AdaDex: Rebuttal Materials

Anonymous submission · under double-blind review

Closed-Loop Real-World Rollouts

In addition to the open-loop rollouts on the project page, we run the policy closed-loop on the real arm-hand system at 1x speed (30 Hz). Videos are played at 1x. Background regions are blurred for anonymity.

Cup · pour (closed-loop)

Hammer · use (closed-loop)

Reference-plus-Noise vs. Random Goals

Both policies combine motion tracking with goal-reaching training (50/50 TrackEnv/GenEnv). They differ only in how goals are sampled in the goal-reaching environments.

Left — Reference-plus-noise (AdaDex): goals are sampled by perturbing future poses from the demonstration.
Right — Random goals: object goals are sampled independently of the demonstrated trajectory.

Each policy with its native training goal sampler

Each policy receives the type of goals used during its training. The left follows perturbed demonstration goals; the right follows independently sampled random goals. This shows the behavior each training objective encourages.

AprilCube

Marker

Pen

Same test commands

Both policies receive the same goal sequence, with matched Stage-2 iteration counts. These examples examine how the two training approaches retain and extend the demonstrated manipulation skills.

AprilCube

The same goal sequence asks both policies to perform two consecutive in-hand rotations.

Marker

The same noisy-reference goal sequence is given to both policies.

Pen

The same noisy-reference goal sequence is given to both policies.

These are selected qualitative examples, not a random sample of all benchmark outcomes. Some paired videos hold the final frame when one rollout ends earlier.