Here’s a detailed summary of what you receive after submitting your raw model metrics to the Trace Metric API. This document explains your results, highlights key findings, and guides you on how to access and use your report.
What You Get After Using the Trace Metric API
After submitting your raw metrics, the Trace Metric API processes your data and generates a comprehensive evaluation of your AI model. Here’s what your report includes:
Health Index & Star Rating:
A quick snapshot of your model’s overall quality and risk level.Pillar Scores:
Assessment across critical dimensions such as performance, fairness, safety, task adherence, reliability, robustness, and privacy.Metric Breakdown:
Detailed analysis of key metrics like answer relevance, contextual precision and recall, faithfulness, bias, and hallucination.Risks & Business Impacts:
Identification of potential risks and real-world business consequences.Recommended Actions:
Tailored next steps for engineers, managers, and compliance teams.Summary Table:
At-a-glance view of your model’s strengths, weaknesses, and priorities.
Your Model’s Results
Health Index & Star Rating
Health Index: 57 (out of 100)
Star Rating: ★★★☆☆ (3 out of 5)
Risk Colour: 🟠 (Moderate)
Pillar Scores
Pillar | Score | Colour | Metrics Count | Meaning |
Performance | 0.49 | 🔴 | 3 | Needs improvement in speed or relevance |
Fairness & Bias | 0.55 | 🟠 | 5 | Some bias detected, moderate risk |
Safety & Truthfulness | 0.67 | 🟠 | 3 | Acceptable, but could be more reliable |
Task Adherence | 0.0 | 🔴 | 0 | Not meeting task requirements |
Reliability | 0.0 | 🔴 | 0 | Unreliable performance |
Robustness | 0.0 | 🔴 | 0 | Not robust to changes or errors |
Privacy | 0.0 | 🔴 | 0 | Privacy risks present |
Metric Breakdown
Metric | Score | Risk Level | What It Means / Action Needed |
Answer Relevance | 0.88 | Medium | Generally relevant, but improve alignment |
Precision | 0.75 | Medium | Some irrelevant context, optimize retrieval |
Recall | 0.74 | Medium | Some context missing, ensure completeness |
Faithfulness | 1.0 | Low | Fully factual, maintain this standard |
Bias | 0.0 | Low | No harmful bias detected, keep monitoring |
Hallucination | 1.0 | High | No hallucinated content, maintain checks |
Risks & Business Impacts
Your report highlights several risks and their potential business impacts, such as:
User Churn: Poor or irrelevant answers can drive users away.
Control Gaps: Inadequate controls may lead to compliance failures.
Support Cost Spike: Inaccurate answers increase customer support needs.
Public Backlash: Harmful or offensive outputs can damage your reputation.
Regulatory Risks: Non-compliance with laws like the EU AI Act can result in legal consequences.
Recommended Actions
For ML Engineers:
Refine prompts and improve data quality.
Fine-tune models for your specific domain.
Optimize retrieval and ranking strategies.
Implement fact-checking and bias mitigation protocols.
For Business/Product Managers:
Align product requirements with user expectations.
Monitor relevance and satisfaction metrics.
Ensure retriever systems deliver complete and relevant responses.
For Compliance Managers:
Verify outputs meet content relevance and audit standards.
Maintain documentation for retrieval and ranking logic.
Track and mitigate bias for regulatory compliance.
How to Access Your Report
After you submit your raw metrics, you receive a unique Report ID (for example: gW3eQpV63sRTrQey9uJPNp). To access your detailed report:
Save Your Report ID:
This ID is your key to retrieve the full evaluation at any time.Access the Report:
Go to the Trace Metric API dashboard or portal.
Enter your Report ID in the “View Report” section.
Instantly view, download, or share your comprehensive evaluation.
Share with Your Team:
You can provide your Report ID to colleagues, auditors, or stakeholders for transparency and collaboration.
Why This Report Matters
Pinpoints Strengths & Weaknesses:
Quickly see where your model excels and where it needs work.Guides Compliance:
Helps you meet standards like the EU AI Act and NIST AI RMF.Drives Better User Experience:
Focuses your improvements on what matters most to users.Informs Business Decisions:
Connects technical risks to real-world impacts, helping you prioritize effectively.
Summary Table
Metric | Score | Risk Level | Action Needed |
Answer Relevance | 0.88 | Medium | Improve alignment & accuracy |
Precision | 0.75 | Medium | Optimize context inclusion |
Recall | 0.74 | Medium | Ensure complete responses |
Faithfulness | 1.0 | Low | Maintain factual grounding |
Bias | 0.0 | Low | Continue monitoring |
Hallucination | 1.0 | High | Maintain strict fact-checking |
Keep your Report ID safe and use it to revisit your evaluation any time. The Trace Metric API gives you the clarity and guidance you need to make your AI models safer, more reliable, and more aligned with your goals.