Research question
Can current methods make acceptable selection and timing decisions under an editorial goal, and where does performance break down?
Purpose
Determine which decisions current methods already handle, identify reproducible failures, and establish whether a new representation is needed for those failures.
Acceptability is evaluated under a specified goal and allowed editing operations. Multiple cuts may be valid; preference among valid edits is assessed separately.
Same-task comparison: Wan 2.7
Same-task comparison: Agent-LTX
Same-task comparison: MiniMax H3