One-shot Learning is a setting where the model sees a single example of the task in addition to the task description before being asked to perform it.
Example
- Prompt:
- “Sea otter translates to loutre de mer” (Example)
- “Translate cheese into French” (Target Task)
- Output: The model uses the single example as a hint or guide to generate the translation.
Related
- Zero-shot Learning: No examples provided.
- Few-shot Learning: More than one example provided.
