Add response eval case

Name Prompt/context you responded to Response to score Rubric JSON (optional) Create eval case

New fine-tuning run (LoRA / QLoRA on AWS)

Builds an instruction dataset from the published taste corpus, then trains Llama 3.1 8B with QLoRA on a g4dn.xlarge. Cheap, fast validation before scaling up.

Run name Base model Create run & build dataset