# probe_s12A_440k — decay probe of `s12A_arcmix_pool/` at step 440,000

**Why it exists:** a short side run that measures what the arm would score if training stopped here. The arm (ARC-MIX pool, building pair 1 arm A) trains at a constant learning rate, and constant-LR checkpoints are not comparable with decayed models. This probe copies the arm's step-440,000 checkpoint and runs only the decay phase: the learning rate decays from 3e-4 to 6e-5 over 4,000 steps with a 1−sqrt schedule (as in the trainer's WSD decay) (440,000 → 444,000), on the same data and with the same optimizer state. The arm itself is not affected and keeps training.

- **Use:** segment 2 of the pair is decided on the difference between this probe and the decay probe of arm B2 (`probe_s12B2_440k/`) on the held-out selection sets.
- **Status:** research checkpoint, **not a leaderboard submission**.
- **Format:** PyTorch checkpoint dict with `model`, `opt`, `step`, `config`; `train_gpt_ref.py` in the repository root rebuilds the model from `config`.
- **Data:** as in `s12A_arcmix_pool/`; see the root card of this repository.