What is zero-shot learning?
Zero-shot learning tackles tasks with no examples.
How It Works:
Models generalize to unseen classes or tasks by leveraging semantic embeddings or descriptive prompts, mapping novel inputs to known concepts.
Key Benefits:
- No labeled data needed for new categories
- Rapid support for evolving requirements
- Leverages rich pretraining
Real-World Use Cases:
- Classifying new product types without retraining
- Understanding novel user intents in chatbots
FAQs
How accurate is zero-shot?
What models support it?
.png)
.svg)
