AnswerRefusalScorer
Evaluates whether the model appropriately refuses to answer harmful or inappropriate requests. Assesses if the model correctly identifies dangerous queries and provides appropriate refusals. Important for compliance and guardrails.
Overview
Evaluates whether the model appropriately refuses to answer harmful or inappropriate requests. Assesses if the model correctly identifies dangerous queries and provides appropriate refusals. Important for compliance and guardrails.
Use Cases
- Safety and compliance monitoring
- Conversational AI quality assessment
How It Works
This scorer uses LLM-as-Judge technology to evaluate responses. It prompts a large language model with specific evaluation criteria and the content to assess, then analyzes the LLM's judgment to produce a score and detailed reasoning.
Input Schema
| Parameter | Type | Required | Description |
|---|---|---|---|
| output_text | str | Yes | The model's response to evaluate |
| input_text | str | Yes | The original question/request |
| context.system_guidelines | str | No | System guidelines for refusal behavior |
Output Schema
| Field | Type | Description |
|---|---|---|
| score | float | Refusal appropriateness (0-10) |
| passed | bool | True if refusal behavior is correct |
| reasoning | str | Analysis of refusal appropriateness |
| metadata.should_refuse | bool | Whether question should be refused |
Score Interpretation
Default threshold: 7/10
Related Scorers
ContentModerationScorer
Evaluates content against moderation guidelines and policies. Checks for violations across multiple ...
AgentRoleAdherenceScorer
Assesses how well the agent adheres to its defined role and persona throughout the conversation. Eva...
ConversationalConversationalRoleAdherenceScorer
Measures how well an AI maintains its assigned role or persona throughout a conversation. Evaluates ...
SafetyContentSafetyViolationScorer
Detects specific safety policy violations in generated content. More granular than general moderatio...
Frequently Asked Questions
When should I use this scorer?
Use AnswerRefusalScorer when you need to evaluate safety and conversational aspects of your AI outputs. It's particularly useful for safety and compliance monitoring.
Why doesn't this scorer need expected output?
This scorer evaluates quality aspects that don't require comparison against a reference answer. It uses the system prompt and context as the implicit ground truth.
Can I customize the threshold?
Yes, the default threshold of 7 can be customized when configuring the scorer.
Quick Info
Ready to try AnswerRefusalScorer?
Start evaluating your AI agents with Noveum.ai's comprehensive scorer library.
Explore More Scorers
Discover 106 calibrated LLM-as-Judge scorers for comprehensive AI evaluation