Instruction Tuning
Training a base model to follow human instructions by showing it examples of good responses.
Definition
Instruction tuning (also called Supervised Fine-Tuning or SFT) is the first stage of making a base model useful — feeding it (instruction, ideal-response) pairs so it learns the format of helpful answering. Comes before RLHF in the typical pipeline.
Examples
- Writing a model-quality answer to "Plan a 7-day Tokyo trip for $2k"
- Producing reference solutions to math problems
Related terms
RLHF
Reinforcement Learning from Human Feedback — the technique that made ChatGPT helpful instead of just plausible.
Fine-Tuning
Adapting a pretrained model to a specific task or domain by continuing training on new data.
SFT (Supervised Fine-Tuning)
The technical name for instruction tuning — supervised learning on (input, ideal-output) pairs.
Looking for actual gigs that use Instruction Tuning? Browse all live listings or filter by $100+/hr.