Galileo Guardrail Metrics
Overview of Galileo Guardrail Metrics
Utilize Galileo’s Guardrail Metrics to monitor generative AI models, ensuring adherence to quality, correctness, and alignment with project goals.
Understand Galileo’s Guardrail Metrics in LLM Studio
Galileo has built a menu of Guardrail Metrics to help you evaluate, observe and protect your generative AI applications. These metrics are tailored to your use case and are designed to help you ensure your application quality and behavior. The 
Galileo’s Guardrail Metrics are a combination of industry-standard metrics and an outcome of Galileo’s in-house ML Research Team.
Scorer definition for each metric is listed immediately below.
Output Quality Metrics
- Correctness (Open Domain Hallucinations)
-
Instruction Adherence:
Scorers.instruction_adherence_plus - Uncertainty
-
Ground Truth Adherence:
Scorers.ground_truth_adherence_plus -
Completeness
-
Completeness Luna:
Scorers.completeness_luna -
Completeness Plus:
Scorers.completeness_plus
-
Completeness Luna:
- BLEU and ROUGE
Agent Quality Metrics
-
Action Completion:
Scorers.action_completion_plus -
Action Advancement:
Scorers.action_advancement_plus -
Tool Selection Quality:
Scorers.tool_selection_quality_plus -
Tool Error:
Scorers.tool_errors_plus
RAG Quality Metrics
-
Context Adherence (Closed Domain Hallucinations)
-
Context Adherence Luna:
Scorers.context_adherence_luna -
Context Adherence Plus:
Scorers.context_adherence_plus
-
Context Adherence Luna:
-
Chunk Attribution
-
Chunk Attribution Luna:
Scorers.chunk_attribution_utilization_luna -
Chunk Attribution Plus:
Scorers.chunk_attribution_utilization_plus
-
Chunk Attribution Luna:
-
Chunk Utilization
-
Chunk Utilization Luna:
Scorers.chunk_attribution_utilization_luna -
Chunk Utilization Plus:
Scorers.chunk_attribution_utilization_plus
-
Chunk Utilization Luna:
Input Quality Metrics
Safety Metrics
- Input & Output PII
- Input & Output Tone
- Input & Output Toxicity
- Input & Output Sexism
- Prompt Injection