datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
biosciences-competency-questions-sample
Open Biosciences Competency Questions (Sample)
Dataset Description
A curated collection of 15 competency questions (CQs) for evaluating and guiding knowledge graph construction in biosciences research. Each question includes structured entities with standardized CURIEs, gold-standard knowledge graph paths using BioLink predicates, executable multi-API workflow steps, and source provenance.
What Are Competency Questions?
Competency questions are natural language… See the full description on the dataset page: https://huggingface.co/datasets/open-biosciences/biosciences-competency-questions-sample.ARC-Bind-Energy
ARC-Bind Energy Prediction
Overview
ARC-Bind is a synthetic 2D binding dataset designed to study how machine learning models learn to predict molecular interactions. The goal is to understand model reasoning about binding in a setting small enough to fully inspect and debug.
Inspiration: ARC-AGI Format
Inspired by François Chollet's ARC-AGI benchmark, ARC-Bind uses 16×16 pixel grids with discrete color tokens to represent binding:
Protein: Light blue (token 8)… See the full description on the dataset page: https://huggingface.co/datasets/Leash-Biosciences/ARC-Bind-Energy.ARC-Bind-Pose
ARC-Bind Pose Prediction
Overview
ARC-Bind is a synthetic 2D binding dataset designed to study how machine learning models learn to predict molecular interactions. The goal is to understand model reasoning about binding in a setting small enough to fully inspect and debug.
Inspiration: ARC-AGI Format
Inspired by François Chollet's ARC-AGI benchmark, ARC-Bind uses 16×16 pixel grids with discrete color tokens to represent binding:
Protein: Light blue (token… See the full description on the dataset page: https://huggingface.co/datasets/Leash-Biosciences/ARC-Bind-Pose.