polyguard-rl/submission_bundle/README.md

PolyGuard Submission Bundle

This folder contains push-friendly training evidence from the Hugging Face runs.

Included:

• grpo_training_cycle/: GRPO run plots, reports, and docs result JSONs.
• sft_baseline/: SFT baseline sweep plots, reports, and docs result JSONs.
• local_results/: Current local result mirrors from outputs/plots, outputs/reports, and docs/results.
• notebooks/: Project notebooks, including 09_training_loop.ipynb.

Not included:

• Model weights, adapters, checkpoints, optimizer states, or tokenizer/model binary payloads.
• These were intentionally removed after download cancellation so the repository can be pushed without large model artifacts.