ad68c9f01146bca14b944d2dd6eef2be58d7496ff35693e2ec5f69065bf97864  .gitattributes
1956260cd25b3acdbe64322d5ba61f6afbc005a165ff8d0cb2da76f9aba5124a  .gitignore
cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30  LICENSE
b8e977ba93f14f391c6199a1fcb2ee7247b1cb227aff98634089ab7f4d88674b  NOTICE
b0788198a43f5b7b40ac0b7ec9a48a7324ed121cb936be2efe52d398b2e2e845  README.md
1fd951534cbd784f42c82942242864bfe85ceaa3ca751ff1fe154febede6d967  archive/create_dataset.original.py
a2ca02c9091f048195ae3749dbdb984485d9c4a931a59ce4627f106cea377041  archive/launch_char_count.original.sanitized.sh
4a3cd8e8b623e8e2b0a2a5d3dce32a97be2222dad18de7e6b09e17780ce89285  archive/training.sanitized.log
d30f7e35c9f077c3c27aae1012711a753939db81a9a2cf4bcdbdf03fab59d0a9  archive/uv.lock
3c107d544c23314389bdcb7a1bf5cde8787ec4dc2fc35f54113faa8f62484ecf  archive/verl-pyproject.toml
a3d4c278b084813891bba225402d771c121ca0e674540f10d9cd3ee62e9a2e05  data/rl/test.parquet
e9d5fa57da19d61bd13a187ee3b8e9da971f95f6ebcc17647163f287f41f7c7b  data/rl/train.parquet
3102943a3d288c833ee39af5a5084cbbb62c42ced3bffac29d86846ab27f5685  docs/conversation-excerpt.md
025d10dfdfa5f6a6d8308ae39dcedb462d50b0352962031e750f43df1ebad7c2  docs/experiment.md
cade8973f63f639ad6a7bb3f5aa9558d9ae8dc7b48d5fab4d2258e71afb51243  docs/provenance.md
4032dfbdef79e22e258c274bae83fd41ad2f23fab539c2f59a5b2f7464fe5c77  requirements-analysis.txt
9dd140c58768611c26f3b33225e4b53736c794505ae97060641e3d21b7f16bfc  results/metrics.csv
f78b9ddf2f7b2ea44cc2bd2584df16acad0b9a37420e118782abfa445fb8ce60  results/training-curves.png
96ce11685dc1dd344a6ab792af32ce344a3732b3d08703296b24e8b76db15b2f  scripts/create_dataset.py
d9d57e30803c49c35add60a372124aa89ccae8673fff2a831717f7c04bd3a7c0  scripts/extract_metrics.py
f180425d96bae670a05e90b6178e152717dff8e21e50f632371601b39f5c6b70  scripts/plot_metrics.py
560d575e7e7fffc39fe7aee2725e6d2642d8e8ef689e1ffb297c9f58b308336d  scripts/run_grpo_base.sh
4de190cc3294d41c76f8c9a6020e2995f0f8faa77981ab3049538676ce4f388c  scripts/setup.sh
e1ad95c0bfa57c0f2af9454e875dd0a4df937c6b5206b5e2244845b79f2296c5  scripts/train.sh
01dccdccb780655a199ccb7f8ef3ebb5668f1dc7260f7fc199c86c71ecdb92cf  src/reward_tolerant.py
00f793a497acd270ce9a822640a45507d11533e4dbb1c39e96f4cac7ff26756a  tests/test_release.py
