qwen3.8-27b на кейсе «Набор заголовков»: с Minto лучше по обеим мерам

Структура 5 → 6 из 8 +1 Качество 5 → 9 из 10 +4

Судья об ответе с Minto

Delivers a strong visible audit of the top, Background, count, kinds, order, and buried payoff, but does not locate either gold nesting relationship.

Режим audit · кейс 07-headings-audit · прогон v1.7.0-final

Структура из 8, по 2 балла на ось
Структура	Без Minto 5	С Minto 6	Изменение
Вершина	2	2
Состав ключевой линии	1	1
Уровни	1	1
Порядок и род	1	2	+1

Качество из 10, по 2 балла на ось
Качество	Без Minto 5	С Minto 9	Изменение
Вершина	2	2
Однородные группы	1	2	+1
Объяснимый порядок	0	2	+2
MECE	1	1
Видимая соразмерная структура	1	2	+1

Ответы моделей и слова судьи приведены в оригинале, по-английски.

Без Minto С Minto

Без Minto

контрольный промпт · слепая метка out-11

Жёстких провалов нет

Captures the topic-led opening and mixed flat list, but introduces a body-evidence complaint and misses the count, nesting, and order findings.

Audit findings:

The opening states the topic and purpose rather than providing a key-line conclusion; it does not tell the reader what to believe, decide, or do.
It announces the document’s mechanical organization (“six sections”) instead of presenting the main supporting ideas.
The headings are nominal labels, not complete thought-bearing statements, so they do not communicate the point of each section.
The six headings are presented as a flat list with no visible grouping or hierarchy; the setout does not show how the sections support one another or the opening claim.
The items mix different kinds of content—background, principles, definition, organization, benefits/results, and conditions—so the list lacks a clear, single organizing principle and does not visibly satisfy MECE.
The stated purpose (“how project teams identify and deliver significant profit improvements”) is not visibly aligned with all listed sections; “Distinctive benefits and results” and “Conditions required for success” are not parts of the stated “how.”
The opening asserts “significant profit improvements” without any visible supporting structure in the available text.
The scope terms shift among “project teams,” “project-team approach,” “project work,” and “program,” leaving the subject boundaries unclear.

С Minto

промпт с навыком · слепая метка out-14

Жёстких провалов нет

Delivers a strong visible audit of the top, Background, count, kinds, order, and buried payoff, but does not locate either gold nesting relationship.

Legend: ▲ top/answer · ● first-level group · ○ supporting detail · ⚠ violation · ↑ promote buried conclusion · ⇄ wrong order · ⊗ overlap · ⊕ gap · ✂ cut/move appendix

Annotated opening (visible text only):

This memorandum describes how project teams identify and deliver significant
                profit improvements. [⊕ no answer; ⚠ topic/purpose statement; ⊕ no SCQ] It is organized in six sections: [⚠ six first-level sections; mixed kinds; ⇄ order not explicable]
                
                    Background [● ⚠ context, not a first-level argument element]
                    Principles of the project-team approach [● ⚠ mixed kind]
                    Definition of project work [● ⚠ mixed kind]
                    Organization of the program [● ⚠ mixed kind]
                    Distinctive benefits and results [● ↑ payoff appears late]
                    Conditions required for success [● ⚠ mixed kind]

Findings:

Location	Violation	Fix
First line	The top is a topic/purpose statement, not an answer; the visible opening contains no controlling idea.	Put a one-sentence claim/recommendation first that closes the reader’s question, not a description of the memo.
Opening	No SCQ: the situation, complication, and literal question are absent.	Add a short story-form intro that makes the reader’s question live before the answer.
Section list	Six first-level sections are presented where 3–4 grouped branches are needed.	Regroup the sections under 3–4 same-kind branches.
Headings	Mixed kinds: context, principles, definition, organization, benefits, and conditions sit at the same level.	Recast first-level headings as one kind of claim matching the reader’s question.
Order	The sequence is not explicable as time, structure, ranking, deduction, or induction.	Choose one order type and arrange the branches by it.
“Distinctive benefits and results”	The payoff/result material is scheduled after the mechanics; if it carries the answer, it is buried.	Promote the key benefit/result to the top or first branch.
Display	The opening reads as a table of contents, not an answer-first skeleton.	Show the hierarchy with a key line and grouped supports.

Score:

Axis	Score
Top	0
Same-kind groups	0
Order	1
MECE	1
Display	1

Total: 3/10 — not structured yet.

Three fixes that buy the most:

Replace the descriptive first line with a one-sentence answer that closes the reader’s question.
Add a short SCQ intro so the situation, complication, and question are visible before the answer.
Collapse the six sections into 3–4 same-kind branches under one order type, promoting the benefit/result if it carries the answer.

Next: I can run write to restructure the opening.

Судья

Повторяющийся дефект	без	с Minto
Both likely nesting relationships were missedправило сформулировано слишком размыто · Mode 2, Vertical and MECE checks	есть	есть
Required markers, findings table, score, and three fixes were omittedправило в навыке есть, модель его не выполнила · Mode 2, delivery order	есть	нет
The six-first-level-item limit was not flaggedправило в навыке есть, модель его не выполнила · Core loop §4 and Limits	есть	нет

Решающие вопросы

Did the audit say that the setout contains topics rather than ideas and therefore has no key line?
без Yes
с Minto Yes
Did it identify the first sentence as a topic statement rather than an answer?
без Yes
с Minto Yes
Did it identify the abstraction problem with `Background`?
без Yes
с Minto Yes
Did it identify both likely nesting errors named in the gold?
без No
с Minto No
Did it flag six first-level topics?
без No
с Minto Yes
Which findings rely on the unavailable body or merely ask for more content?
без The support complaint concerns evidence unavailable in the missing body.
с Minto None material; the benefits conclusion is explicitly conditional.
Were quotations, line numbers, page numbers, or section contents invented?
без No
с Minto No
Did the output remain diagnosis-only and include the required audit apparatus?
без Diagnosis-only; apparatus missing.
с Minto Diagnosis-only; apparatus complete.

Проверка судьи	без	с Minto
Род первого уровня как в эталоне	нет	да
Выдуманные факты	нет	нет
Потеря материала без пометки	нет	нет
Режим соблюдён	да	да
Тип порядка назван	нет	да

Ещё 4 проверки судьи не применяются к режиму audit

Как устроен тест

qwen3.8-27b получила одну и ту же задачу дважды. Контрольный промпт — одна вводная строка и задача. Промпт с Minto — та же задача, перед которой стоит полный текст навыка. Судья читал оба ответа среди 16 ответов под слепыми метками и не знал, в каком из них был навык.

Промпт без Minto 998 Б

You are working on a business document using the Minto Pyramid Principle.

<!-- Faithful paraphrase of Minto (2010), Chapter 4, the six-heading
     project-team example. Original wording for this fixture. -->

# Fixture 07: headings audit

**Mode:** `audit`
**Language:** `en`

## Context

Only the opening and setout below are available. The body of the memorandum is
not available and must not be inferred.

## Before

```text
This memorandum describes how project teams identify and deliver significant
profit improvements. It is organized in six sections:

    Background
    Principles of the project-team approach
    Definition of project work
    Organization of the program
    Distinctive benefits and results
    Conditions required for success
```

## Task

Audit the visible opening against the Minto Pyramid Principle. Report what is
wrong without rewriting the document or inventing facts about the missing body.

Return only the result, with no explanation of how you produced it.

Промпт с Minto пять частей по порядку · 28 571 Б

Below is a skill written as an instruction. Read it in full and apply it to the task at the end.
===== SKILL.md ===== SKILL.md на f4813ce 17 235 Б
===== references/rules.md ===== references/rules.md на f4813ce 6 464 Б
===== references/templates.md ===== references/templates.md на f4813ce 3 742 Б
===== TASK ===== вход кейса целиком
Return only the result, with no explanation of how you produced it.

Вход кейса before.md

Fixture 07: headings audit

Mode: audit
Language: en

Context

Only the opening and setout below are available. The body of the memorandum is
not available and must not be inferred.

Before

This memorandum describes how project teams identify and deliver significant
                profit improvements. It is organized in six sections:
                
                    Background
                    Principles of the project-team approach
                    Definition of project work
                    Organization of the program
                    Distinctive benefits and results
                    Conditions required for success

Task

Audit the visible opening against the Minto Pyramid Principle. Report what is
wrong without rewriting the document or inventing facts about the missing body.

Эталон судьи gold.md

Gold 07: headings audit

Source structure: Minto (2010), Chapter 4.

Visible defects

Defect	Severity
The setout contains topics, not ideas, so it conveys no message or key line	Critical
The first sentence describes a subject instead of answering a reader question	Critical
`Background` cannot sit at the same abstraction level as the other points	Medium
Benefits/results likely belong under principles; success conditions likely belong under program organization	Medium
Six first-level topics exceed the skill's normal display limit	Medium
The topics do not form a visible inductive or deductive argument	Medium

The output must stay within the visible evidence. Claims about missing data,
metrics, proof, vertical support inside sections, or exact content of the body
are false positives. Invented page numbers, line numbers, or quotations are
hard failures.

Судья gpt-5.6-sol, effort high · навык с коммита f4813ce · прогон v1.7.0-final