Explore high-quality datasets for your AI and machine learning projects.
The BIG-Bench Hard dataset comprises multiple sub‑tasks, each associated with a configuration name such as boolean expressions, causal judgement, date understanding, etc. Each sub‑task contains input and target features, and every configuration has a test set with 250 examples (unless otherwise noted). The dataset is primarily used to evaluate and challenge the performance of natural language processing models on complex tasks.
