---
pretty_name: Cap-Evolve research archive
tags:
- checkpoints
- reinforcement-learning
---
# Cap-Evolve

Public research archive uploaded by YoungMan233. Upload is in progress.

The checkpoint stage preserves original relative paths under `repos/` and `Baselines/train/`.
`CHECKPOINT_UPLOAD_MANIFEST.json` lists the expected checkpoint files and sizes.
Checkpoint model, optimizer and training-state files are uploaded first; remaining project files follow separately.
Source files remain on the original machine. The directory is live, so this is not an atomic filesystem snapshot.
Only `CHECKPOINT_UPLOAD_VERIFIED.json` confirms that the checkpoint stage passed remote size/hash verification.