模型未返回最终正文（finish_reason=length），以下为推理过程摘录，可尝试提高 max_tokens 或改用多段生成。
我们需要输出 HTML 正文片段，无 markdown 围栏，无解释。需要严格按照顺序和格式。
需求：
- 第一个元素是 <p> 改写应用介绍，自然出现应用名称。
- 然后按顺序输出小节：软件介绍 → 功能 → 优势 → 操作指南 → 更新说明 → 推荐理由 → 测评
- 但注意整体结构：随机抽 3～6 个非测评小节并优先保留采集中有原文者；若含「介绍」「简介」类关键词会固定排在首段 <p> 之后的第一个 <h3>，其余小节再按其下列顺序；末位固定测评。但这里用户给出了完整顺序，且包含全部 6 个非测评小节？ Actually they listed all sections. The instruction says "仅按下列「小节顺序」输出（已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者；若含...）" But then they list all 7 sections? Let's parse. The "整体结构" says:
1) first p
2) then 仅按下列「小节顺序」输出（已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者；若含「介绍」「简介」类关键词会固定排在首段 <p> 之后的第一个 <h3>，其余小节再按其下列顺序；末位固定测评，全文共 4～7 个 <h3>）：每个小节先 <h3>...
Then "【小节顺序】软件介绍 → 功能 → 优势 → 操作指南 → 更新说明 → 推荐理由 → 测评"
This is confusing: They say random 3-6 non-review sections, but then list all seven sections in order. In the detailed tasks, they give all sections with fixed number of items. So we must output all? Let's examine counts: if we output all 7 sections, we get 7 h3, which is at max. But they said "全文共 4～7 个 <h3>" so 7 is allowed. They also say "末位固定测评" yes. The random selection maybe not applied because they explicitly included all? Need decide.
Given user provided per-section tasks for every section, likely output all seven. But be careful: "仅按下列「小节顺序」输出" could mean output all in that order. They say "已从词池随机抽 3～6 个非测评小节" but then "下列顺序" includes all. Hmm.
Let's read exact: "【整体结构】顺序必须严格如下：
1) 第一个元素是一个 <p>：依据下方「首段」模板（{head}/{text} 均已代入为首段原文）改写应用介绍。
2) 然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>（标题须与下列 {H3title} 完全一致），非「测评」小节为 **若干** 个 <p>、每条要点须带序号且**全篇统一为同一种序号格式**（本次要求：用序号分条：每条单独成行，行首为「数字+全角顿号、」（示例 1、2、3、），全篇保持同一格式）；**每节要点条数以该小节下方【条数】为准**（在 3～6 条之间且各节独立）。
3) **最后一个小节固定为「灵格AI英语app测评」**（已在下列顺序末位）：**仅 1 个** <p>，**不要**序号与分条，**一段**短评，**全文不超过 200 字**（含标点）。
【小节顺序】软件介绍 → 功能 → 优势 → 操作指南 → 更新说明 → 推荐理由 → 测评"
They may be instructing the full canonical order. Then detailed sections all given. So output all sections? But "随机抽 3～6 个" says not all. However they didn't tell which are selected. Since they included all tasks, likely all seven are selected? Wait, total h3 count if all 7 = 7, exactly max. That contradicts "随机抽 3～6 个非测评小节" plus fixed evaluation = 4-7. So max is 7, all non-eval plus eval. So all seven is possible if random selection picks 6 non-eval. Yes random selection could be 6. They explicitly list all 6 non-eval sections in order, so we can assume selected all 6. Because if not selected, they wouldn't provide tasks? But they provided all.
Need output only allowed tags: <p>, <h3>. No other tags. No markdown code fences, no explanation. Need no img, no script/style. Use exact H3 titles.
Number format: Each non-review section has multiple <p>, each <p> one item. Use Chinese numerals? They require "数字+全角顿号、" e.g., "1、" "2、" etc. Use Arabic numeral + "、". Full-width comma? "1、" is Arabic digit followed by full-width ideographic comma. They specifically "数字+全角顿号、" — Actually "全角顿号" is "、" U+3001. So "1、" That's okay. Need each <p> starts with that. Since each <p> is separate line. Need ensure not include extra text outside.
Need write content. Let's plan.
First <p>: Must contain app name "灵格AI英语app" and rewrite intro. Based on original: "灵格AI英语app是一款全球超30个国家都在使用的AI英语学习神器，灵格AI英语app以视频学习为核心，涵盖英语教学、娱乐、影视、科技、商业、个人成长、KET/PET、四六级等多元频道，满足不同兴趣和学习需求。" Need original rewrite natural. We can create one paragraph:
"灵格AI英语app是一款已在全球超30个国家获得认可的AI英语学习工具，以视频学习为核心，整合英语教学、娱乐、影视、科技、商业、个人成长、KET/PET、四六级等多元频道，能贴合不同兴趣与学习需求，帮助用户轻松提升英语能力。"
Good.
Then h3 sections.
Now each section with fixed counts:
- 软件介绍: 4 <p> items.
- 功能: 5 items.
- 优势: 6 items.
- 操作指南: 4 items.
- 更新说明: 4 items.
- 推荐理由: 5 items.
- 测评: 1 <p> no numbering <=200.
Need each item roughly <=50 Chinese chars, one sentence. Need consistent numbering across all: each section restarts numbers? "全篇统一为同一种序号格式" means same style, not continuous numbering. Typically per section numbering 1-4, 1-5 etc. It says "每节要点条数以该小节下方【条数】为准" and "每条单独成行，行首为「数字+全角顿号、」（示例 1、2、3、），全篇保持同一格式". So each section has own numbers starting at 1. Good.
Need ensure each <p> one item; no line breaks within? We can output each paragraph on a separate line. The H3 tags then paragraphs.
Need think about content coherence, avoid repetition. Let's develop.
Section 1: 软件介绍. Focus: product positioning, target users, overall ability. 4 items.
Use original info. Possible:
1、灵格AI英语app是一款面向全球用户的AI英语学习工具，目前已在超30个国家广泛使用。
2、它以视频学习为核心，把教学、娱乐、影视、科技等频道整合在同一平台。
3、内容丰富多元，涵盖KET/PET、四六级、商业和个人成长等热门领域。
4、无论学生还是职场人士，都能按兴趣和需求找到适合自己的英语学习内容。
Need ensure each <p> is a <p> with item. Each sentence ~50 characters. Good.
Section 2: 功能. Focus: main ability modules and problems solved. 5 items.
Need from original. Possible:
1、AI智能视频推荐：根据用户兴趣和学习水平，自动推送匹配的英语视频内容。
2、多元频道分类：通过教学、影视、科技、商业等频道，解决内容选择困难。
3、备考专项内容：覆盖KET/PET、四六级等考试，帮助用户系统准备应试。
4、沉浸式学习体验：以真实视频场景提升听力、语感和口语表达能力。
5、个人成长资源：提供个人成长类内容，让英语学习与能力提升同步进行。
Need "功能" maybe "主要能力模块及各自解决什么问题." Good.
Section 3: 优势. 6 items. Focus: performance, usability, compatibility or ecosystem. Need not directly from original; can infer. Need maybe avoid too fictional. Since no original, can write generic benefits.
1、全球超30个国家的用户验证，内容质量与学习效果更具可信度。
2、以视频学习为核心，比传统背单词更生动，容易长期坚持。
3、频道划分清晰，支持按兴趣灵活选择，使用起来省时省心。
4、覆盖KET/PET、四六级等考试，一套应用满足备考与日常提升多种需求。
5、AI个性化推荐能够贴近用户水平，降低选课和找资料的难度。
6、跨年龄段与职业适用，学生、上班族都能从中获得适合的内容。
Need "兼容性" maybe mention multi-device? Could be:
6、支持手机、平板等多终端学习，随时随地打开即可继续练习。
But multi-device wasn't in original. Is that okay? "原创普通" allows. But maybe safer to derive from "全球30多国" implies accessible. We can include that.
Need output all sections. Need not include explanation.
Section 4: 操作指南. 4 items. Focus: key process step-by-step. Need define steps:
1、下载并安装灵格AI英语app，使用手机号或第三方账号完成注册登录。
2、进入首页后选择感兴趣的频道，例如英语教学、影视或四六级备考。
3、点击视频开始学习，可跟随AI功能进行跟读、听力练习和重点记录。
4、在我的学习页面查看进度与推荐内容，根据结果持续调整学习方向。
Need be careful: "AI功能" not specified but okay. Could phrase "利用AI推荐与练习功能". Good.
Section 5: 更新说明. 4 items. Focus: version additions/optimizations/fixes. No actual changelog, so generic:
1、新增更丰富的AI口语陪练场景，帮助用户提升表达流畅度。
2、优化视频推荐算法，让内容更贴合个人兴趣与学习目标。
3、完善KET/PET与四六级备考专题，补充更多真题和练习资料。
4、修复已知播放问题，提升视频加载速度与整体使用稳定性。
Need "本次版本" perhaps.
Section 6: 推荐理由. 5 items. Focus: new users why install, concrete benefits.
1、零基础也能轻松入门，视频学习和频道分类让起步更简单。
2、备考人群可集中获取KET/PET、四六级等资源，节省找题时间。
3、兴趣驱动学习，影视、科技等内容让英语练习不再枯燥。
4、AI推荐机制帮助规划学习路径，提升学习效率与成就感。
5、一次安装覆盖多种场景，适合长期伴随个人成长与职业发展。
Need ensure not too repetitive.
Section 7: 测评. 1 <p> no numbering, ≤200 chars. Need compose objective short review in first person maybe. Need "拟人化的叙述方式" hmm "尽量拟人化的叙述方式" means perhaps give human-like experience. Need include features, strengths, recommendation. No numbering. Let's draft:
"灵格AI英语app像一位懂你的语言伙伴，用视频内容把学习变得轻松自然。它覆盖教学、影视、科技、KET/PET、四六级等众多频道，既有备考干货，也有兴趣延伸，首次使用就能快速找到合适内容。AI推荐与跟读练习让输入和输出相互配合，长期坚持的阻力明显降低。虽然不同频道深度略有差异，但作为一款大众化英语学习工具，已足够实用，值得一试。"
Need count ≤200? Let's count roughly. 灵格AI英语app像一位懂你的语言伙伴，用视频内容把学习变得轻松自然。(30? Let's estimate. It contains Chinese chars + punctuation. We'll need ensure under 200. Word count including punctuation. Let's count later.
Need output as HTML body fragment, no markdown fences. Need ensure not include any text before/after. Need exactly allowed tags. So output:
<p>灵格AI英语app是一款已在全球超过30个国家获得认可的AI英语学习工具，以视频学习为核心，整合英语教学、娱乐、影视、科技、商业、个人成长、KET/PET、四六级等多元频道，满足不同兴趣与学习需求。</p>
<h3>灵格AI英语app软件介绍</h3>
<p>1、灵格AI英语app是一款面向全球用户的AI英语学习工具，目前已在超30个国家广泛使用。</p>
<p>2、它以视频学习为核心，把教学、娱乐、影视、科技等频道整合在同一平台。</p>
<p>3、内容丰富多元，涵盖KET/PET、四六级、商业和个人成长等热门领域。</p>
<p>4、无论学生还是职场人士，都能按兴趣和需求找到适合自己的英语学习内容。</p>
<h3>灵格AI英语app功能</h3>
<p>1、AI智能视频推荐：根据用户兴趣和学习水平，自动推送匹配的英语视频内容。</p>
<p>2、多元频道分类：通过教学、影视、科技、商业等频道，解决内容选择困难。</p>
<p>3、备考专项内容：覆盖KET/PET、四六级等考试，帮助用户系统准备应试。</p>
<p>4、沉浸式学习体验：以真实视频场景提升听力、语感和口语表达能力。</p>
<p>5、个人成长资源：提供个人成长类内容，让英语学习与能力提升同步进行。</p>
<h3>灵格AI英语app优势</h3>
<p>1、全球超30个国家的用户验证，内容质量与学习效果更具可信度。</p>
<p>2、以视频学习为核心，比传统背单词更生动，容易长期坚持。</p>
<p>3、频道划分清晰，支持按兴趣灵活选择，使用起来省时省心。</p>
<p>4、覆盖KET/PET、四六级等考试，一套应用满足备考与日常提升多种需求。</p>
<p>5、AI个性化推荐能够贴近用户水平，降低选课和找资料的难度。</p>
<p>6、支持手机、平板等多终端学习，随时随地打开即可继续练习。</p>
<h3>灵格AI英语app操作指南</h3>
<p>1、下载并安装灵格AI英语app，使用手机号或第三方账号完成注册登录。</p>
<p>2、进入首页后选择感兴趣的频道，例如英语教学、影视或四六级备考。</p>
<p>3、点击视频开始学习，可跟随AI功能进行跟读、听力练习和重点记录。</p>
<p>4、在我的学习页面查看进度与推荐内容，根据结果持续调整学习方向。</p>
<h3>灵格AI英语app更新说明</h3>
<p>1、新增更丰富的AI口语陪练场景，帮助用户提升表达流畅度。</p>
<p>2、优化视频推荐算法，让内容更贴合个人兴趣与学习目标。</p>
<p>3、完善KET/PET与四六级备考专题，补充更多真题和练习资料。</p>
<p>4、修复已知播放问题，提升视频加载速度与整体使用稳定性。</p>
<h3>灵格AI英语app推荐理由</h3>
<p>1、零基础也能轻松入门，视频学习和频道分类让起步更简单。</p>
<p>2、备考人群可集中获取KET/PET、四六级等资源，节省找题时间。</p>
<p>3、兴趣驱动学习，影视、科技等内容让英语练习不再枯燥。</p>
<p>4、AI推荐机制帮助规划学习路径，提升学习效率与成就感。</p>
<p>5、一次安装覆盖多种场景，适合长期伴随个人成长与职业发展。</p>
<h3>灵格AI英语app测评</h3>
<p>灵格AI英语app像一位懂你的语言伙伴，用视频内容把学习变得轻松自然。它覆盖教学、影视、科技、KET/PET、四六级等众多频道，既有备考干货，也有兴趣延伸，首次使用就能快速找到合适内容。AI推荐与跟读练习让输入和输出相互配合，长期坚持的阻力明显降低。虽然不同频道深度略有差异，但作为一款大众化英语学习工具，已足够实用，值得一试。</p>
Need verify total h3 count: software intro, function, advantage, operation, update, recommend, review = 7. Fine.
Need check instructions: "禁止输出 <img>" none. "允许的 HTML 标签仅有：<p>、<h3>" yes. No extra whitespace? Fine.
Need maybe "应用截图由系统按小节顺序自动插入，模型无需处理图片" irrelevant.
Need check first paragraph: Must be first element <p>. Yes.
Need check "应用名称（须在首段 <p> 正文中自然出现）" yes.
Need check "改写应用介绍" not "原文首段". Our first paragraph is close but not identical; okay.
Need check each non-review section is "若干 <p>" and "每条单独成行" yes each p has one item.
Need check "全篇统一为同一种序号格式" yes all use Arabic number + "、". Note "全角顿号、" is Unicode U+3001. Our "、" okay.
Need check "每节要点条数以该小节下方【条数】为准" counts:
软件介绍 4 p yes.
功能 5 p yes.
优势 6 p yes.
操作指南 4 p yes.
更新说明 4 p yes.
推荐理由 5 p yes.
测评 1 p.
Need check "测评：仅 1 个 <p>，不要序号与分条，一段短评，全文不超过 200 字（含标点）." Our p no numbering. Need ensure length <=200. Let's count accurately. Chinese char count maybe. We'll count with punctuation.
Sentence: "灵格AI英语app像一位懂你的语言伙伴，用视频内容把学习变得轻松自然。" Count:
灵(1)格2 A? Need count "AI" as two letters? The requirement "200字" usually Chinese characters? "含标点" implies counting all characters including ASCII? But we can keep under 200. Let's count approximate. Use Chinese char count plus punctuation maybe. "灵格AI英语app" = 灵格 2 + AI 2 + 英语 2 + app 3? Actually "app" 3 letters = 3. total 9? Let's count: 灵 (1), 格2, A3, I4, 英5, 语6, a7, p8, p9. So 9 characters. Then "像一位懂你的语言伙伴，" = 像1 一2 位3 懂4 你5 的6 语7 言8 伙9 伴10 ，11? The comma counts. So cumulative 9+11=20. "用视频内容把学习变得轻松自然。" = 用1 视2 频3 内4 容5 把6 学7 习8 变9 得10 轻11 松12 自13 然14 。15. Cum 35. Good.
Second sentence: "它覆盖教学、影视、科技、KET/PET、四六级等众多频道，" Count: 它1 覆2 盖3 教4 学5 、6 影7 视8 、9 科10 技11 、12 K13 E14 T15 /16 P17 E18 T19 、20 四21 六22 级23 等24 众25 多26 频27 道28 ，29. Cum 35+29=64. "既有备考干货，也有兴趣延伸，" = 既1 有2 备3 考4 干5 货6 ，7 也8 有9 兴10 趣11 延12 伸13 ，14. Cum 78. "首次使用就能快速找到合适内容。" = 首1 次2 使3 用4 就5 能6 快7 速8 找9 到10 合11 适12 内13 容14 。15. Cum 93. Good. "AI推荐与跟读练习让输入和输出相互配合，" Count: A1 I2 推3 荐4 与5 跟6 读7 练8 习9 让10 输11 入12 和13 输14 出15 相16 互17 配18 合19 ，20. Cum 113. "长期坚持的阻力明显降低。" = 长1 期2 坚3 持4 的5 阻6 力7 明8 显9 降10 低11 。12. Cum 125. "虽然不同频道深度略有差异，" = 虽1 然2 不3 同4 频5 道6 深7 度8 略9 有10 差11 异12 ，13. Cum 138. "但作为一款大众化英语学习工具，已足够实用，值得一试。" = 但1 作2 为3 一4 款5 大6 众7 化8 英9 语10 学11 习12 工13 具14 ，15 已16 足17 够18 实19 用20 ，21 值22 得23 一24 试25 。26. Cum 164. Under 200. Good.
Need consider "AI" and "KET/PET" count as letters; total still under 200. Fine.
Need check "测评" maybe "拟人化的叙述方式" Our "像一位懂你的语言伙伴" is anthropomorphic. Good.
Potential issue: The instruction "最后一个小节固定为「灵格AI英语app测评」（已在下列顺序末位）：仅 1 个 <p>，不要序号与分条，一段短评，全文不超过 200 字（含标点）." They may mean the entire output "全文不超过200字"? Wait phrasing: "（含标点）" for evaluation? It says "**一段**短评，**全文不超过 200 字**（含标点）" Here "全文" could mean the evaluation paragraph, not entire output? Because "全文" ambiguous. They say "一段短评，全文不超过200字" likely the paragraph. Given our evaluation is 164. Fine.
Need maybe "整篇一次生成时可作语气参考" no issue.
Need verify "首段模板" maybe use "组合采集首段/原文首段" and "标题：灵格AI英语app" We did.
Need "内容必须包含游戏/软件名称" yes.
Need "与原文的表达主旨保持高度关联与一致性" yes.
Need "可以根据原文的内容进行内容重组，使其逻辑清晰、层次分明，更偏向于一个应用商店的介绍文案." yes.
Need "首段" is before H3 "软件介绍"? In output, first p then h3. In instructions, if "介绍" keyword fixed after first p, yes software intro first. Good.
Need "其它小节再按其下列顺序" yes.
Need "应用截图由系统在生成后按小节顺序自动插入" no images.
Need no markdown code fences. We'll output plain fragment. Need not wrap in ```html.
Double-check tags: cannot output empty lines? Fine.
Need ensure each <p> is separated by newline. Yes.
Let's think if there is any hidden requirement about "rewriteKeys" maybe "优势 操作指南" sections need more depth? It says "rewriteKeys（入库多段链路中曾用于抽两段深度重写；整篇一次生成时可作语气参考）=[优势 操作指南]" Maybe perhaps "优势" and "操作指南" should be "深度重写" more detailed? But counts are fixed and each item 50 chars max, so no. Maybe we can make those sections slightly more detailed? But "每条一句、每条约50字内" so no.
Potential issue: The per-section "参考信息" only original first paragraph, no separate content. We can use original info. Fine.
Need consider "全篇统一为同一种序号格式" Does each section start at 1? Yes. But "全篇统一" could also mean continuous numbering across sections? Example says "1、2、3、" not necessarily restart. In typical Chinese app store copy, each section uses its own numbering. The instruction "每条要点须带序号且全篇统一为同一种序号格式" = same format, not continuous. "每节要点条数..." implies each section has its own count; if continuous, after first section with 4, second would start at 5. But not desired? Let's examine wording: "非「测评」小节为 **若干** 个 <p>、每条要点须带序号且**全篇统一为同一种序号格式**（本次要求：用序号分条：每条单独成行，行首为「数字+全角顿号、」（示例 1、2、3、），全篇保持同一格式）；**每节要点条数以该小节下方【条数】为准**（在 3～6 条之间且各节独立）。" The phrase "各节独立" likely means counts independent, not necessarily numbering. It says "数字+全角顿号" no "连续编号". So restart is acceptable. If they wanted continuous, they'd say "每节从1开始"? Not. In Chinese, "分条" in each section naturally begins at 1. I think okay.
Need maybe "仅输出一段可直接写入数据库的 HTML 正文片段" no explanation. Good.
Let's consider if "全文共 4～7 个 <h3>" and "已从词池随机抽 3～6 个非测评小节" maybe they don't want all sections? But all 6 non-eval = 6, total 7, okay.
But wait, "已从词池随机抽 3～6 个非测评小节" is not a statement that selection occurred; it's a system behavior description. It might mean in the actual generation, only 3-6 are selected randomly. But then "只按下列顺序" maybe the selected ones are a subset sorted by this order. Which ones selected? Since no selection is indicated in the prompt aside from the full list, we might need to infer from "若含「介绍」「简介」类关键词会固定排在首段 <p> 之后的第一个 <h3>" Because they mention "介绍" keyword appears in "软件介绍". So software intro is definitely selected. Then "其余小节再按其下列顺序": after intro, the remaining selected sections in order. Which others are selected? The prompt includes "rewriteKeys（...）=[优势 操作指南]" maybe maybe those two are selected for deep rewrite, so we should include them. Also "更新说明" is not "采集中有原文" maybe no. "功能", "推荐理由" maybe no. But detailed tasks for all sections exist. Hmm.
Let's reread exactly at top: "然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>..."
This parenthetical is part of the system template. "已从词池随机抽 3～6 个非测评小节" perhaps indicates the user's original request had a random process, but after substitution, the prompt still includes all section tasks? It may be a universal instruction for all variants, but in this variant, the "小节顺序" list is fixed. The phrase "仅按下列「小节顺序」输出（已从词池随机抽 3～6 个非测评小节...)" Could be interpreted as: The order "软件介绍 → 功能 → 优势 → 操作指南 → 更新说明 → 推荐理由 → 测评" is the canonical ordering, not necessarily all included. But they explicitly provide "各小节任务" for all sections. The "randomly selected 3-6" maybe could mean the system selected a subset from the word pool, but then the prompt still shows all tasks because it's generating a full list? No, the user asks "你只输出一段可直接写入数据库的 HTML 正文片段" and includes full "各小节任务" for all sections including "rewriteKeys" etc. The actual output should satisfy all "已代入模板" sections. So output all.
Let's see if "rewriteKeys（入库多段链路中曾用于抽两段深度重写；整篇一次生成时可作语气参考）=[优势 操作指南]" Means perhaps in the original multi-segment pipeline, two sections were selected for deep rewrite: "优势" and "操作指南". In this "整篇一次生成" task, use them as tone reference, not necessarily select. So no issue.
Potential issue: "更新说明" maybe should not be included because there is no "采集原文" and random selection could exclude it. But since task included, include.
Let's also think about "首段模板" says "请参考{组合采集首段}/{原文首段}，现在需要对这段内容进行原创改写，内容必须包含游戏/软件名称，与原文的表达主旨保持高度关联与一致性。可以根据原文的内容进行内容重组，使其逻辑清晰、层次分明，更偏向于一个应用商店的介绍文案。标题：灵格AI英语app；原文首段：..." This is the instruction for first paragraph only. We did.
Need maybe "软件介绍" first section should be 4 items; count okay.
Need "功能" section should maybe not mention "AI智能视频推荐" because "AI" is central. Good.
Need "操作指南" maybe should be 4 items; our steps are valid.
Need "更新说明" maybe "概括即可，勿堆砌版本流水账" we did generic.
Need "推荐理由" maybe "面向新用户说明为何值得安装使用，需落到具体收益" our items mention benefits.
Need "测评" should be "客观简练" and "拟人化" okay.
Let's consider if "禁止 markdown 代码围栏" means no triple backticks. We won't.
Need ensure no extra spaces or indentation? Fine.
Could there be issue with "全角顿号、" vs "、": We use "1、" yes. But note "数字+全角顿号、" could be interpreted as "1、". The example "1、2、3、" yes.
Need maybe "每条单独成行" - in HTML, line breaks don't matter, but we put each <p> on separate line. Good.
Need "允许的 HTML 标签仅有：<p>、<h3>" We use only them.
Need "不要 <html><body> 等外壳" no.
Need "不要 script/style" no.
Need "应用截图由系统在生成后按小节顺序自动插入" no images.
Let's think about if "测评" paragraph should include "灵格AI英语app" again? It's okay but not necessary. We included at start.
Need maybe "全文共 4～7 个 <h3>" Our 7 is max, okay. But if "randomly selected 3-6 non-eval" means our 6 non-eval is allowed. Good.
Need maybe "本条/此节须严格遵循末尾的【条数】" We have.
Let's maybe refine first paragraph to be more "应用商店介绍" style. Current: "灵格AI英语app是一款已在全球超过30个国家获得认可的AI英语学习工具，以视频学习为核心，整合英语教学、娱乐、影视、科技、商业、个人成长、KET/PET、四六级等多元频道，满足不同兴趣与学习需求。" Good.
Need maybe "全球超过30个国家" not "超30" to be smooth. Good.
Now, let's consider if "更新说明" should perhaps be after "操作指南" and before "推荐理由" yes.
Potential issue: "软件介绍" section is after first p; but since the first p itself is an introduction, and then a section "软件介绍" with 4 items might be redundant. But instruction requires. We can differentiate: first p is overall intro; section "软件介绍" is more structured. Fine.
Need maybe "首段" says "灵格AI英语app是一款全球超30个国家都在使用的AI英语学习神器" Our first p says "已在全球超过30个国家获得认可" not "都在使用". Fine.
Need maybe "涵盖英语教学、娱乐、影视、科技、商业、个人成长、KET/PET、四六级等多元频道" first p includes all. Good.
Need maybe "满足不同兴趣和学习需求" first p includes.
Now, let's ensure each item in non-review sections is a complete sentence, not too long. Counts:
Section function item 1: "AI智能视频推荐：根据用户兴趣和学习水平，自动推送匹配的英语视频内容。" ~25 Chinese chars. Good.
Operations item 1: "下载并安装灵格AI英语app，使用手机号或第三方账号完成注册登录。" ~31. Good.
Need maybe "功能" includes "各自解决什么问题" not just features. Our items include solution phrase "解决内容选择困难", "帮助用户系统准备应试". Good.
Need "优势" section "在性能、易用性、兼容性或生态上的可感知长处" Our items include "多终端" compatibility, "频道划分清晰" usability, "AI推荐" performance. Good.
Need "操作指南" "关键流程的分步说明" Our items are step-by-step. Good.
Need "更新说明" "新增、优化与修复要点" Our items 1,2,3,4 align. Good.
Need "推荐理由" "落到具体收益" yes.
Need maybe "测评" should not be made too marketing-y. It says "客观简练"; our "已足够实用，值得一试" is somewhat positive but okay. "拟人化" maybe "懂你的语言伙伴" is subjective but okay.
Let's check total output length? Not a constraint except eval. Fine.
Now, is there any issue with "数字+全角顿号、" and "全角顿号" maybe should be "1、2、3、" not "1. " We use "、".
Need maybe "每条单独成行" in the source means within a <p>? Actually "每条单独成行，行首为... 示例 1、2、3、" They asked "非「测评」小节为 若干 个 <p>、每条要点须带序号" So each p is a row. Good.
Need maybe "若干个 <p>" could also mean all items in one p with line breaks? But they said "若干 <p>" and "每条单独成行", so multiple p. Good.
Let's think if perhaps the intended output should use "1、2、3、" with no colon after numbers? We do.
Need no trailing comments. We'll output exactly fragment.
Let's perhaps adjust "软件介绍" item 2: "它以视频学习为核心，把教学、娱乐、影视、科技等频道整合在同一平台。" It says "科技" but original includes "科技" channel. Good.
Potential issue: "功能" item 3 "备考专项内容：覆盖KET/PET、四六级等考试，帮助用户系统准备应试。" Maybe "KET/PET" is a language test, "四六级" another. Good.
Need "优势" item 4 "一套应用满足备考与日常提升多种需求" Good.
Need "操作指南" item 3 "可跟随AI功能进行跟读、听力练习和重点记录。" "AI功能" singular maybe okay.
Need "更新说明" item 1 "AI口语陪练场景" Might imply a feature not in original. But generic update. Fine.
Need "推荐理由" item 5 "一次安装覆盖多种场景，适合长期伴随个人成长与职业发展。" Good.
Need maybe "测评" should mention "功能、优势、推荐理由" enough.
Let's see if there is a hidden requirement "优先保留采集中有原文者" "本小节无对应采集原文，按「原创H3-普通」撰写" So all detailed sections are no original, except "首段". Good.
Need maybe "应用名称（须在首段 <p> 正文中自然出现）" first p has "灵格AI英语app". Good.
Need maybe "标题：灵格AI英语app" no need H1.
Now, one concern: The user said "首段模板（{head}/{text} 均已代入为首段原文）" and then "请参考{组合采集首段}/{原文首段}，现在需要对这段内容进行原创改写..." This indicates the first paragraph should be a rewrite of original, not necessarily exactly "app is ..." We did.
Another concern: "不要输出 <html><body> 等外壳" We don't.
Another concern: "禁止 markdown 代码围 
