CommonGen is a constrained text generation task, associated with a benchmark dataset,
to explicitly test machines for the ability of generative commonsense reasoning. Given
a set of common concepts; the task is to generate a coherent sentence describing an
everyday scenario using these concepts.
CommonGen is challenging because it inherently requires 1) relational reasoning using
background commonsense knowledge, and 2)… See the full description on the dataset page: https://huggingface.co/datasets/allenai/common_gen.
