Datasets:
Download evaluation-plan.json from instagrim/lexsi-trajectory-data: direct link, hf CLI and curl.
- Browser
- Download file 721 Bytes
-
https://huggingface.co/datasets/instagrim/lexsi-trajectory-data/resolve/main/evaluation-plan.json
- Command line
-
hf download hf://datasets/instagrim/lexsi-trajectory-data/evaluation-plan.json
-
curl -L -o evaluation-plan.json https://huggingface.co/datasets/instagrim/lexsi-trajectory-data/resolve/main/evaluation-plan.json
721 Bytes
| { | |
| "heldout_sha256": "b68fbba8e3a70e966749bb1557346dd9cc717d86830044a5b1f8b63764f2f66b", | |
| "train_steps_sha256": "9d640e13f1df15ce5e2fa7ca74260352a1bf0800a9953bae8b0cf3f55e0e5d99", | |
| "system_prompt_sha256": "edfc484ec663368416e176f90c6c6508ae5d770017ab4bfc571b99eb52a7dadc", | |
| "max_episode_steps": 8, | |
| "max_new_tokens": 64, | |
| "decoding": "greedy", | |
| "training_optimizer_steps": 168, | |
| "model": "CohereLabs/tiny-aya-global", | |
| "revision_claim": "af89d219b53ed9b13b8a4645f9c8028973510324", | |
| "metrics": [ | |
| "safe_completion", | |
| "unsafe_attempts", | |
| "invalid_calls", | |
| "case_outcome", | |
| "steps" | |
| ], | |
| "qualification": "reference policy passed all 18 synthetic cases; no model held-out results used for setup" | |
| } | |