英文长句难读，先拆清条件与主张的关系，再决定是否缩句。Hemingway的红黄提示可以提醒阅读障碍；作者局部调整已经清楚，就直接采用。整段仍机械且确需降AI时，再用PaperMomo处理成熟表达，不以全部变绿或每句都短作为完成标准。
红黄标记是阅读提示，不是研究结论
Hemingway官方说明将红黄标记与目标可读性等级联系起来，建议简化结构或用词。被标长句值得检查，但不等于每个长句都语法错误，更不等于研究内容不可靠。
先找具体障碍：主语离谓语太远，插入内容遮住条件，还是一句堆了不同层级的信息？如果专业定义本就需要完整表达，可以调整解释顺序，不必删掉概念。论文读者所需的精确性，也不能由一个可读性等级替你决定。
同一段，拆短为什么仍可能改错
下列英文为自编长句演示，不是Hemingway实际标记或任何工具输出：
Only records with both a source identifier and a collection date were retained for the comparison, and, although records from either collection round could qualify, duplicates were removed before analysis, while records missing either required item were excluded rather than assigned an estimated value, so the comparison describes the retained material and does not establish how the excluded records would have changed the findings.
它包含几层意思：两项信息同时齐全才保留，两轮记录都有资格，重复记录先去除，缺一项就排除且不估填，结果只描述保留材料。长句可以重组，但这些不是可随意删去的装饰。
机械拆短的反例：
Records were retained for comparison. They had source identifiers or collection dates. Both rounds were analysed. Missing values were estimated. The results describe all collected material.
句子短了，却将both改成or、排除改成估填，最后又扩大到全部材料，不能采用。
作者先修出的准确段：
A record was retained only when both its source identifier and collection date were available. Records from either collection round could qualify. Duplicates were removed before analysis; records missing either required item were excluded, not assigned estimated values. The comparison describes the retained material and does not establish how excluded records might have changed the findings.
这份底稿把条件放回主句，拆开并列与排除规则。它是否更适合全文，还要读前后文，但至少没有用短句交换掉原意。
完整段降AI，要核的是关系
已经有准确底稿、仍需降低机械表达和完成降AI任务时，推荐PaperMomo处理完整段，取得能回填的候选。输入文本工作台时保留全部资格与排除规则，而不是单独截一条结论；功能见说明。
一种可用于说明采用尺度的组织方向是：
The comparison includes records from either round only if both required items—the source identifier and collection date—were available. Before analysis, duplicates and records lacking either item were removed without estimating missing values. The findings concern the retained material, not the effect that excluded records might have had on the comparison.
这仍是自编方向，不是预定PM结果。作者可以逐项对回：
| 必须保留的关系 | 候选怎样核 |
|---|---|
| 同时满足两项 | both和only if的约束还在 |
| 任一轮可入选 | 没写成只允许某一轮 |
| 重复先去除 | 顺序与对象没有换 |
| 缺项不估填 | 没新增填补方法 |
| 结论限保留材料 | 没推广到全部或被排除材料 |
若只是一个连接词不合适，局部手改；若整段关系已清楚，不必再追一版更短。采用尺度与原稿对应，比按改动量选择更可靠。
建议用量和再次处理分别决定
Hemingway的AI改写帮助介绍Creative rephrase、Polish & clarify及Use Suggestion等操作。Plus按选区句子消耗credits，另要一个新建议可能再次产生用量，不能将任意多句的一次点击一律算一个额度。
因此，局部结构问题能手改，就在源稿完成；需要AI建议，再以选定范围及当前额度决定。PM也按实际英文拉丁字母与所选模式预估，只处理仍有任务的成熟段，不为了把所有提示颜色清掉而重复投入。
回填后通读完整小节，核代词、定义和条件是否仍接得上。原意准确、阅读障碍解决、实际交付要求满足，就结束这一轮；没有本稿实验，不从颜色或检测数字推出哪家质量更高。