Download model_outputs.jsonl from auraone/evalkit-tutorial-v0.1: direct link, hf CLI and curl.
- Browser
- Download file 785 Bytes
-
https://huggingface.co/datasets/auraone/evalkit-tutorial-v0.1/resolve/main/model_outputs.jsonl
- Command line
-
hf download hf://datasets/auraone/evalkit-tutorial-v0.1/model_outputs.jsonl
-
curl -L -o model_outputs.jsonl https://huggingface.co/datasets/auraone/evalkit-tutorial-v0.1/resolve/main/model_outputs.jsonl
785 Bytes
| {"output_id":"out-001","prompt":"Review this synthetic parser change.","output":"The change handles empty input and cites parser.py, but it logs a token in debug mode and needs a regression test.","model":"tutorial-model-a","metadata":{"benchmark":false,"dataset":"synthetic tutorial"}} | |
| {"output_id":"out-002","prompt":"Review this synthetic API change.","output":"Looks fine overall. Maybe rename a variable.","model":"tutorial-model-b","metadata":{"benchmark":false,"dataset":"synthetic tutorial"}} | |
| {"output_id":"out-003","prompt":"Review this synthetic migration.","output":"The review names the compatibility risk, cites migration.ts, and asks for one targeted test in a professional tone.","model":"tutorial-model-c","metadata":{"benchmark":false,"dataset":"synthetic tutorial"}} | |