Hi Jorge,

Great question. we've noticed that inferring the entity from the name of the label alone doesn't work well in case of ambigous entity or when the context is not present in vicinity of the entity. We've recently added an update, where you can attach a description for each label so you can instruct and guide the LLM better. You can for example, instruct it to look at the entire document and only label the entity if it finds the corresponding context.