AI engineering glossary · LLM training alignment
What is instruction backtranslation?
Instruction backtranslation: A method that takes unlabelled text such as web documents, has a model guess what instruction would have produced each one, and then keeps the best pairs as training data. It builds instruction datasets from existing writing.
Instruction backtranslation: an example
A blog post on Diwali recipes gets the generated instruction 'Give me a Diwali sweets recipe', forming a training pair.
Related terms
More LLM training alignment terms
- Instruction tuning
- Kahneman-Tversky optimisation
- KL penalty
- Knowledge distillation
- LIMA
- Long-context fine-tuning
- LoRA
- Loss masking
Where you learn it
Instruction backtranslation is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.