Definition
Plain language
A rectangle that marks exactly where an item sits and how big it is on a page or image.
As stated in the literature
The coordinate rectangle (position and extents) enclosing an element in a layout or image; grounded layout checkers read bounding boxes to detect overlaps and off-page overflow that a screenshot judge misses.
Also called: bounding boxes
Why it matters: It lets a system reason about exactly where things sit and how big they are, catching overlaps and overflow that a plain visual glance would miss.
For example, a layout checker can spot that a photo's rectangle runs past the edge of the page or sits on top of a caption.
Heard on the show
“Then they pointed a grounded tool at the same figures, one that reads the actual bounding boxes of every element.”Episode 217 — Why an AI Called Fourteen Broken Figures Perfect, And What It Reveals About Test-Time Compute