Instruction Tuning
Fine-tuning a model on examples of instructions and their responses.
Visão geral
Instruction tuning trains a pretrained model on a large collection of tasks phrased as instructions, so it learns to follow a request rather than continue text. It solves the problem that a model trained only to predict the next token treats a question as something to extend rather than something to answer. It exists as the step that turns a language model into one that can be asked to do things.
Grafo de tecnologias
Como Instruction Tuning se liga ao resto do stack.
Usa e depende de
- Fine-tuning technique Techniques Instruction Tuning based on Fine-tuning High ✓ source
Aplicado a
- Alignment technique Safety & Alignment Instruction Tuning used for Alignment High ✓ source
- Chatbot application Applications Instruction Tuning used for Chatbot High ✓ source