About Me

I am Andrew Gambardella (Japanese name: Akira Takanami), postdoctoral researcher at the University of Tokyo, under the supervision of Professor Yutaka Matsuo.

Prior to joining UTokyo I spent a year as a postdoctoral researcher at KAIST.

I obtained my DPhil (PhD) in Engineering Science at Oxford, advised by Professor Phil Torr.

During my undergrad years I studied EECS at UC Berkeley, where my undergraduate research was advised by Professor Satish Rao.

I have lived in the USA, Japan, Korea, and the UK. I speak fluent English and Japanese and less-fluent Korean and Chinese.

My full CV can be found here.

A bio for media coverage and similar purposes can be found here.

Research

My research interests include:

• Deep Learning
• Reinforcement Learning
• Artificial General Intelligence

Publications

Journal Articles

• Real-World Robot Applications of Foundation Models: A Review. Kento Kawaharazuka, Tatsuya Matsushima, Andrew Gambardella, Jiaxian Guo, Chris Paxton, Andy Zeng. Advanced Robotics, 2024. [arXiv]
• gOd, mOther and sOldier: A Story of Oppression, Told Through the Lens of AI. Andrew Gambardella, Meeyung Chung, Doyo Choi, Jinjoon Lee. To Appear in Leonardo, 2023. [MIT Press Direct]

Conference Proceedings

• Which Programming Language and What Features at Pre-training Stage Affect Downstream Logical Inference Performance?. Fumiya Uchiyama, Takeshi Kojima, Andrew Gambardella, Qi Cao, Yusuke Iwasawa, Yutaka Matsuo. In The 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2024.
• Language Models Do Hard Arithmetic Tasks Easily and Hardly Do Easy Arithmetic Tasks. Andrew Gambardella, Yusuke Iwasawa, Yutaka Matsuo. ACL, 2024. [ACL Proceedings] [arXiv]
• Multitask Soft Option Learning. Maximilian Igl, Andrew Gambardella, Nantas Nardelli, N. Siddharth, Wendelin Böhmer, Shimon Whiteson. In Uncertainty in Artificial Intelligence, 2020. [UAI Proceedings] [arXiv]

Workshops

• Detecting and Quantifying Malicious Activity with Simulation-based Inference. Andrew Gambardella, Bogdan State, Naeemullah Khan, Leo Tsourides, Philip H. S. Torr, Atılım Güneş Baydin. In International Conference on Machine Learning: Workshop on Socially Responsible Machine Learning, 2021. [arXiv]
• Hierarchical Semantic Sonification for the Visually Impaired. Andrew Gambardella, Jinjoon Lee. In IJCAI: Workshop on Artificial Intelligence for Social Good, 2021. [AI4SG-21]
• Simulation-Based Inference for Global Health Decisions. Christian Schroeder de Witt, Bradley Gram-Hansen, Nantas Nardelli, Andrew Gambardella, Rob Zinkov, Puneet Dokania, N. Siddharth, Ana Belen Espinosa-Gonzalez, Ara Darzi, Philip Torr, Atılım Güneş Baydin. In International Conference on Machine Learning Workshop: ML for Global Health, 2020. [arXiv]

Preprints

• Inconsistent Tokenizations Cause Language Models to be Perplexed by Japanese Grammar. Andrew Gambardella, Takeshi Kojima, Yusuke Iwasawa, Yutaka Matsuo. In submission, 2024.
• Crypto-LLM: Two-Stage Language Model Pre-training with Ciphered and Natural Language Data. Yohei Kobashi, Takeshi Kojima, Andrew Gambardella, Qi Cao, Yusuke Iwasawa, Yutaka Matsuo. In submission, 2024.
• Answer When Needed, Forget When Not: Language Models Pretend to Forget via In-Context Knowledge Unlearning. Shota Takashiro, Takeshi Kojima, Andrew Gambardella, Qi Cao, Yusuke Iwasawa, Yutaka Matsuo. In submission, 2024.
• Inside Head: Uncertainty Quantification for LLMs based on Internal Discrepancy. Qi Cao, Andrew Gambardella, Takeshi Kojima, Yutaka Matsuo, Yusuke Iwasawa. In submission, 2024.
• Efficient Data Mosaicing with Simulation-based Inference. Andrew Gambardella, Youngjun Choi, Doyo Choi, Jinjoon Lee. ArXiv Preprint, 2023. [arXiv]
• Transflow Learning: Few-shot Adaptation for Normalizing Flow Models. Andrew Gambardella, Atılım Güneş Baydin, Philip H. S. Torr. ArXiv Preprint, 2019. [arXiv]

Thesis

• Deep Transfer Learning with Bayesian Inference. Andrew Gambardella. DPhil Thesis, 2021. [Oxford University Research Archive]

Invited Talks

• Achieving Human-centric AI through Research. At “The Responsibility of Humanity with Regard to AI in the Future”, 2024. [Hosted by 社団法人Privacy by Design Lab]

Contact

You can reach me at my UTokyo email address and GitHub.

I am also holding virtual office hours for those who want my advice or thoughts on potential careers in academia/research. Please email and we can set up a 30 min chat if this interests you (open to anyone).