固定文字只探索布局,按实际数量验收
本页展示随包原生输入和验收参考。目标产品运行状态:未运行。表格与图片不代表产品已经生成的结果。
input.csv / practice-data.json
本课实际输入字段
| field | value | variantCount | creativeRange | aspects |
|---|---|---|---|---|
| event | 周末手作夜 | |||
| date | 2026-11-07 19:00 | |||
| price | ¥68 | |||
| 3 | EXPLORE | ["LAYOUT"] |
设置变体数量
reference.csv:作者验收映射,不是目标API回包
| returnedScreens | fixedText | difference |
|---|---|---|
| 3 requested, actual must be read | ||
| ["周末手作夜", "2026-11-07 19:00", "¥68"] | ||
| structure, not just hue |
规则与输入文件原文
{
"inputRows": [
{
"field": "event",
"value": "周末手作夜"
},
{
"field": "date",
"value": "2026-11-07 19:00"
},
{
"field": "price",
"value": "¥68"
},
{
"variantCount": 3,
"creativeRange": "EXPLORE",
"aspects": [
"LAYOUT"
]
}
],
"referenceRows": [
{
"returnedScreens": "3 requested, actual must be read"
},
{
"fixedText": [
"周末手作夜",
"2026-11-07 19:00",
"¥68"
]
},
{
"difference": "structure, not just hue"
}
],
"invalidRows": [
{
"variantCount": 20,
"error": "outside 1..5"
},
{
"aspects": [
"TEXT_CONTENT"
],
"error": "violates fixed content exercise"
}
],
"notes": [
"全部输入为作者自制教学资料,不含真实客户或账号。",
"referenceRows是校验基准,不是目标模型预测或识别结果;targetProduct=not_run。"
],
"files": {
"input.json": "[\n {\n \"field\": \"event\",\n \"value\": \"周末手作夜\"\n },\n {\n \"field\": \"date\",\n \"value\": \"2026-11-07 19:00\"\n },\n {\n \"field\": \"price\",\n \"value\": \"¥68\"\n },\n {\n \"variantCount\": 3,\n \"creativeRange\": \"EXPLORE\",\n \"aspects\": [\n \"LAYOUT\"\n ]\n }\n]\n",
"reference.json": "{\n \"identity\": \"作者验收参考,非模型返回\",\n \"rows\": [\n {\n \"returnedScreens\": \"3 requested, actual must be read\"\n },\n {\n \"fixedText\": [\n \"周末手作夜\",\n \"2026-11-07 19:00\",\n \"¥68\"\n ]\n },\n {\n \"difference\": \"structure, not just hue\"\n }\n ]\n}\n",
"invalid.json": "{\n \"identity\": \"作者构造负例,非实际错误回包\",\n \"rows\": [\n {\n \"variantCount\": 20,\n \"error\": \"outside 1..5\"\n },\n {\n \"aspects\": [\n \"TEXT_CONTENT\"\n ],\n \"error\": \"violates fixed content exercise\"\n }\n ]\n}\n",
"case.md": "# Stitch设计变体:固定文字只探索布局,按实际数量验收\n\nStitch设计变体通过variantCount、creativeRange和aspects限定探索。本课固定活动文案与价格,只比较布局差异,避免把同页改三个颜色误判为三种结构方案。\n\n## 固定内容基线\n\n保存活动名称、日期、价格和CTA文本,作为所有变体共同输入。\n\n价格不可因设计探索被AI改写。\n\n## 选择一个探索方面\n\n在当前SDK将aspects限定为LAYOUT,creativeRange选择REFINE或EXPLORE。\n\n不同时开放TEXT_CONTENT和所有方面,否则很难隔离布局变化。\n\n## 设置变体数量\n\n将variantCount设为3,读取当前SDK允许的1到5范围。\n\n范围外数字作为本地参数负例,提交前先拦截。\n\n## 执行并登记实际返回\n\n调用variants并保存返回Screen数组。\n\n实际返回不足3个时记为未达标,不补复制截图充数。\n\n## 比较结构变化\n\n逐张记录主区和侧栏的位置、CTA顺序与信息分组。\n\n只有色彩差异而结构不变时,本课布局探索尚不成立。\n\n## 核对固定文本\n\n把标题、日期和价格逐项与基线比对。\n\n任何内容变化都单独提出,不用“设计更好看”替代内容正确。\n\n## 选择并保留未选方案\n\n写明选中理由,保存三套身份和截图。\n\n未选中方案不立即删除,便于回看选择依据。\n\n身份:输入与参考为作者材料;localCli有实测记录时仅证明对应CLI运行,目标云端与模型推理未执行。\n"
}
}应当保留的约束
| returnedScreens | fixedText | difference |
|---|---|---|
| 3 requested, actual must be read | ||
| ["周末手作夜", "2026-11-07 19:00", "¥68"] | ||
| structure, not just hue |
故意构造的错误样例
| variantCount | error | aspects |
|---|---|---|
| 20 | outside 1..5 | |
| violates fixed content exercise | ["TEXT_CONTENT"] |
先保存真实产品输出,再逐项对照。actualOutput / actualResult 在未执行时保持 not_run,不能把这里的期望值复制成已测结果。