SFT
Technology & Computing abbreviation · Artificial Intelligence
| Stands for | Supervised fine-tuning |
|---|---|
| Categories | Technology & Computing |
SFT stands for Supervised fine-tuning, an artificial intelligence abbreviation.
Continued training of a pretrained model on curated examples of inputs paired with desired outputs, such as instructions and ideal answers, to teach it a task, format or style. It is usually the first step in turning a base language model into an assistant.
Example: Five thousand SFT examples taught the model to answer in our support team's tone.
Usage notes
Called instruction tuning when the data consists of instructions and responses. SFT is typically followed by preference-based methods such as RLHF or DPO. Data quality matters more than quantity: small, carefully written datasets often beat large noisy ones.
Related abbreviations
More artificial intelligence abbreviations
- RLHF – Reinforcement learning from human feedback
- RMSE – Root mean square error
- RNN – Recurrent neural network
- ROC – Receiver operating characteristic
- RPA – Robotic process automation
- SD – Stable Diffusion
- SGD – Stochastic gradient descent
- SHAP – SHapley Additive exPlanations
- SLAM – Simultaneous localization and mapping
- SLM – Small language model
- SOTA – State of the art
- STT – Speech-to-text