How does the comp extraction work? 8-K prose has no standard format so curious whether you're running it through an LLM or using a rules-based parser, and how you handle amendments where the actual figures show up in a later filing.
The 8-K text gets run through an LLM to extract the structured fields. It handles the variation in format pretty well. For amendments, the system picks up the amended filing and updates the record — the comp figures sometimes only appear in the amendment.