# Streaming RL Persistent Worker 3

- Updated: `2026-07-25T10:35:40.108303+00:00`
- Queue: `diffusion-training-acceleration`
- Lane: `causal`
- Current experiment: `srl_w03_defassign200`
- Current status: `failed`

| Experiment | Status | Return code | Latest log step |
|---|---|---:|---:|
| `srl_w03_zero10` | completed | 0 |  |
| `srl_w03_shuffle20` | completed | 0 |  |
| `srl_w03_assign100` | failed | 1 |  |
| `srl_w03_zero100` | failed | 1 |  |
| `srl_w03_shuffle100` | failed | 1 |  |
| `srl_w03_assign200` | failed | 1 |  |
| `srl_w03_tail100` | failed | 1 |  |
| `srl_w03_defassign200` | failed | 1 |  |

Rollout manifests and compact evaluations are mirrored to Hugging Face.
Trajectory tensors remain node-local; checkpoints and timers remain on S3.