Loading…
Glossary term
Fine-tuning a model on examples of instructions and good responses so it follows directions.
Instruction tuning fine-tunes a base model on many examples of a request paired with a good response. The model learns to treat input as an instruction to follow instead of text to continue. Diverse instructions help it generalize to new requests.
| Instruction | Good response | |
|---|---|---|
| Example 1 | Summarize this paragraph | A two-sentence summary |
| Example 2 | Translate to French: hello | Bonjour |
It is the step that turns a text-completion model into something you can talk to.