Just give it a try and check out the data with db-out 
In both cases, the example is saved with the respective answer: "answer": "accept" or "answer": "reject", and an empty list of spans: "spans": [].
Accepted examples with no spans are super important, because you always want your model to lean about texts that do not contain entities. When you train from the annotations in ner.batch-train, rejected examples are skipped. See here for more details: When to reject in ner.manual or ner.make-gold?