Question
01 · EVIDENCE TOPOLOGY
The answer exists only after both hops connect
Each node below is grounded to a released page and layout identifier.
02 · SOURCE EVIDENCE
Rendered pages with exact layout highlights
Gold boxes use the layout coordinates stored in the benchmark annotations.
03 · SYSTEM BEHAVIOR
Six baselines on one controlled instance
Every response and retrieved context is reproduced in full below.
04 · RAW JUDGE RECORDS
Per-hop decisions behind Step Coverage
No aggregate score is treated as a substitute for the underlying binary decisions.