Few-Shot Learning
Few-shot learning is a machine learning technique where a model performs a new task after being shown only a small number of examples, typically 2 to 10, without parameter updates or retraining. In LLMs, few-shot learning is implemented by including example input-output pairs in the prompt itself. The model uses its pre-trained understanding of language to infer the task from the examples. Few-shot learning works because modern LLMs have been trained on such diverse data that they can recognize and replicate patterns from minimal demonstrations.
In Financial Services
Real-World Example
A compliance team at a European bank uses few-shot learning to classify vendor contracts under SFDR requirements. With three examples, the model processes 500 contracts in under an hour with 94% accuracy, saving 150 person-hours. When regulation was updated, the team simply modified the three examples.
Why It Matters for Finance
Few-shot learning democratizes AI in financial services by putting powerful capabilities directly in the hands of domain experts without requiring ML expertise. This flexibility is particularly valuable in finance where regulations change frequently.
Related Terms
Explore in Finatune
Frequently Asked Questions
What is few-shot learning in finance AI?
Few-shot learning is a technique where an AI model performs a task after being shown only a few examples, without additional training. In finance, it allows rapid adaptation without ML expertise.
How does few-shot learning help with financial document classification?
A compliance analyst can show an LLM three examples of suspicious transaction patterns and classify new ones. This enables rapid deployment of document classification for financial use cases.
When should I use few-shot vs fine-tuning for finance?
Use few-shot for quick adaptation with fewer than 500 examples or changing task definitions. Use fine-tuning for consistent performance at scale with stable, long-term tasks.