Student Award in recognition of my academic achievements and research contributions. I have also competed in multiple machine learning competitions—earning medals each time—which strengthened both my theoretical grounding and practical skills. My academic experience includes an exchange program at the National University of Singapore and research internships at A*STAR.

Contact me!

Feel free to contact me at!

Work

Publication

Google Scholor

journal

Hiroki Naganuma*, Ryuichiro Hataya*, Kotaro Yoshida, Ioannis Mitliagkas, “An Empirical Study of Pre-trained Model Selection for Out-of-Distribution Generalization and Calibration”, Transactions on Machine Learning Research (TMLR), 2025, [paper]

Kotaro Yoshida*, Hiroki Naganuma*,“Towards Understanding Variant Methods of Invariant Risk Minimization from the Perspective of Calibration”, Transactions on Machine Learning Research (TMLR), 2024, [paper]

International Conference

Kotaro Yoshida, Yuji Naraki, Yoshikazu Ikeda, Takafumi Horie, Ryosuke Yamaki, Ryotaro Shimizu, Yuki Saito, Julian McAuley, Hiroki Naganuma, “Mastering Task Arithmetic: $\tau$Jp as a Key Indicator for Weight Disentanglement”, The 13th International Conference on Learning Representations (ICLR), 2025, [paper]

International Workshop

Kotaro Yoshida, Yuji Naraki, Yoshikazu Ikeda, Takafumi Horie, Ryosuke Yamaki, Ryotaro Shimizu, Yuki Saito, Julian McAuley, Hiroki Naganuma, “Mastering Task Arithmetic: $\tau$Jp as a Key Indicator for Weight Disentanglement”, NeurIPS 2024 Workshop on Fine-Tuning in Modern Machine Learning: Principles and Scalability, 2024, [paper]

Hiroki Naganuma*, Kotaro Yoshida*, Laura Gomezjurado Gonzalez*, Takafumi Horie, Yuji Naraki, Ryotaro Shimizu, “On Fairness of Task Arithmetic: The Role of Task Vectors”, NeurIPS 2025 Workshop on Reliable ML from Unreliable Data, 2025, [paper].

Domestic Conference

Kotaro Yoshida, Takafumi Horie, Yuji Naraki, Ryotaro Shimizu, Ryosuke Yamaki, Hiroki Naganuma, “Evaluating the Effectiveness of Model Linearization in Task Analogies”, The 39th Annual Conference of the Japanese Society for Artificial Intelligence, 2025.

Kotaro Yoshida, Jin Takeshita, Hiroki Naganuma, “Towards Understanding Variant Methods of Invariant Risk Minimization from the Perspective of Calibration”,The 86th National Convention of IPSJ, 2024. (Student Encouragement Award)

Kotaro Yoshida, Keigo Tada, Hiroki Naganuma, “Uncertainty Calibration in Deep Neural Networks through Invariant Risk Minimization”, The 22nd Forum on Information Technology, 2023.

PREPRINT and UNDER REVIEW

Kotaro Yoshida, Atsushi Nitanda, “How Does Preconditioning Guide Feature Learning in Deep Neural Networks?”, 2025, [paper].

Kotaro Yoshida, Yuji Naraki, Takafumi Horie, Ryotaro Shimizu, Hiroki Naganuma, “DisTaC: Conditioning Task Vectors via Distillation for Robust Model Merging”, 2025, [paper].

Kotaro Yoshida, Konstantinos Slavakis, “Robust Invariant Representation Learning by Distribution Extrapolation”, 2025, [paper].

Yuji Naraki, Ryosuke Yamaki, Yoshikazu Ikeda, Takafumi Horie, Kotaro Yoshida, Ryotaro Shimizu, Hiroki Naganuma, “Augmenting NER Datasets with LLMs: Towards Automated and Refined Annotation“, 2024, [paper]

Awards

Undergraduate Outstanding Student Award, the Department of Information and Communications Engineering of the Institute of Science Tokyo (Mar. 2025)
Singapore International Pre‑Graduate Award (SIPGA), Agency for Science, Technology and Research (A*STAR) (Jan. 2025)