Children retrieving a stray football · different NAVSIM junction

What to verify

• Exactly 3 children and one persistent football throughout the 47 seconds.
• Children start playing off the road; the stray ball rolls into the road and the only road interaction is retrieval.
• Every kick happens off the roadway. Children who enter turn back immediately after the one-second pickup.
• All children and the ball have mapped sidewalk clearance from 10 seconds through the end.
• After returning, repeated football passes and child movement continue through the ending; no frozen post-clearance cast.
• Ego front bumper stops 1.1 m before the approximate line, remains outside the intersection, and resumes after two seconds of clearance.
• Unaffected lead car moves at 1.8 m/s throughout; related vehicles avoid all authored actors and context vehicles.
• Final meta-action is NORMAL and ego has passed the children.
• The NAVSIM anchor token, source log, map geometry and original camera photos differ from the first three rehearsals.
• Children avoid every background actor, including source pedestrians, throughout the full episode.

Review notes

Inspected all 21 event, stop-line, calibrated front-camera and football-contact views from committed revision 2, plus both native Higgsfield artifacts. This is a distinct NAVSIM anchor 38da4e087bf4599f and source log, with a different four-way junction and its own original camera photos. Exactly three children and one persistent ball: off-road play, one brief retriever, a one-second pickup, immediate return, and continued sidewalk passes and footwork through 40 s. Children and the ball are clear from 10 s. Ego stops with its front bumper 1.1 m before the approximate line upstream of mapped intersection 45225, waits two seconds after clearance and ends in NORMAL driving at 3 m/s past the children. The unaffected lead keeps moving at 1.8 m/s. The ball travels 33.88 m after return and is rolling at 1.4 m/s in the final frame; the third child is still moving. All 941 exported samples pass position, facing, sidewalk, relevant vehicle and child-to-background footprint checks. Visual optimization moved two overlapping source pedestrians to clear mapped sidewalk holding positions, recorded as counterfactual context changes, and gave the third child a lime shirt. The minimum child-to-background gap is 1.114 m. The source map and camera photos are unchanged; the approach line remains an approximation.

Check action, relative position, facing, posture and visibility. Approximate offsets are acceptable. Sampled visual checks do not prove every contact or continuous collision clearance.