ContextRecallMetric<T>
Models & Types ยท AiDotNet.RetrievalAugmentedGeneration.Evaluation
Evaluates context recall (RAGAS): what fraction of the ground-truth answer is supported by the retrieved contexts.
Inheritance: RAGMetricBase<T> โ ContextRecallMetric<T>
How It Works
Context recall measures retrieval completeness. The ground-truth answer is broken into individual claims (sentences); each claim is checked for whether it can be attributed to the retrieved contexts. The score is attributable-claims / total-claims. A low score means the retriever missed information that the correct answer depends on.
Attribution of each claim is decided by the text generator (a yes/no judge) when one is supplied, otherwise by cosine similarity against an embedding model, otherwise by an offline lexical word-coverage threshold.
Constructors
| Constructor | Summary |
|---|---|
ContextRecallMetric(ITextGenerator generator, IEmbeddingModel<T> embeddingModel, double attributionThreshold) |
Initializes a new instance of the ContextRecallMetric class. |
Properties
| Property | Summary |
|---|---|
string Description { get; } |
Gets the description of what this metric measures. |
string Name { get; } |
Gets the name of this metric. |