Toxicity

Identify toxic language in AI outputs. Use toxicity detection to enforce safe and respectful responses.

The Toxicity metric is a span-level metric used to measure the degree of offensive, harmful, or inappropriate content in the response generated by the Agentic application. This metric is essential for ensuring that the model's outputs align with ethical and inclusive communication standards.

---

How to Run the Metric

1. Access the Dataset

   • Navigate to the dataset where you wish to evaluate toxicity.

   • Click on the Evaluate button.

2. Select the Metric

   • Choose Toxicity-Alteryx from the list of metrics.

   • Optionally, rename the metric for easier identification.

3. Choose the Evaluation Type

   • Select the evaluation type based on the component to evaluate:

     • LLM: For spans related to language model outputs.

     • Agent: For agent-level responses.

     • Tool: For outputs generated by specific tools.

4. Define the Schema

   • Specify the span name to evaluate.

   • Choose the parameter to analyse, such as:

     • input

     • output

     • ground truth

5. Configure the Model

   • Select the model configuration for the evaluation.

6. Set Passing Criteria

   • Define the pass/fail threshold to establish acceptable toxicity levels.

7. Run the Metric

   • Click on Run to initiate the evaluation process.

---