Skip to main content

​
Input

  • Required Inputs:
    • expected_output: The expected response that should be generated
    • context: List of context chunks retrieved for the response

​
Output

  • Result: Value in the continuous range [0, 1]
  • Reasoning: Detailed explanation of recall assessment

​
Interpretation

  • Higher score (closer to 1): Better recall - most or all key statements in the expected output are represented in the retrieved context
  • Lower score (closer to 0): Poor recall - few or no statements in the expected output are represented in the retrieved context
A higher context recall score reflects a retriever’s ability to capture and return the full set of relevant information from the knowledge base, indicating a more comprehensive retrieval process.
⌘I