Definition
Plain language
Special markers around an answer that tell software exactly where the answer starts and ends.
As stated in the literature
Delimiter tokens enclosing a model's response field so it can be parsed programmatically; rigid wrapper-tag formats can suppress free-form reasoning text entirely.
Also called: wrapper-tag, wrapper tags
Why it matters: These markers make model outputs machine-readable, but overly rigid formats can squeeze out the reasoning a model would otherwise show, hiding how it reached its answer.
For example, a model might be told to put its final answer between markers like <answer> and </answer> so a program can pull out exactly that text and ignore everything else.
Heard on the show
“Then the author uses a rigid wrapper-tag format.”Episode 251 — When a Fake Dashboard Makes an AI Agent Just as Confident