Design an evaluation framework for: [YOUR AI APPLICATION]

**Context:**
- Output type: [text/code/structured data/etc.]
- Quality dimensions: [accuracy, helpfulness, safety, style]
- Stakeholders: [who defines "good"]

Create a comprehensive framework:

1. **Evaluation Dimensions**
   - Define 5-7 quality dimensions
   - Rubric for each (1-5 scale with descriptions)
   - Weighting by importance

2. **Automated Metrics**
   - Programmatic checks (format, length, keywords)
   - Reference-based metrics (BLEU, ROUGE, etc.)
   - Model-based evaluation (LLM as judge)
   - Consistency checks

3. **Human Evaluation Protocol**
   - Evaluator selection criteria
   - Task instructions
   - Annotation interface design
   - Inter-rater reliability targets

4. **Test Set Design**
   - Coverage across use cases
   - Edge cases and adversarial inputs
   - Golden answer creation
   - Periodic refresh strategy

5. **Reporting & Analysis**
   - Score aggregation methods
   - Trend visualization
   - Failure analysis workflow
   - Improvement prioritization

6. **Integration into Development**
   - Pre-deployment gates
   - Continuous monitoring
   - Regression alerts
   - Feedback to prompt improvement

Output as an actionable evaluation plan with templates.