MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering

MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering | OpenAI

Skip to main content

[](https://openai.com/)

* Research

* Products

* Business

* Developers

* Company

* Foundation(opens in a new window)

Log inTry ChatGPT(opens in a new window)

Try ChatGPT(opens in a new window)Login

OpenAI

October 10, 2024

Publication

MLE-bench

Evaluating Machine Learning Agents on Machine Learning Engineering

Read paper(opens in a new window)

We introduce MLE-bench, a benchmark for measuring how well AI agents perform at machine learning engineering. To this end, we curate 75 ML engineering-related competitions from Kaggle, creating a diverse set of challenging tasks that test real-world ML engineering skills such as training models, preparing datasets, and running experiments. We establish human baselines for each competition using Kaggle's publicly available leaderboards. We use open-source agent scaffolds to evaluate several frontier language models on our benchmark, finding that the best-performing setup — OpenAI's o1‑preview with AIDE scaffolding — achieves at least the level of a Kaggle bronze medal in 16.9% of competitions. In addition to our main results, we investigate various forms of resource-scaling for AI agents and the impact of contamination from pre-training. We open-source our benchmark code⁠(opens in a new window) to facilitate future research in understanding the ML engineering capabilities of AI agents.

October 10, 2024

Publication

MLE-bench

Evaluating Machine Learning Agents on Machine Learning Engineering

Authors

Chan Jun Shern, Neil Chowdhury, Oliver Jaffe, James Aung, Dane Sherburn, Evan Mays, Giulio Starace, Kevin Liu, Leon Maksin, Tejal Patwardhan, Lilian Weng, Aleksander Madry

* o1

* Software & Engineering

* Learning Paradigms

* Reasonings & Policy

Authors

Chan Jun Shern, Neil Chowdhury, Oliver Jaffe, James Aung, Dane Sherburn, Evan Mays, Giulio Starace, Kevin Liu, Leon Maksin, Tejal Patwardhan, Lilian Weng, Aleksander Madry