Learn about the AutoML SOTA

[gpsaggese]

We want to collect / learn all the interesting info about AutoML

[gpsaggese]

• AutoML to Date and Beyond: Challenges and Opportunities https://arxiv.org/pdf/2010.10777

  • This paper explains the SOTA as of 2021
  • The highlight is that EDA, data on-boarding, and potential feature ideation are still open issues (which is what we are working on)

• [ ] @gpsaggese

• [ ] @sonniki

• [ ] @Shaunak01

• [ ] @tkpratardan

[PranavShashidhara]

Hi Team,

I came across this recent ICML 2025 paper that might be relevant to the discussion on AutoML SOTA, AutoML-Agent: A Multi-Agent LLM Framework for Full-Pipeline AutoML

It proposes a new LLM-based multi-agent system that automates the whole ML pipeline, from data retrieval to deployment using smart planning and specialized agents for tasks like data cleaning, model design, and code generation.