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。
我们需要输出一个可直接写入数据库的 HTML 正文片段。要求严格。我们必须遵守：没有 markdown 围栏，没有解释。允许的 HTML 标签仅有 <p> 和 <h3>。不能有 <img>，不能有 html/body/script/style。应用名称必须在第一个 <p> 中自然出现。整体结构：第一个元素是一个 <p> 基于首段改写。然后 h3 小节顺序：更新说明 → 推荐理由 → 功能 → 操作指南 → 使用说明 → 特色 → 测评。但是用户说“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”。然而在“小节顺序”中列出了全部 7 个小节？ Wait user says “【小节顺序】更新说明 → 推荐理由 → 功能 → 操作指南 → 使用说明 → 特色 → 测评” and then provides tasks for each. But in overall structure 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「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>）”. It might mean we need to randomly select 3-6 non-review sections? But then in the template each non-review section says “本小节固定 6 条” etc. It seems contradictory. Let’s parse carefully.
The prompt: "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+全角句号．」（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）。"
This says: "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；...）" There is a fixed list of h3 titles for all possible subsections? The "小节顺序" is the ordering if included. Then "全文共 4～7 个 <h3>" indicates we should not output all 7? But then they provided detailed instructions for each of the 7 sections. Wait, the sections listed in "小节顺序" include exactly 7: 更新说明, 推荐理由, 功能, 操作指南, 使用说明, 特色, 测评 (last). They say "已从词池随机抽 3～6 个非测评小节" — maybe the system in a previous step selected 3-6 non-review sections. But here the prompt itself contains "各小节任务与已代入模板" for all 7. Maybe we need to output all sections shown? But the overall says 4-7 h3: If we output all 7, that is 7 h3 (including 测评), okay 4-7? It's exactly 7. But "随机抽 3～6 个非测评小节" would produce 4-7 total (3-6 non-review + 1 review). If we output all 6 non-review + 1 review = 7, which is within 4-7. But "随机抽" implies not all? Yet the instruction says "仅按下列「小节顺序」输出" and the entire list is included. The "小节顺序" may be the full ordered list of all possible sections, but the model should choose 3-6 non-review? However no explicit instruction to omit any; and they provide "【条数】" for each. But "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" could be a note that the list was already filtered? Wait "下列「小节顺序」" includes all sections, not a filtered list. Hmm.
Let's look at first sentence: "应用名称（须在首段 <p> 正文中自然出现）：工会干教网app" Then "【整体结构】顺序必须严格如下：" The "顺序必须严格如下" might refer to the order of the list below: 更新说明 → 推荐理由 → 功能 → 操作指南 → 使用说明 → 特色 → 测评. It doesn't say "all must be included"; it says "顺序必须严格如下" for the sections selected. But then "仅按下列「小节顺序」输出" maybe "输出" all sections? The parenthetical says "已从词池随机抽 3～6 个非测评小节" — that seems like a meta instruction to the generator: from the pool, random selection has already been made. But if in the prompt all seven templates are included, maybe the selection is "all non-review sections" due to "优先保留采集中有原文者"? There is no collection original for any; but the template says "本小节无对应采集原文，按「原创H3-普通」撰写". So no original.
Need to decide. The safest is to output all seven sections? Let's see constraints: "全文共 4～7 个 <h3>" — if we output 7, it satisfies "4~7". If output fewer, despite instructions for each, we might omit required sections. But "已从词池随机抽 3～6 个非测评小节" says random selection from word pool; here the "词池" may have included all six non-review sections? The system likely generated this prompt with all sections? Actually the prompt includes sections: 更新说明, 推荐理由, 功能, 操作指南, 使用说明, 特色, 测评. It doesn't include 介绍/简介, so no special first h3. It says "优先保留采集中有原文者"; none have original. It says "其余小节再按其下列顺序" — perhaps the selected sections are those in the "【小节顺序】" list? In this case list includes six non-review plus 测评, so output all six non-review plus 测评. That equals 7 h3, within 4-7. It might be what they expect: "每个小节先 <h3>..." and "每节要点条数以该小节下方【条数】为准" for all given sections. The "随机抽" is generic boilerplate; we can ignore because the concrete list is full.
But "全文共 4～7 个 <h3>" allows 7. If we output all seven, no violation. Also "非「测评」小节为 若干 个 <p>" and "每节要点条数" defines all. I think output all seven sections, each with required number of <p> items: 更新说明 6, 推荐理由 6, 功能 6, 操作指南 5, 使用说明 3, 特色 5, 测评 1. That gives 6+6+6+5+3+5+1=32 <p> plus 7 <h3>. That's large but acceptable.
Need ensure first element is a <p> containing app name and original rewrite. Then h3 "工会干教网app更新说明", etc. Each non-review section: each <p> contains one numbered item using full-width dot "．" after number. The example "1．" with Chinese full stop. Need use same format throughout: "1．", "2．", "3．", etc. We'll number reset per section? The instruction says "每条单独成行，行首为「数字+全角句号．」（示例 1．2．3．），全篇保持同一格式". "each section's points counted from 1? It doesn't explicitly say reset per section, but usually in such content, each H3 subsection starts from 1. "每节要点条数" and examples. Need decide. They say "全篇保持同一格式" meaning numbering style, not continuous numbering. It likely means each section uses 1,2,3... We can reset per section to make sense. But "全篇统一为同一种序号格式" not "全篇连续编号". I'll reset per section because each section independent.
Need content original, not too generic. Need include app name in first paragraph. First paragraph based on original: "工会干教网app是全国工会干部教育培训网的官方移动端平台，由全总组织部主管、中国工人出版社运营，是专为工会干部打造的一站式在线学习工具。工会干教网app集学习、考核、组织管理功能于一体，提供思想政治、理论素养、法律法规、实务技能等多元化课程资源。" Need rewrite as app store intro, logical clear. We can write: "工会干教网app是由全国工会干部教育培训网推出的官方移动端学习平台，由全总组织部主管、中国工人出版社运营，专注于为工会干部提供一站式在线学习服务。应用整合学习、考核与组织管理功能，覆盖思想政治、理论素养、法律法规、实务技能等多元课程，助力干部高效提升综合能力。" That's fine. "应用名称须在首段 <p> 正文中自然出现" yes.
Need h3 titles exactly:
<h3>工会干教网app更新说明</h3>
<h3>工会干教网app推荐理由</h3>
<h3>工会干教网app功能</h3>
<h3>工会干教网app操作指南</h3>
<h3>工会干教网app使用说明</h3>
<h3>工会干教网app特色</h3>
<h3>工会干教网app测评</h3>
Need content for each:
Update notes (6 items, each ~50 chars): Since no original, generic based on "本版本新增、优化与修复要点". Need mention app improvements, not version log. Could be:
1．优化课程加载速度，视频播放更流畅，减少学习过程中的等待与卡顿。
2．新增学习进度提醒功能，帮助用户及时掌握未完成课程与考核任务。
3．完善组织管理模块，管理员可更便捷地查看成员学习情况和统计数据。
4．更新课程资源库，增加法律法规与实务技能等最新专题课程内容。
5．修复已知问题并提升系统稳定性，兼容更多主流移动设备。
6．改进个人中心界面，学习记录与证书查询入口更加清晰直观。
Need ensure each item "one sentence" ~50 chars. Good.
Recommendation reasons (6 items): Why install:
1．官方平台权威可靠，课程内容由全总组织部主管，学习数据更有公信力。
2．课程体系覆盖思政、理论、法规、实务，满足工会干部多样化学习需求。
3．学习、考核、管理一体化设计，减少切换成本，提升培训组织效率。
4．支持随时随地在线学习，碎片时间也能完成继续教育与能力提升。
5．个人学习进度自动同步，便于制定计划并追踪培训完成情况。
6．面向基层和各级工会干部免费开放，是提升业务能力的实用工具。
Need "具体收益" yes.
Features (6 items): Main capability modules:
1．在线课程学习模块提供视频、图文等多种形式内容，支持按需选学。
2．考核测评模块可在线完成练习与考试，帮助检验理论学习成果。
3．组织管理模块支持管理员创建培训班、分配任务并查看学习进度。
4．课程分类清晰，涵盖思想政治、理论素养、法律法规、实务技能等。
5．学习记录自动归档，支持查看历史成绩、证书与培训档案。
6．个人中心集成消息通知与待办提醒，重要学习任务不再遗漏。
Operation guide (5 items): Key processes step by step:
1．下载安装工会干教网app后，使用工会干部账号或手机号完成注册登录。
2．进入“课程中心”可按分类浏览课程，选择目标课程后点击“加入学习”。
3．在“我的学习”中可查看进行中课程，点击课程卡片即可继续视频学习。
4．完成课程学习后，进入“考核测评”模块参加在线考试并提交答卷。
5．管理员可在“组织管理”中创建班级、添加成员并导出学习统计数据。
Need "关键流程的分步说明". Good.
Usage instructions (3 items): Installation, permissions, account basic:
1．应用支持Android和iOS主流系统，在官方应用商店搜索工会干教网app下载安装。
2．首次使用需授予网络等必要权限，并同意用户协议方可正常访问学习内容。
3．账号需与全国工会干部教育培训网实名信息一致，登录后可同步学习记录。
Features/特色 (5 items): Distinctive:
1．依托全国工会干部教育培训网官方体系，课程资源权威且持续更新。
2．学习、考核与组织管理三端融合，打造工会培训工作闭环。
3．课程内容紧贴工会实际，涵盖维权服务、民主管理等实务案例。
4．支持按岗位和层级推荐课程，实现个性化学习路径规划。
5．学习数据与培训档案贯通联动，便于组织掌握干部培养情况。
Need avoid too many "特色" similar to features. Need differentiate. Good.
测评 (1 paragraph, ≤200 chars, no numbering, objective, personified). Need "拟人化叙述方式" and objective. Example: "工会干教网app给我的整体印象是“专业且省心”。作为工会干部专属的学习平台，它把课程学习、考核测评和组织管理安排得井井有条，课程方向也贴近实际工作。虽然界面风格偏稳重，但胜在功能实用、数据同步顺畅，适合长期使用。对需要系统提升业务能力的工会干部来说，是一个值得信赖的随身课堂。" Count chars? Let's count roughly. Need <=200 including punctuation. Let's count carefully. Chinese characters and punctuation all count. "工会干教网app给我的整体印象是“专业且省心”。" Let's count: 工(1)会2干3教4网5a6p7p8给9我10的11整12体13印14象15是16“17专18业19且20省21心22”23。24. Actually "app" 3 chars, so 24. Next "作为工会干部专属的学习平台，它把课程学习、考核测评和组织管理安排得井井有条，课程方向也贴近实际工作。" Count: 作1为2工3会4干5部6专7属8的9学10习11平12台13，14它15把16课17程18学19习20、21考22核23测24评25和26组27织28管29理30安31排32得33井34井35有36条37，38课39程40方41向42也43贴44近45实46际47工48作49。50. Next "虽然界面风格偏稳重，但胜在功能实用、数据同步顺畅，适合长期使用。" Count: 虽51然52界53面54风55格56偏57稳58重59，60但61胜62在63功64能65实66用67、68数69据70同71步72顺73畅74，75适76合77长78期79使80用81。82. Next "对需要系统提升业务能力的工会干部来说，是一个值得信赖的随身课堂。" Count: 对83需84要85系86统87提88升89业90务91能92力93的94工95会96干97部98来99说100，101是102一103个104值105得106信107赖108的109随110身111课112堂113。114. Total ~114 incl punctuation. Good.
First paragraph maybe "工会干教网app是全国工会干部教育培训网的官方移动端平台，由全总组织部主管、中国工人出版社运营，专注于为工会干部打造一站式在线学习工具。它集学习、考核与组织管理于一体，提供思想政治、理论素养、法律法规、实务技能等多元课程资源，帮助用户随时随地高效提升业务能力。" Count fine. Need "原文首段" includes "提供思想政治..." and "集..." Good.
Need ensure no extra text, no markdown. Just start with <p> and end with </p> after evaluation. Need not include blank lines? HTML fragment can have newlines. Fine.
Potential issue: "禁止输出 <img>（应用截图由系统在生成后按小节顺序自动插入，模型无需处理图片）。" So no img. OK.
Need "全文共 4～7 个 <h3>" We have 7 h3. That's okay. But is "总字数" for evaluation only? "全文不超过 200 字" refers to the evaluation paragraph only, yes.
Need perhaps not include "更新说明" if "随机抽 3~6" but all included is okay? Let's think again. The "小节顺序" includes "更新说明 → 推荐理由 → 功能 → 操作指南 → 使用说明 → 特色 → 测评" and the prompt says "顺序必须严格如下". It then says "然后仅按下列「小节顺序」输出". The parenthetical says "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>". This parenthetical may be part of template instructions from a larger system, and "下列「小节顺序」" is the full list after random selection? But if the list in the prompt includes 6 non-review sections, then random selection was 6? They say "随机抽 3～6" can include 6. If it includes all 6, total 7. So no violation. If they expected a subset, they'd not provide detailed tasks for all sections. Better output all to satisfy "顺序必须严格如下" and "每个小节先 h3" for each provided. Since all tasks are provided, output all.
Need "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）=[操作指南 功能]" This might mean for those two sections, use deeper rewrite style? No special instruction. We can write natural.
Need maintain "style" not markdown. Use Chinese punctuation. Need "序号分条：每条单独成行，行首为「数字+全角句号．」" We'll use exactly "1．" in each <p>. Ensure no extra spaces after punctuation? We can put a space? Better no extra spaces. Each <p> content starts with "1．..." etc. Since p has block display, "单独成行" satisfied. Need not include line breaks within p. Good.
Need "每节要点条数以该小节下方【条数】为准" We'll match.
Let's draft full output.
First p:
<p>工会干教网app是全国工会干部教育培训网推出的官方移动端学习平台，由全总组织部主管、中国工人出版社运营，专注为工会干部提供一站式在线学习服务。应用集课程学习、考核测评与组织管理功能于一体，覆盖思想政治、理论素养、法律法规、实务技能等多元化课程资源，助力用户随时随地高效提升能力。</p>
Then h3 and p.
Need maybe "由全总组织部主管" - "全总" is okay. "中国工人出版社运营" okay.
Update:
<h3>工会干教网app更新说明</h3>
<p>1．优化课程视频加载与播放流畅度，减少在线学习时的等待与卡顿。</p>
<p>2．新增学习进度提醒功能，及时推送未完成课程与考试待办事项。</p>
<p>3．完善组织管理模块，管理员可更便捷地查看成员学习数据和统计报表。</p>
<p>4．扩充课程资源库，增加法律法规、实务技能等最新专题内容。</p>
<p>5．修复已知问题并提升系统稳定性，更好兼容主流手机设备。</p>
<p>6．改进个人中心界面，学习记录、证书查询等入口更加清晰直观。</p>
Recommend:
<h3>工会干教网app推荐理由</h3>
<p>1．作为官方移动学习平台，课程资源权威可靠，学习结果认证更有公信力。</p>
<p>2．课程覆盖思想政治、法律法规、实务技能等方向，满足工会干部多元提升需求。</p>
<p>3．学习、考核、组织管理集成在一处，省去多平台切换的麻烦。</p>
<p>4．支持碎片化在线学习，方便随时充电，缓解工学矛盾。</p>
<p>5．学习进度自动同步，可清晰追踪培训完成情况，便于查漏补缺。</p>
<p>6．免费提供优质培训内容，是基层工会干部提升业务能力的实用工具。</p>
Need "免费" maybe not in original but okay? "面向工会干部" likely. Should be okay.
Feature:
<h3>工会干教网app功能</h3>
<p>1．在线课程模块支持视频、图文等多样学习形式，用户可按需选学。</p>
<p>2．考核测评模块提供在线练习与考试，帮助巩固和检验学习成果。</p>
<p>3．组织管理模块支持创建培训班、分配学习任务并查看进度。</p>
<p>4．课程中心按思想政治、理论素养、法律法规、实务技能等分类展示。</p>
<p>5．学习记录自动归档，历史成绩、学习时长和证书均可随时查询。</p>
<p>6．个人中心集成消息通知与待办提醒，重要安排不再遗漏。</p>
Operation:
<h3>工会干教网app操作指南</h3>
<p>1．下载安装工会干教网app后，使用已注册的工会干部账号或手机号登录。</p>
<p>2．在课程中心按分类或搜索找到目标课程，点击“加入学习”开始选课。</p>
<p>3．进入“我的学习”可查看进行中课程，点击课程卡片即可继续学习。</p>
<p>4．完成课程后，在考核测评模块参加考试，提交后即可查看成绩。</p>
<p>5．管理员在组织管理模块创建班级、添加成员，并导出学习统计数据。</p>
Need include quotation marks? "加入学习" allowed. HTML no issue. Could use Chinese quotes.
Use:
<h3>工会干教网app使用说明</h3>
<p>1．工会干教网app支持Android和iOS系统，可在官方应用商店下载安装。</p>
<p>2．首次启动时需授予网络等必要权限，并同意用户协议后正常使用。</p>
<p>3．账号需与全国工会干部教育培训网实名信息一致，登录后自动同步学习记录。</p>
Feature:
<h3>工会干教网app特色</h3>
<p>1．依托全国工会干部教育培训网官方体系，课程内容权威且持续更新。</p>
<p>2．学习、考核与组织管理形成闭环，贴合工会培训业务实际场景。</p>
<p>3．课程紧贴工会干部工作需求，包含维权服务、民主管理等实务案例。</p>
<p>4．支持按岗位与层级推荐课程，帮助用户规划个性化学习路径。</p>
<p>5．学习数据与培训档案联动，为组织评价干部培养情况提供参考。</p>
Eval:
<h3>工会干教网app测评</h3>
<p>工会干教网app给我的整体印象是“专业且省心”。作为工会干部专属的学习平台，它把课程学习、考核测评和组织管理安排得井井有条，课程方向也贴近实际工作。虽然界面风格偏稳重，但胜在功能实用、数据同步顺畅，适合长期使用。对需要系统提升业务能力的工会干部来说，这是一个值得信赖的随身课堂。</p>
Need check evaluation total <=200. We counted 114. Good.
Need ensure "全文共 4~7 h3" yes 7.
Potential issue: "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。" We did.
Need "不得输出 <html><body> 等外壳" okay.
Need "不要 script/style" okay.
Need "没有 markdown code fences" okay.
Need "直接写入数据库的 HTML 正文片段" We output HTML fragment. Good.
Need think about "应用名称（须在首段 <p> 正文中自然出现）：工会干教网app" We have "工会干教网app是..." yes.
Need perhaps "app" maybe should be "工会干教网app" exactly with lowercase? We use exact.
Need "全总组织部" maybe "中华全国总工会组织部" would be more formal but original says "全总组织部", okay.
Need "中国工人出版社" original says "中国工人出版社" (not "中国工人出版社运营"? Yes). Good.
Need "全篇保持同一种序号格式" We use "1．" fullwidth period. Good. But note "．" is not Chinese "。" It is "fullwidth full stop" U+FF0E. The instruction says "全角句号．" Example "1．2．3．" — yes U+FF0E. We use that. Ensure no dot after 10? We only go to 6 per section. OK.
Need "每条单独成行" Since p tags each item on separate line. OK.
Need "若干 个 <p>、每条要点须带序号" yes.
Need "每节要点条数以该小节下方【条数】为准" yes.
Need "本小节无对应采集原文，按「原创H3-普通」撰写" We did.
Need "《更新说明》写作侧重：本版本新增、优化与修复要点（概括即可，勿堆砌版本流水账）" Our update note is generic but okay. We didn't mention version numbers. Good.
Need "推荐理由" We mention "免费" and "权威". Good.
Potential issue: "功能" and "特色" might be too similar. But okay.
Need "操作指南" key flow "分步说明" — we have steps.
Need "使用说明" "安装、权限、账号与基础上手要点" yes.
Need "测评" "不超200字" We should verify exactly count to ensure no accidental 200. Let's count evaluation paragraph in detail with punctuation and "app" letters? The instruction says "不超过 200 字（含标点）" Usually "字" includes Chinese characters and punctuation; "app" letters count as characters. Let's count precisely.
Evaluation paragraph:
工会干教网app给我的整体印象是“专业且省心”。作为工会干部专属的学习平台，它把课程学习、考核测评和组织管理安排得井井有条，课程方向也贴近实际工作。虽然界面风格偏稳重，但胜在功能实用、数据同步顺畅，适合长期使用。对需要系统提升业务能力的工会干部来说，这是一个值得信赖的随身课堂。
Let's count characters:
1 工
2 会
3 干
4 教
5 网
6 a
7 p
8 p
9 给
10 我
11 的
12 整
13 体
14 印
15 象
16 是
17 “
18 专
19 业
20 且
21 省
22 心
23 ”
24 。
25 作
26 为
27 工
28 会
29 干
30 部
31 专
32 属
33 的
34 学
35 习
36 平
37 台
38 ，
39 它
40 把
41 课
42 程
43 学
44 习
45 、
46 考
47 核
48 测
49 评
50 和
51 组
52 织
53 管
54 理
55 安
56 排
57 得
58 井
59 井
60 有
61 条
62 ，
63 课
64 程
65 方
66 向
67 也
68 贴
69 近
70 实
71 际
72 工
73 作
74 。
75 虽
76 然
77 界
78 面
79 风
80 格
81 偏
82 稳
83 重
84 ，
85 但
86 胜
87 在
88 功
89 能
90 实
91 用
92 、
93 数
94 据
95 同
96 步
97 顺
98 畅
99 ，
100 适
101 合
102 长
103 期
104 使
105 用
106 。
107 对
108 需
109 要
110 系
111 统
112 提
113 升
114 业
115 务
116 能
117 力
118 的
119 工
120 会
121 干
122 部
123 来
124 说
125 ，
126 这
127 是
128 一
129 个
130 值
131 得
132 信
133 赖
134 的
135 随
136 身
137 课
138 堂
139 。
So 139 characters. Good.
Need perhaps output first paragraph and all sections no "前言后记". Good.
Let's think about "本小节无对应采集原文，按「原创H3-普通」撰写（变量已全部代入；{text} 为首段原文与其它小节摘录）" It says "请根据原文中的内容，针对工会干教网app写一段关于该款游戏/软件的工会干教网app更新说明" They refer "游戏/软件" generic. Fine.
Need "允许的 HTML 标签仅有：<p>、<h3>" We only use those.
Need no "markdown code fences" We won't.
Need "应用截图由系统在生成后按小节顺序自动插入，模型无需处理图片" So no img. OK.
Need perhaps "由全总组织部主管" in first paragraph and "全国工会干部教育培训网" not "全国工会干部教育网"? Original says "全国工会干部教育培训网". We use "全国工会干部教育培训网". Good.
Need "组织管理" vs "组织管理功能" Good.
Need ensure "更新说明" h3 after first p. Yes.
Need maybe "最终" not "最后"? no.
One concern: The prompt says "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+全角句号．」（示例 1．2．3．），全篇保持同一格式）" It might mean no <p> wrapping each item? "每个小节先 <h3>，非测评小节为若干 个 <p>" Yes each p is a point. Good.
Need "不要使用 1、2、3 或分条序号" for evaluation. We use no numbering.
Need "语气客观简练" okay.
Potential issue: "标题必须与下列 {H3title} 完全一致" We have exact.
Potential issue: "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>" None.
Need "更新说明" is before "推荐理由" etc. yes.
Let's consider if "全文共 4～7 个 <h3>" means total h3, not total sections? We have 7 (including evaluation). Good.
Need "推荐理由" section maybe "free" could be unsupported. Original says official platform; it doesn't explicitly say free. But training for cadres likely free. It's okay. But to be safe, we could avoid "免费" if not in original? The prompt says "可参考信息" only original; we can infer. "免费" is a benefit but might be inaccurate. Better to avoid unverified specifics. Let's alter item 6 to not state free unless known. Maybe "依托工会官方培训体系，内容贴近基层实际，是提升履职能力的可靠选择。" But "官方" already item 1. Let's keep item 6 as "面向各级工会干部开放，是持续提升履职能力的综合工具。" This doesn't say free. Good. Let's adjust.
Recommend item 6 originally: "免费提供优质培训内容，是基层工会干部提升业务能力的实用工具。" To avoid unsupported, use "面向各级工会干部开放，是提升履职能力的实用工具。" Good.
Need "推荐理由" item 2 "法律法规" etc. Fine.
Need "更新说明" item 6 "改进个人中心界面，学习记录、证书查询等入口更加清晰直观" good.
Need "使用说明" item 1 "Android和iOS" maybe "Android" and "iOS" are Latin letters; okay. "官方应用商店" general. Good.
Need "特色" item 3 "维权服务、民主管理等实务案例" okay.
Need maybe "组织管理" "管理员" might require role; okay.
Need "操作指南" item 4 "参加考试" maybe "答题" okay.
Need "功能" item 1 "视频、图文" original didn't mention but likely. Fine.
Need "测评" "界面风格偏稳重" maybe subjective but okay.
Need "原创改写" We didn't copy exact original. First paragraph similar but rephrased. Fine.
Need not include "应用名称" in every h3? Titles do. Good.
Need no HTML entity? Fine.
Let's ensure no accidental Markdown "---" or backticks. No.
Potential issue: "行首为「数字+全角句号．」（示例 1．2．3．），全篇保持统一" In our HTML, <p> tags are not "行" strictly? But visually line per p. Good.
Need "每个 p 一条要点" yes.
Need "每条单独成行" yes.
Need "禁止输出 <img>" no.
Let's think about whether to include "更新说明" as h3? Yes.
Now, one more nuance: "整体结构 1) 第一个元素是一个 <p> ... 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；... 全文共 **4～7** 个 <h3>）" If the system expects not all seven, our 7 h3 = 7 "小节" including 1 evaluation, so exactly 4-7? Wait "全文共 4～7 个 <h3>" includes evaluation h3. 7 is maximum. So it's valid. "随机抽 3～6 个非测评小节" — we have 6 non-review sections, which is within 3-6. So it matches. Good.
Need "优先保留采集中有原文者" — no original but we still have all six; perhaps "no original" means should not "优先保留", but still can randomly select 6. Okay.
Potential issue: The "小节顺序" includes "更新说明 → 推荐理由 → 功能 → 操作指南 → 使用说明 → 特色 → 测评". If we include all, order exactly. Good.
Need maybe "推荐理由" after "更新说明" before "功能" yes.
Need "测评" last yes.
Need "全文共 **4～7** 个 <h3>" yes.
Let's consider if "应用名称（须在首段 <p> 正文中自然出现）：工会干教网app" maybe means "必须且仅在首段出现"? It says "应用名称（须在首段 <p> 正文中自然出现）" not necessarily only there, but our h3 titles include name, so it appears many times. That's required for h3 titles. Fine.
Need "可直接写入数据库" no special encoding. OK.
Now let's produce final answer as plain text. Ensure no preceding/trailing text. We'll include newlines between tags for readability. That's acceptable in HTML fragment.
Need perhaps "全角句号．" after number should be "1．" not "1." We use "1．". Good.
Let's now write final.
But let's double-check content for "更新说明" "本版本新增、优化与修复要点" We have:
1 "优化课程视频加载..." = optimization
2 "新增学习进度提醒" = new
3 "完善组织管理模块" = improvement
4 "扩充课程资源库" = new
5 "修复已知问题..." = fix
6 "改进个人中心" = improvement. Good.
Need "每条约50字内" Let's count item lengths roughly. "1．优化课程视频加载与播放流畅度，减少在线学习时的等待与卡顿。" Count: 1 + punctuation + 26? Let's count: "优化课程视频加载与播放流畅度，减少在线学习时的等待与卡顿。" 30 characters? Under 50. Good.
Need "测评" "全篇不超过200字" we counted.
Potential issue: 
