Last Edited : Aug 24, 2023

The AutoML Cup is a competition that will challenge participants to design AutoML methods that perform well on a diverse set of tasks and data settings. It is a direct follow-up to the AutoML Decathlon 2022, comprising diverse, real-world problems whose domains are far afield from standard machine learning testbeds such as vision and language.

While the 2022 edition of the AutoML Decathlon competition explored the intersection of AutoML and data-centric AI along the axis of task diversity, this year’s AutoML Cup goes a step further by challenging competitors to cope with additional real-world data issues. With the key goal of maximizing scientific progress, the competition will feature several phases, similar to the impactful first AutoML challenge by ChaLearn. We will mandate participants to open-source their solutions after each phase in order to claim prizes, enabling participants to bootstrap from each others’ solutions. The phases will vary in data dimensions and resource settings (e.g., limited access to labeled training data).


Winning teams will be awarded exciting hardware!!!

To be eligible for prizes, you must make a submission for each of the three competition phases and make your submission open source at the conclusion of each phase. After the competition, the top ranked teams will also be featured in a retrospective paper that dives into the results and key takeaways of the competition.

To avoid conflicts of interest, all competition organizers are ineligible for prizes.

Important Dates

The AutoML Cup will include three phases.

Phase 1: June 6st — July 10th, 2023 September 10th, 2023

Phase 2: July 1st — August 5th, 2023 July, 15th — September 10th, 2023

Phase 3: August 10th — September 10th, 2023

  • The third and final phase will build on the first two phases but will additionally require handling situations in which access to labeled data is limited.
  • Make your phase 3 submission here!


  • Spencer Schoenberg* (technical lead) - University of Wisconsin-Madison
  • Nicholas Roberts* (organizing lead) - University of Wisconsin-Madison
  • Dyah Adila - University of Wisconsin-Madison
  • Tzu-Heng Huang - University of Wisconsin-Madison
  • Changho Shin - University of Wisconsin-Madison
  • Sonia Cromp - University of Wisconsin-Madison
  • Jeffrey Li - University of Washington
  • Cong Xu - Hewlett Packard Enterprise
  • Samuel Guo - Carnegie Mellon University
  • Adrien Pavao - University of Paris-Saclay
  • Ameet Talwalkar - Carnegie Mellon University
  • Frederic Sala - University of Wisconsin-Madison


University of Wisconsin—Madison Data Science Institute