AI engineering glossary · LLM training alignment
What is self-rewarding language models?
Self-rewarding language models: A training approach where a model both generates answers and judges them, using its own judgements as preference labels for further preference training, so its answering and judging abilities improve together over iterations.
Self-rewarding language models: an example
Round one: the model writes 4 answers to each prompt and scores them; the best and worst form a preference pair for the next round.
Related terms
More LLM training alignment terms
- Self-Taught Reasoner
- Sequence packing
- Sequence-level distillation
- Simple preference optimisation
- Soft labels
- Span corruption
- Superficial alignment hypothesis
- Supervised fine-tuning
Where you learn it
Self-rewarding language models is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.