DiversityExampleSelector<T>
Models & Types ยท AiDotNet.PromptEngineering.FewShot
Selects diverse examples to maximize coverage of different patterns.
Inheritance: FewShotExampleSelectorBase<T> โ DiversityExampleSelector<T>
For Beginners
Picks examples that are different from each other.
Instead of picking similar examples, this ensures variety:
- If you have 100 sentiment examples
- Random might give 3 positive examples
- Diversity gives 1 positive, 1 negative, 1 neutral
Example:
Use this when:
- You want broad coverage of the example space
- Your examples cluster into natural groups
- Model needs to understand the full range of possibilities
How It Works
This selector picks examples that are different from each other to provide broad coverage. It uses a greedy algorithm to iteratively select examples that are most different from already-selected examples.
Constructors
| Constructor | Summary |
|---|---|
DiversityExampleSelector(Func<string, Vector<T>> embeddingFunction) |
Initializes a new instance of the DiversityExampleSelector class. |
DiversityExampleSelector(Func<string, Vector<T>> embeddingFunction, T diversityThreshold) |
Initializes a new instance of the DiversityExampleSelector class. |
DiversityExampleSelector(IEmbeddingModel<T> embeddingModel) |
Initializes a new instance of the DiversityExampleSelector class. |
DiversityExampleSelector(IEmbeddingModel<T> embeddingModel, T diversityThreshold) |
Initializes a new instance of the DiversityExampleSelector class. |
Properties
| Property | Summary |
|---|---|
T DiversityThreshold { get; } |
Gets the diversity threshold used for selection. |