A deep read of the SONIC paper's training section revealed that every run since June carried a set of reward overrides — originally a symptomatic patch for "policy can't keep up with the reference" — that pushed tracking weights 2–4× away from paper values. The patch cured the June symptom but its side effect became the main disease: heavy global-position pressure teaches the policy to hover near the anchor with fast small steps instead of reproducing the motion.
| reward term | modified (since June) | paper / official default |
|---|---|---|
| tracking_body_linvel | w 4.0 / std 0.4 | w 1.0 / std 1.0 |
| tracking_anchor_pos | w 2.0 / std 0.5 | w 0.5 / std 0.3 |
| rel_body_pos / ori | w 2.5 | w 1.0 |
| vr_5point | w 4.0 / std 0.1 | not listed in the paper's table |
Official philosophy: don't press hard on global position — position correction belongs to the upstream reference-stream closed loop; the low-level tracker's job is to reproduce the motion itself. Batch arithmetic also matters: official SONIC trains at 524,288 parallel envs (4096 × 128 GPUs, ~12.6M samples per update, 3 days); our single-GPU batch went from 256 env × 24 steps to 2048 env × 96 steps (8×) and the official rewards started working immediately.