Camera, command,
response.

Four recordings. Every decision retained.

The four teacher diagnostics complete placement. They cover two starting positions, each with two destinations. These are recorded simulation frames, not a live arm or a learned-policy demonstration.

RGB display is sampled at 2.5 Hz from the original 25 Hz camera. The slider and measurements retain every 50 Hz motor decision. Frame time is shown separately; display sampling does not change the controller's observations.

Object estimate

—

Source observation age

—

Teacher phase (offline)

—

Decision authority

—

The target is not the measurement.

When the object estimate is available

Available Missing. Availability says an estimate was returned; it does not certify its accuracy.

The blind spot is the finding.

Across these four diagnostics, the object tracker returns no pose during 749 of 749 close, lift, transfer, place and release decisions. Missing stretches last 7.24–7.96 simulated seconds. Submillimetre errors on available estimates do not mean the object is tracked throughout manipulation.

The frozen learner must use its explicitly historical initial visual position and robot measurements through that gap. Offline teacher-phase labels explain the chart; they are excluded from learner inputs. Teacher actions are synthetic labels from an authored controller with privileged simulator state.

Qualification remains pending. This subset supplies no student success rate, superiority result, physical-camera accuracy, hardware latency or hardware qualification. The registered protocol keeps the full demonstration, correction and confirmation budgets separate.

All plotted measurements retain their original values. Issued targets and final rate-limited servo targets are different commands at different times; the gripper value is one finger's slide position, not total jaw opening. Camera capture and arrival use the simulation clock. No wall-clock timing is inferred from playback.