Describe the bug
Grader appears to be looking for an exact prior solution match rather than task success.
Seen +ve only from unsandboxed Claude Code smoke test.
Even acknowledged in the task metadata:
and emits the accumulated energy shift as a trailing GlobalPhase reduced mod 4*pi (RZ(-phi) on the control wire in the controlled variant). Many mathematically equivalent circuits produce the same unitary but a different op stream, and the merging/skip/phase conventions have no in-repo precedent to copy (labs research code), so a plausible correct-from-first-principles implementation fails the exact-match assertions.
Describe the bug
Grader appears to be looking for an exact prior solution match rather than task success.
Seen +ve only from unsandboxed Claude Code smoke test.
Even acknowledged in the task metadata: