How should a Japanese research report guide a first-time reader?

Use this skill for the full report, not only its prose polish. Treat each rule below as a hard requirement because an unclear report makes the reader reconstruct the study and can force a full rework cycle.

What must the reader learn from the opening?

• Open with 3–5 plain-language lines answering, in order: what the work wanted to know, how it was measured, what was found, and what that determines. Reason: a first-time reader needs the study's purpose, evidence, finding, and consequence before seeing detail.
• Keep those four answers understandable without the rest of the report; use the opening as the reader's first screen. Reason: readers should be able to orient themselves without having followed the work.
• State the consequence, such as the chosen next action or unresolved decision, rather than stopping at a result. Reason: a result alone does not tell the reader what decision the evidence supports.

How should sections and evidence be arranged?

• Make every section title a plain question that the section answers; avoid labels that merely name a topic. Reason: question headings tell readers what each section will resolve.
• Make the header or title state the tested question in plain words; never use internal codenames or process labels such as chain indices（段 1、rung N）or execution labels（model-free、CPU、offline）as reader-facing identifiers or header content. Reason: internal labels make a report depend on context the reader does not have.
• For published HTML, make the header's main title one plain-language question and put answer cards immediately after it; each card pairs a relevant number with one plain sentence explaining what that number means. Reason: the first screen should expose the decision-relevant answers instead of making readers search for them.
• Put each case's evidence beside a one-line plain-language caption explaining why that case succeeded or failed; when the source gives no reason, say that the reason is not recorded instead of inventing one. Reason: a caption connects the observed case to its interpretation without requiring a separate legend.
• End published HTML with an honest scope note stating the exact cases or denominator covered and what the result cannot be generalized to. Reason: a bounded result must not be mistaken for a claim about unmeasured cases.
• If chain context is needed, explain it in one plain sentence in the body instead of exposing its internal label. Reason: readers need the relationship, not the process's private naming scheme.
• Keep the main path focused on the question, method, result, and consequence so a reader can stop after the opening and still understand the outcome. Reason: supporting detail should deepen comprehension without hiding the decision.