InitAI
InitAI · Human video for robot learning

Rebuilt targets, same score.

Tracked, side-verified targets fully recover the detection experiment's ground and land level with the first fine-tune — and the flat line was itself a finding.

2026-08-20 · 3 min read
doc RUN-03 rev A issued

Rebuilt targets π0.5 pilot corpus 3,000 steps · 4.9 passes

Trained on the pilot corpus: 39 episodes / 26,040 frames, one take per task — the corpus that proved the line end to end. About the data and the exam ↗

Highlights

Overview

After the detection experiment, the detector line was rebuilt around two rules: no detection becomes a target unless it is tracked across frames, and no left/right assignment survives unless the side view confirms it. The rebuilt line passed a blind gate first: precision 99% and side accuracy 96.3% on frames no earlier round had seen, with genuinely ambiguous sides abstained rather than guessed. This run trained on those targets with the same footage and budget as the two runs before it.

Results

Three runs, one model, only the targets changed

Left hand — ratio to the assume-no-motion floor, lower is better

Detection experiment 1.1
First fine-tune 1.1
Rebuilt targets 1.1
The rebuild recovers the detection experiment's ground and lands level with the first fine-tune on the left hand.

Right hand

Detection experiment 1.4
First fine-tune 1.2
Rebuilt targets 1.2
Same picture on the right hand: corrected, not improved.

The ruler itself was audited

A dataset rebuild changes the exam as well as the training data, because the baselines are computed on the exam’s own clips. So the rebuilt-targets run was additionally scored on the archived exam base, where it reads 1.096 left and 1.32 right. Any improvement claim has to hold on both bases or it is an artefact of the ruler moving. This one held: level is level on either ruler.

Limitations

Single seed per run, so “level with the first fine-tune” is bounded by unmeasured run-to-run variance. Offline prediction metrics on held-out human video throughout.

Where this goes next

The matched-budget run holds everything fixed except the architecture. The detector gates and the corpus record live in the data write-up.

Sources

Scored by the programme’s frozen evaluation pipeline; the detector gates were graded on a blind 200-frame answer key marked by two independent squads. Full scoring artifacts are available to buyers on request.