Few-Shot Learning¶
Few-shot learning refers to the ability to learn or adapt to a new task from a very small number of examples (typically 1-100). For language models, few-shot learning occurs through in-context learning, where a few task demonstrations are provided in the model's input prompt.
Key papers¶
- Language Models are Few-Shot Learners — Foundational work demonstrating that scaling up language models dramatically improves few-shot learning performance without task-specific fine-tuning
- Atlas: Few-shot Learning with Retrieval Augmented Language Models — Shows that retrieval-augmented models achieve competitive few-shot performance with 50× fewer parameters