ExpansionRx–OpenADMET Blind Challenge: Webinar Recap and How to Participate

The community ADMET benchmarking challenge is live. Download the training set, build a model, and submit predictions to a live leaderboard hosted on Hugging Face. A held-out test split determines final rankings in January 2026. Data access and onboarding materials are linked below.

Open benchmarking only works if the underlying data are consistent and well-annotated. The ExpansionRx–OpenADMET challenge is built around an industry dataset contributed by ExpansionRx and released for community use. The goal is to shift effort from incremental algorithm tweaks toward rigor in data quality, evaluation, and reproducibility.

Participants build models for real-world ADMET endpoints commonly used in discovery (e.g., logD, solubility, microsomal stability, permeability, plasma protein binding). The training set is public; the test set is blinded for scoring.

How to participate

• Register: Create an account with Hugging Face.
• Walk through the tutorials: We have prepared a Tutorial showing how to train a model and submit to the leaderboard.
• Download the Public Dataset: Download the ExpansionRx training and blinded test sets from Hugging Face.
• Train Your Model: Use the provided training data for each ADMET property of your choice.
• Submit Predictions: Follow the instructions in the Submit tab to upload your predictions.
• Join the discussion on the Challenge Discord!