Acknowledgments
We would like to thank the International Mathematical Olympiad Organization for their support.
Thang Luong led the overall technical direction of the advanced Gemini model with IMO’s detailed thinking and co-led with Edward Lockhart on the overall coordination of IMO 2025 efforts.
The IMO 2025 system was not possible without the following technical leads: Dawsen Hwang, co-led training data and expert evaluations from Junyuk Jung. Jonathan Lee, Nate Kushman, Pol Moreno and Yi Tay co-led training on advanced Gemini Deep Think models while Lei Yu was leading the model evaluation. Golnaz Ghiazi, Garrett Bingham and Lalit Jain co-led with Dawsen Hwang and Dawsen Hwang of Vincent Cohen-Addad, co-led with strengthening the inference approach.
The IMO 2025 system was developed with a significant contribution from Ankesh Anand’s Theophane Weber for modelling. Inferences from Vinay Ramasesh, Andreas Kirsch, Jieming Mao, Zicheng Xu, Wilfried Bounsi and Vahab Mirrokni. Training data for Hoang Nguyen, Fred Zhang, Mahan Malihi and Yangsibo Huang.
We appreciate the contribution and efforts from relevant teams. Alpha Sexual Metric Team of Yuri Chervonyi (Lead), Trieu Trinh, Hoang Nguyen, Junsu Kim, Mirek Olšák, Marcelo Menegali and Xiaomeng Yang. Miklós Z. For the formal mathematics of Horváth, Aja Huang, Goran Palace. Fabian Pedregosa, Richard Song, Alex Zai, Sarah Javanmaldi, Yaguan Lee, Filipe Miguel de Almeida, Silvio Ratanzi, Ashcan Norowzi Fard, Tar Schuster, Honlu Fan, King Xuezhi, Adity Mavalancal, Tom Scharcal, Komara, Verjour, Barabari.
In particular, we thank the other core members of the Deep Think team (Archit Sharma, Tong He, Shubha Raghvendra), post-training efforts (Tianhe Kevin Yu, Siamak Shakeri, Hanzhao Lin, Cosmo Du, Sid Lall), and research in the area where the IMO 2025 system appears to have been built.
This effort was advised by Cooc Le and Pushmeet Kohli with programmatic support from Kristen Chiafullo and Alex Goldin.
We would also like to thank the experts for providing the data and evaluation: Insuk Seo (Lead), Jiwon Kang, Donghyun Kim, Junsu Kim, Jimin Kim, Seongbin Jeon, Yoonho Na, Seunghwan Lee, Jihoo Lee, Yunghun Jo, Yongsuk Moon, Seojin Kim, Yuuen Lee, Taehoon Kim, Jiho Liu, Seungumo Lee, Daein Kim, Sangha Lee, Hyun Woo Choi, Aiden Jung, Youngbom Jin, Yonghyun An, Junwee Bae, Gin Kim, Nam Dan Trang, Kari Qian Tang Ran, Kari Thun Tan Yahya Tabesh, Hamed Mahdavi, Azin Nazari, Xiangzhuo Ding, Chu-Lan Kao, Steven Creech, Tony Feng, and Ciprian Manolescu.
Emmanuel Talopa, Charlie Chen, Joe Stanton, Ship Betu, Alvin Abdasik, Federico LeBron, Iona Mihailescu, Soheil Hassas Yegane, and Min Gang.
We would like to thank Jessica Law and Sajad Zafar for their support in providing and managing calculations. Jane Labanowski, Andy Forbes, Sean Nakamoto for legal and logistics. Omar Levi, Timothy Lyriclap, Jack Ray, Ifen Lu, Hen Tsen, Ed Chi, Vahab Mirokuni, Tulsi Dosi, Madavi Swarm, Melvin Johnson, Korey Kabuk Ugur, Oriol Vinval, Jeff Dean, Demis Hassavis, Sergie Briin.
Finally, we would like to thank Professor Gregor Drinner of the IMO Committee for their support and approval.
The IMO has confirmed that the answers submitted are the complete and correct solution. It is important to note that their reviews have not been extended to validating systems, processes, or underlying models (see details).
