sortletters_parallel_121

STOP — rejected for behavior training. This derived retimed dataset may be used only for loader, mask, and runtime-throughput engineering. It must not support claims about a valid real-robot policy, sortletters behavior quality, or real-robot success rate.

警告：该数据已被行为训练资格门拒绝。 仅允许用于 loader、mask 和运行时吞吐工程验证； 不得用于声称有效真机策略或真机成功率。

For immutable source revision 76fd84d2f254f7acc95d5a5b4caafc00a03207a1, independent recomputation from all 121 meta/retime_source_indices/episode_*.npz files found:

• 65,532 / 66,053 frames (99.211%) pair the right-front observation with a different source timestamp from the reference arm timeline.
• 65,533 frames compose left and right arm sources from different timestamps.
• 121 / 121 episodes are affected.
• Median source-time delta: 531 frames (17.7 seconds at 30 FPS); maximum: 868 frames.

There is no jointly re-executed world observation or native success receipt that repairs this causal mismatch. Exact normalization and action loss masks do not make the trajectories coherent.

Allowed use: loader tests, padding-mask tests, and training-throughput engineering only.