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

In three linesOpenAI introduces MLE-bench, a benchmark for evaluating AI agents' performance on machine learning engineering tasks. The tool measures agents' ability to complete complex ML work.

Summary generated by Claude — human-verified