What happens next
We tune the selected open model on your masked gold set, then run the same comparison again. The results appear here after the run.
Nothing has been tuned in this demo, so there are no fine-tune numbers yet. The request button is for members; this demo is read-only.
Exactly what leaves this server
{
"type": "finetune_request",
"request_id": "(set when sent)",
"workflow_id": "7f3c1a52-0b6e-4d0e-9a41-2c5d8e9f1a01",
"gold_version_id": "a0000000-0000-4000-8000-000000000001",
"rubric_version_id": "a0000000-0000-4000-8000-000000000002",
"base_model": "dvar:gpt-oss-120b",
"gold_size": 300,
"pass_rate": "—"
}✓ Ids and counts only. No rows, no rubric text, no scores, no tokens. The Dvar connector adds its envelope: schema version, message id, data plane id and send time.
Conversations
None yet. After you send a request, its progress shows here:
- Started
- Engineer replied
- Scope agreed
- Model delivered