The 2025 Google Code Golf Championship

Michael D. Moffitt, Divy Thakkar, Ryan Burnell, Orhan Firat, Walter Reade, Sohier Dane, Addison Howard, Lucas Deutschmann, Luke Gustafson, Adam Kues, Mukundan Senthilkumar, Nicolas Posada, Lyndon Ho, Larry Wu, Lav Patel, Hunter Richardson, Wei Heng Kuan
Proceedings of the NeurIPS 2023-2025 Competition Track, PMLR 276:123-173, 2026.

Abstract

For over half a decade, the Abstraction and Reasoning Corpus has played a critical role in evaluating the fluid intelligence and skill acquisition efficiency of machine learning systems. While the dataset has proven especially useful in advancing program synthesis, few investments have been made in distilling the transformational logic of these tasks into their most irreducible computational forms. In conjunction with the NeurIPS 2025 Competition Track, we organized the 2025 Google Code Golf Championship, a three-month online challenge designed to elicit minimal Python solutions for all training puzzles in the seminal ARC-AGI-1 benchmark suite. Sponsored by Google DeepMind and funded with $100,000 in prizes, the contest attracted over 1,000 teams and yielded tens of thousands of verified, open-source programs that collectively establish tight upper bounds on the Kolmogorov complexity of each task. In this paper, we offer a comprehensive retrospective of the event, including an overview of the contest format, a survey of approaches taken by top teams, and a discussion of potential implications for the future design of intelligent agents.

Cite this Paper


BibTeX
@InProceedings{pmlr-v276-moffitt26a, title = {The 2025 Google Code Golf Championship}, author = {Moffitt, Michael D. and Thakkar, Divy and Burnell, Ryan and Firat, Orhan and Reade, Walter and Dane, Sohier and Howard, Addison and Deutschmann, Lucas and Gustafson, Luke and Kues, Adam and Senthilkumar, Mukundan and Posada, Nicolas and Ho, Lyndon and Wu, Larry and Patel, Lav and Richardson, Hunter and Kuan, Wei Heng}, booktitle = {Proceedings of the NeurIPS 2023-2025 Competition Track}, pages = {123--173}, year = {2026}, editor = {Qin, Tao and Zhang, Kun and Frellsen, Jes}, volume = {276}, series = {Proceedings of Machine Learning Research}, month = {02--07 Dec}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v276/main/assets/moffitt26a/moffitt26a.pdf}, url = {https://proceedings.mlr.press/v276/moffitt26a.html}, abstract = {For over half a decade, the Abstraction and Reasoning Corpus has played a critical role in evaluating the fluid intelligence and skill acquisition efficiency of machine learning systems. While the dataset has proven especially useful in advancing program synthesis, few investments have been made in distilling the transformational logic of these tasks into their most irreducible computational forms. In conjunction with the NeurIPS 2025 Competition Track, we organized the 2025 Google Code Golf Championship, a three-month online challenge designed to elicit minimal Python solutions for all training puzzles in the seminal ARC-AGI-1 benchmark suite. Sponsored by Google DeepMind and funded with $100,000 in prizes, the contest attracted over 1,000 teams and yielded tens of thousands of verified, open-source programs that collectively establish tight upper bounds on the Kolmogorov complexity of each task. In this paper, we offer a comprehensive retrospective of the event, including an overview of the contest format, a survey of approaches taken by top teams, and a discussion of potential implications for the future design of intelligent agents.} }
Endnote
%0 Conference Paper %T The 2025 Google Code Golf Championship %A Michael D. Moffitt %A Divy Thakkar %A Ryan Burnell %A Orhan Firat %A Walter Reade %A Sohier Dane %A Addison Howard %A Lucas Deutschmann %A Luke Gustafson %A Adam Kues %A Mukundan Senthilkumar %A Nicolas Posada %A Lyndon Ho %A Larry Wu %A Lav Patel %A Hunter Richardson %A Wei Heng Kuan %B Proceedings of the NeurIPS 2023-2025 Competition Track %C Proceedings of Machine Learning Research %D 2026 %E Tao Qin %E Kun Zhang %E Jes Frellsen %F pmlr-v276-moffitt26a %I PMLR %P 123--173 %U https://proceedings.mlr.press/v276/moffitt26a.html %V 276 %X For over half a decade, the Abstraction and Reasoning Corpus has played a critical role in evaluating the fluid intelligence and skill acquisition efficiency of machine learning systems. While the dataset has proven especially useful in advancing program synthesis, few investments have been made in distilling the transformational logic of these tasks into their most irreducible computational forms. In conjunction with the NeurIPS 2025 Competition Track, we organized the 2025 Google Code Golf Championship, a three-month online challenge designed to elicit minimal Python solutions for all training puzzles in the seminal ARC-AGI-1 benchmark suite. Sponsored by Google DeepMind and funded with $100,000 in prizes, the contest attracted over 1,000 teams and yielded tens of thousands of verified, open-source programs that collectively establish tight upper bounds on the Kolmogorov complexity of each task. In this paper, we offer a comprehensive retrospective of the event, including an overview of the contest format, a survey of approaches taken by top teams, and a discussion of potential implications for the future design of intelligent agents.
APA
Moffitt, M.D., Thakkar, D., Burnell, R., Firat, O., Reade, W., Dane, S., Howard, A., Deutschmann, L., Gustafson, L., Kues, A., Senthilkumar, M., Posada, N., Ho, L., Wu, L., Patel, L., Richardson, H. & Kuan, W.H.. (2026). The 2025 Google Code Golf Championship. Proceedings of the NeurIPS 2023-2025 Competition Track, in Proceedings of Machine Learning Research 276:123-173 Available from https://proceedings.mlr.press/v276/moffitt26a.html.

Related Material