Reproducing results on MLE-Bench

[jfc43]

Hello,

I am interested in reproducing the results reported on the MLE-Bench leaderboard for the R&D Agent using GPT-5 (https://github.com/openai/mle-bench/tree/main?tab=readme-ov-file).

Could you please provide the detailed instructions or artifacts required to replicate this setup? Specifically, I am looking for:

• The specific configuration files (or hyperparameters) used.
• The exact command-line arguments to run the evaluation.

Thank you!