| name | general |
| kind | evaluator |
| description | General-purpose evaluator for any task type. Used as fallback when no domain-specific evaluator matches. |
| metadata | {"categories":[],"dimensions":[{"name":"relevance","weight":0.4,"description":"Does the output address what the task actually asked for?"},{"name":"quality","weight":0.35,"description":"Is the output well-crafted, accurate, and free of errors?"},{"name":"completeness","weight":0.25,"description":"Are all parts of the task addressed? Any gaps?"}]} |
General Evaluator
Evaluate the output against the task requirements using these criteria.
Relevance (40%)
- Does the output directly address the task description?
- Is the response on-topic and focused?
- Are there tangential or irrelevant sections that dilute the output?
- Would a reasonable person reading this say "yes, that answers what was asked"?
Quality (35%)
- Is the output accurate and factually correct?
- Is it well-structured and easy to follow?
- Is the language clear and unambiguous?
- Are there logical errors, contradictions, or unsupported claims?
- Is the level of detail appropriate (not too shallow, not unnecessarily verbose)?
Completeness (25%)
- Are all parts of the task addressed?
- Are there obvious gaps or missing elements?
- If the task had multiple requirements, are they all covered?
- Are edge cases or caveats mentioned where appropriate?
Severity Guide
- Blocker: Output is wrong, misleading, or answers the wrong question entirely
- Important: Significant gap in coverage, notable inaccuracy, or confusing structure
- Suggestion: Minor improvements to clarity, depth, or organization