Event knowledge in large language models: the gap between the impossible and the unlikely

Open in new window