Datasets:
Exxe
/
crossword-construction-logic

Dataset card Data Studio Files
xet
 Community
instruction
stringlengths
50
473
input
stringlengths
11
21.5k
output
stringlengths
475
2.61k
task_type
stringclasses
9 values
language
stringclasses
5 values
quality
stringclasses
3 values
puzzle_style
stringclasses
5 values
grid_size
stringclasses
8 values
difficulty
stringclasses
4 values
source_format
stringclasses
7 values
verified
stringclasses
2 values
id
stringlengths
6
6
Итоговая строка блока дизайна: соберите ценообразование маски воедино. Что дороже: открытая сетка с длинными записями или закрытая с короткими? Возьмите русскую 4×4 как эталон и объясните, почему она заполняется быстро, а трёхбуквенно-тяжёлые маски - нет.
Данные: русская 4×4 (профиль 3-4-4-3 / 4-4-3-3) заполняется решателем за 7 решений; её трёхбуквенная когорта - 90 слов, четырёхбуквенная - 192. Вариант «все трёхбуквенные»: сетка, где каждая запись требует слово из когорты в 90 слов, сужаемой каждой буквой.
Механизм цены: стоимость заполнения определяется не количеством слов в словаре, а глубиной когорт ПО ПАТТЕРНУ после каждой буквы. В 4×4 стартовые когорты богаты (192 четырёхбуквенных), и одна запись быстро создаёт лестницу: ЛУК превратил соседние когорты в 9-16-4 кандидата - решателю хватило семи шагов, три из которых ...
design_grid
ru
good
mini
4x4
advanced
original
yes
cw_234
एक रिपोर्ट अंग्रेज़ी रूढ़ि (न्यूनतम तीन) से इस ग्रिड को जाँचकर 'चार प्रविष्टियाँ बहुत छोटी' का फ़ैसला सुनाती है; दूसरी हिंदी रूढ़ि (न्यूनतम दो) से उसी को 'स्वच्छ' कहती है। दोनों सही हैं, दोनों अधूरी हैं — निरीक्षक की भूमिका निभाइए।
MASK: MASK: ..# ... #.. FILL: FILL: बल# सड़क #काल रिपोर्ट क (अंग्रेज़ी रूढ़ि, न्यूनतम तीन): चार दो-अक्षर प्रविष्टियाँ पाई गईं — मशीन की गिनती short entries: 4। रिपोर्ट ख (हिंदी रूढ़ि, न्यूनतम दो): सब ठीक, ग्रिड स्वच्छ।
दोनों रिपोर्टों की गिनती में कोई झगड़ा नहीं। मशीन ने दोनों बार सच गिना: न्यूनतम-तीन का पैमाना दें तो चार प्रविष्टियाँ (1A बल, 1D बस, 4D कल, 5A काल — चारों दो-अक्षर) पारी खोलती हैं — short entries: 4; न्यूनतम-दो दें तो वही चार अंदर हैं — शून्य शिकायत। झगड़ा गिनती का नहीं, पैमाने का है — और पैमाना प्राकृतिक नियम नहीं, घो...
validate_puzzle
hi
good
mini
3x3
advanced
original
yes
cw_357
एक पहेला-पैकेट में संकेत-टोली एक ग्रिड की है और ग्रिड दूसरे का। तीन दिशाओं की साझा जाँच (संकेत ↔ प्रविष्टि ↔ ग्रिड) चलाइए और बताइए कौन-सा क़दम सबसे पहले घोटाला खोलता है।
ग्रिड (मशीन-सत्यापित): FILL: नल# सड़क #काल साथ भेजी संकेत-टोली: 1A: शारीरिक या मानसिक ताक़त (2) 3A: ज़मीन पर बनी राह (3) 5A: समय-युग (2) 1D: यात्रियों का बड़ा वाहन (2) 2D: पुरुष बच्चा (3) 4D: बीता/आने वाला दिन (2)
पहला क़दम — सबसे मज़बूत: ग्रिड से प्रविष्टियाँ दुबारा निकालिए। पंक्तियाँ: 1A नल, 3A सड़क, 5A काल; स्तंभ: 1D नस, 2D लड़का, 4D कल। यह नक़्शा किसी संकेत की राय नहीं माँगती — खाने पढ़ने भर से बनती है। दूसरा क़दम: संकेतों से जवाब निकालिए। 'ताक़त' = बल; 'राह' = सड़क; 'समय-युग' = काल; 'बड़ा वाहन' = बस; 'पुरुष बच्चा' = लड़का;...
diagnose_failure
hi
good
mini
3x3
advanced
original
yes
cw_353
Разберитесь с этим «готовым» заполнением 4×4: слова настоящие, а сетка не собирается. Проверьте в правильном порядке - сначала совпадение с маской, затем полноту, затем записи - и назовите каждый дефект с координатами.
МАСКА: МАСКА: #... .... .... ...# ЗАПОЛНЕНИЕ: ЗАПОЛНЕНИЕ: #ЛУ. ДЕЛО АТОМ .РОВ WORD LIST: 3-letter: АГА, АИР, АКТ, АРА, АУЛ, БИТ, БОР, БЫК, ВАЛ, ВЕК, ВИД, ВОР, ГАД, ГАЗ, ГАМ, ГОД, ДАР, ДОМ, ДУБ, ДЫМ, ЕДА, ЖАР, ЖИР, ЖУК, ЗАЛ, ЗУБ, ИВА, КИТ, КОМ, КОТ, КУБ, ЛЁД, ЛАК, ЛЕВ, ЛЕС, ЛОБ, ЛОМ, ЛУК, ЛУЧ, МЁД, МАК, МИГ, МИР, МО...
Шаг 1 - совпадение с маской: сверяем каждую клетку с маской. Одна грубая рана: (3,3) - в маске БЛОК, а в заполнении стоит буква В - слово «РОВ» наехало на несуществующую клетку. Любая буква вне маски - автоматический брак. Шаг 2 - полнота: в якобы готовом заполнении две дыры - (0,3) и (3,0): маска открыта, а клетки пу...
diagnose_failure
ru
flawed
mini
4x4
beginner
original
no
cw_249
Two masks are shown. Both are legal. Explain which one a solver would find more enjoyable and why, using: word count, entry-length distribution, block percentage, and the concentration of 3-letter entries. Then recommend which to build for a Monday-difficulty mini.
MASK A: #...# ..... ..... ..... #...# MASK B: ##... ##... ..... ...## ...##
Mask A statistics: words: 10, blocks: 4. Across 3,5,5,5,3; down 3,5,5,5,3. 3-letter entries: 4 of 10. Block percentage: 16 percent. Mask B statistics: words: 10, blocks: 8. Row lengths: 3,3,5,3,3; columns: c0 r2-r4 (3), c1 r2-r4 (3), c2 full (5), c3 full (5), c4 full (5)... wait - check columns for B: blocks at (0,0),...
design_grid
en
good
mini
5x5
intermediate
original
yes
cw_012
A constructor wants a 5x5 mini whose across entries are 4,4,5,4,4 letters and whose down entries are 3,5,5,5,3. Design the mask, show that it is rotationally symmetric, and explain why this shape is friendlier to a small word list than a fully open grid.
Target slot profile: across 4,4,5,4,4; down 3,5,5,5,3. Grid 5x5.
Work from the slot profile backwards. Rows of 4 starting at column 1 mean column 0 is blocked in those rows; rows of 4 starting at column 0 mean column 4 is blocked. Downs of 3 at the edges mean rows 0-1 (or 3-4) are blocked at the outer columns. Rows 0-1 need length 4 starting at c1: blocks at (0,0) and (1,0). Rows 3...
design_grid
en
good
mini
5x5
beginner
crosshare
yes
cw_003
下面是一位「模型构造者」提交的填充轨迹与成品格盘，它宣称每个交叉都检查过了。请重新推导这份证词，找出它究竟错在哪里、错的是哪条规则。
MASK: MASK: .. .. 模型提交的轨迹: 第1步: 1A 落「中文」——「中文」是词，好 第2步: 「中文」以「文」结尾，接龙规则，下方续「文学」 第3步: 「文学」以「学」结尾，再续「国学」 成品格盘: 中文 文学 模型结论: 四条词条全部是词，交叉一致，验证通过。
先重推导，再下结论。成品格盘是两行「中文」「文学」——逐列读出：1D = (0,0)(1,0) 读作「中」「文」＝「中文」；2D = (0,1)(1,1) 读作「文」「学」＝「文学」。机器裁决：duplicate_words——「中文」同时占据 1A 和 1D，「文学」同时占据 2D 和 3A。一份格盘四条词条，实际只有两个词，各出现两次。 错在哪条规则？错在模型把「成语接龙」当成了填充规则。接龙说：下一个词从上一个词的末字起头——它是链式的一维文化游戏；格盘说：纵向词条必须从共享格向下读——它是二维的约束系统。在 2×2 里，把「文学」放在「中文」正下方，得到的不是「续上了」，而是两条纵向词条与两条横向词条字字相同。「接」这个动...
fill_trace
zh
bad
mini
2x2
advanced
original
yes
cw_294
Why forbid duplicate answers? Both placements are legal words - whose business is it if the constructor wants EERIE twice?
(no artifact - pure methodology question)
The solver's business. The duplicate converts one clue into a memory aid for another: a solver who fills one copy gets the second free, making the grid easier than its difficulty was budgeted to be - an accounting error, not a style choice. It also corrupts the answer key (a duplicate line reads as an error even when l...
methodology_qa
en
good
general
na
beginner
original
no
cw_200
हिंदी वर्ग-पहेली के संकेत परंपरा से तीन शैलियों में लिखे जाते हैं: परिभाषा, पर्याय, रिक्त-स्थान। तीनों शैलियों में एक-एक शब्द के संकेत लिखिए और समझाइए कि कौन-सी शैली किस कठिनाई पर बैठती है।
लक्ष्य-शब्द (सब पूल में): जल, दिल, फूल शैलियाँ: परिभाषा ('यह है ...'), पर्याय ('... का दूसरा नाम'), रिक्त-स्थान ('___ की तरह')
परिभाषा-शैली: जल — 'प्यास बुझाने वाला पारदर्शी तरल'। जाँच: जल = वह तरल? हाँ। यह शैली सीधी है — संकेत शब्द की जाति बताती है, उम्मीदवार जाति में से मिलता है। पर जाति चौड़ी हो तो ग्रिड की मदद चाहिए: पानी भी तो प्यास बुझाता है — ग्रिड के साझा खाने फ़ैसला करेंगे। पर्याय-शैली: दिल — 'मन का दूसरा नाम; शायरों की दुकान'। जाँच:...
write_clues
hi
good
general
na
intermediate
original
yes
cw_346
Teach the HOMOPHONE mechanism: the answer sounds like a different word clued in the surface. Parse the worked example and state the two fairness duties unique to sound-clues.
Worked clue: 'Reportedly, to sport some clothes - that's the past plural of "be" (4)' Answer: WERE
The parse: WERE sounds like WEAR. - 'to sport some clothes' -> WEAR (to wear is to sport clothes). - 'Reportedly' -> the HOMOPHONE indicator: what you just heard is not the answer but its sound-alike. - 'the past plural of "be"' -> the DEFINITION: WERE. - The machine: WEAR spoken aloud is indistinguishable from WERE; t...
cryptic_clue
en
good
cryptic
na
beginner
cryptic-guide
yes
cw_150
一つのパズル・パッケージの中に、カギの一式と盤が入っています — ただし カギは目録盤用、盤は転置盤でした。三方向監査 (カギ↔項目↔盤面) を 実行し、どの方向から攻めると取り違えが最も早く露見するかを説明してください。
盤 (機械検証済み — 転置盤): アワ# サカナ #メシ 同梱のカギ: 1A 朝 (2 マス) 3A 海藻の一種、味噌汁の具材 (3 マス) 5A 果物 (2 マス) 1D 穀物、ヒエの仲間 (2 マス) 2D 魚一般 (3 マス) 4D 食事 (2 マス)
第一方向、盤からの再導出。行と列を丁寧に読みます: 1A = ア・ワ = アワ。 3A = サ・カ・ナ = サカナ。5A = メ・シ = メシ。1D = ア・サ = アサ。 2D = ワ・カ・メ = ワカメ。4D = ナ・シ = ナシ。盤の六項目: アワ、 サカナ、メシ、アサ、ワカメ、ナシ。これは転置盤です。 第二方向、カギから答えを復元。「朝」→ アサ。「海藻の一種」→ ワカメ。 「果物」→ ナシ。「ヒエの仲間」→ アワ。「魚一般」→ サカナ。「食事」→ メシ。カギの六答: アサ、ワカメ、ナシ、アワ、サカナ、メシ。語の集合 は盤と同一 — しかし項目番号への割り当てが全部ずれています。カギは 1A にアサを要求しますが、盤の 1...
diagnose_failure
ja
good
mini
3x3
advanced
original
yes
cw_401
填格的手艺最后都落到一条查询上：「以某字开头/收尾的词有哪些」。请完成下面的队列查询表，并用它解释为什么前向索引与后向索引缺一不可。
词池查询（真实数据）: 以 文 收尾的双字词 = ? 以 文 开头的双字词 = ? 以 学 收尾的双字词 = ? 以 学 开头的双字词 = ? 以 工 收尾的双字词 = ? 以 工 开头的双字词 = ? 以 员 开头的双字词 = ?
逐条作答。以「文」收尾：只有「中文」——1 个。以「文」开头：文件、文学 ——2 个。同一字，两端队列一瘦一胖。以「学」收尾：中学、化学、国学、大学、数学、文学、科学——7 个；以「学」开头：学习、学校、学生——3 个。以「工」收尾：人工、员工、民工——3 个；以「工」开头：工人——仅 此一个。以「员」开头：员工——也只有它。 这张表为什么重要？因为交叉会把词条的两端都变成约束。横向词条落词后，它把「末字」灌进共享格；纵向词条从那个格子出发，被约束的是「首字」。反过来，一条纵向词条先落，横向词条就会被「首字」约束。所以队列查询必须双向：只有首字索引，你算不出「文」收尾时的 1 个候选；只有末字索引，你算不出「文」开头时的 2 个。...
fill_grid
zh
good
mini
2x2
intermediate
original
yes
cw_285
Preview the world's clue traditions in one pass: for Russian, Japanese, Hindi and Chinese crosswords, state what the clue layer inherits from the language's structure - length counting, definition style, enumeration - and what stays invariant across all traditions.
The four construction traditions and their units: Russian (Cyrillic letters, scanword culture allows proper nouns and smaller grids), Japanese (kana cells, mora-counted answers, クロスワード puzzle books), Hindi (Devanagari akshars per cell), Chinese (one character per cell, lattice-based interlock).
RUSSIAN: letters as in English, but the scanword tradition (сканворд) leans on proper nouns, encyclopedic trivia and category clues far more than the Anglo deflection ladder; clue style is definitional and permissive. What transfers: the fill discipline is identical - cell geometry, crossing checks, no-duplicate rules....
write_clues
en
good
general
na
advanced
original
no
cw_134
あるモデルが 2×2 の盤を「完璧な小盤」として提出しました: すべての 交差は合法、どの語も実在、と。機械は二組の重複を宣告しました。この「円環盤」が何を教えるか — 重複禁令がなぜ絶対なのかの証明と あわせて説明してください。
MASK: .. .. 提出された盤: イカ カニ モデルの主張: 「四項目すべて実在の語。交差も全部正しい。」 機械の評決: duplicate_words: イカ [1A, 1D], カニ [2D, 3A]。
盤を丁寧に読み直します。行は イカ と カニ。列は — (0,0)(1,0) の イ・カ = イカ、(0,1)(1,1) の カ・ニ = カニ。つまり横の二語と縦の 二語が完全に同一です。機械の評決は二組の重複: イカ が 1A と 1D に、 カニ が 2D と 3A に。モデルの主張は嘘ではありません — 四項目とも 実在の語で、すべての交差 (0,0) のイ、(0,1) のカ、(1,0) のカ、(1,1) のニ が合法。それでもこの盤は不成立です。 なぜか。交差の価値は「二つの独立した証言が一つのマスで一致する」 ことにあります。この円環盤では、縦の項目の証言は横の項目の証言の 反復です — イカ の縦証言は イカ の横証言...
diagnose_failure
ja
bad
mini
2x2
advanced
original
yes
cw_398
Подстановочный тест по-русски: для каждой пары «подсказка - ответ» проверьте, можно ли подставить ответ в определяемую фразу. Найдите единственную негодную пару и почините её.
1. «Добыча рыбака» - УЛОВ 2. «Канава вокруг крепости» - РОВ 3. «Комок, например земли» - КОМ 4. «Время года перед зимой» - ЛЕТО
Проверка подстановкой (ответ = определяемой фразе?): 1. УЛОВ - добыча рыбака - да, словарно точно. ✓ 2. РОВ - канава вокруг крепости - да. ✓ 3. КОМ - комок земли - да. ✓ 4. ЛЕТО - «время года перед зимой»? Зимой заканчивается осень, а перед зимой - ОСЕНЬ. ЛЕТО стоит ПЕРЕД осенью. Подстановка ложная: определение указыва...
write_clues
ru
good
mini
4x4
beginner
original
no
cw_247
A language model generated clues for three fill answers. Grammatically flawless, editorially confident, magazine-quality surfaces - and semantically wrong. Audit each against its answer, name the mechanism, and derive the correct approach.
Answers: PEW (3 - a church bench), LEMUR (5 - a Madagascar primate), ONUS (4 - a burden of responsibility). THE MODEL'S CLUES: PEW: 'Church official' LEMUR: 'Madagascan mongoose' ONUS: 'Legal oath's beginning'
The audit, one substitution test per clue: 'Church official' substitutes to VICAR, PRIEST, CLERIC - never PEW, which is the BENCH the official stands near. The clue is definition-shaped text aimed one category away from the answer. 'Madagascan mongoose' - a mongoose-like Madagascan animal exists (the fossa), but the le...
write_clues
en
bad
general
na
intermediate
original
no
cw_130
日本語の盤で語族の問いは三種類あります: 「○で始まる」「○で終わる」 「○番目が○」。本盤の 3 項目がそれぞれどの索引を使うか具体的に示し、 三冊の帳簿がなぜ全部必要なのかを説明してください。
語族の実測値: サで始まる 3 拍語: 6 語 (サカナ, サクラ, サトウ, サラダ, サンゴ, サンマ) 真ん中がカの 3 拍語: サカナのみ メで始まる 2 拍語: メシのみ サで終わる 2 拍語: アサのみ シで終わる 2 拍語: メシ, ナシ, ウマ... (複数)
前索引の例: 2D の最初のマス (0,1)。ここにアサのサが降りると、2D は 「サで始まる 3 拍語」— 6 語のリストが回ってきます。1D の最初のマス (0,0) にアワのアが降りたときも同様です。前索引は語の頭の供給を 受けるすべての項目の入口帳簿です。 後索引の例: 1A の最後のマス (0,1)。1A はアサで完成しますが、逆に 考えれば — 2D のカギの頭が (0,1) に入る語の「二拍目」がここに入る ので、1A の立場では「○で終わる語」の帳簿を引きます。実際、サカナ のシード探索では 1A が「サで終わる語」で候補 1 に絞られました — 全 プールで アサ 一語。後索引は、マスの位置によっては前索引よりず...
methodology_qa
ja
good
general
na
intermediate
original
yes
cw_407
中文格盘的招牌词是四字成语，可四字成语到底能不能交叉成格盘？请用两个机器数字回答——成语之间能共享几格、能拼成几张 2×4 格盘——并解释那条 8 字的「连写词条」是什么。
词池成语统计（机器穷举）: 四字成语 32 条；任意两条成语在某个位置共享同字、可形成单格交叉的 方式共 92 种； 两条成语平行摆放、四列全部成词的 2×4 格盘数: 0。 另有 8 字词条一条: 一鸣惊人人山人海。
先说能做的。32 条成语之间，任意一条横向成语与任意一条纵向成语只要在某个位置共享同字，就可以单格交叉——比如「画蛇添足」的第 4 字与「足智多谋」的第 1 字同为「足」，横竖各摆一条，交点放「足」，就是一个合法的单格交叉。机器穷举出 92 种这样的交叉方式。这不是巧合的产物：成语接龙的文化传统（上条末字引下条首字）天然生产共享字，词池里的成语链条 —— 一举两得→得意忘形→形影不离，三心二意→意气风发→发扬光大——摆明了就是格盘交叉的原材料。 再说做不到的。把两条成语平行作两行，要求四个竖列各拼成一个双字词——机器把 32 条成语两两配对全部检查，答案是 0。四种可能全灭：平行摆放的交叉约束是 92 种单格交叉的成倍叠加，而四个...
fill_grid
zh
good
mini
na
intermediate
original
yes
cw_289
Category clues ('Type of X', 'Kind of X') are a tool with a floor and a ceiling. Explain both: when a category clue is the right call, when it is a cop-out, and write one good and one lazy example using verified pool answers (OTTER, CHEF).
The pattern: 'Type of ___' or '___-like' clues point at any member of a set rather than a definition. Pool answers available: OTTER (5), CHEF (4).
When category is right: the answer is a genuinely generic instance - a member of a set with no better individual definition, or when the constructor wants a fast, crossing-friendly clue for a workman entry. 'Type of swimmer at the zoo' for OTTER is a category clue that works: 'swimmer at the zoo' narrows to a small set...
write_clues
en
good
general
na
intermediate
original
no
cw_110
Посев слова-звезды: конструктор хочет ЛУК в первой записи сетки (пусть это будет тематическое слово). Покажите, что меняет посев - и чему учит «почти мгновенное» заполнение после него.
МАСКА: MASK: #... .... .... ...# Требование: 1A = ЛУК. WORD LIST: 3-letter: АГА, АИР, АКТ, АРА, АУЛ, БИТ, БОР, БЫК, ВАЛ, ВЕК, ВИД, ВОР, ГАД, ГАЗ, ГАМ, ГОД, ДАР, ДОМ, ДУБ, ДЫМ, ЕДА, ЖАР, ЖИР, ЖУК, ЗАЛ, ЗУБ, ИВА, КИТ, КОМ, КОТ, КУБ, ЛЁД, ЛАК, ЛЕВ, ЛЕС, ЛОБ, ЛОМ, ЛУК, ЛУЧ, МЁД, МАК, МИГ, МИР, МОХ, МУЛ, НИЗ, НОС, ОДА, О...
Что делает посев: ЛУК раздаёт три буквы пересечениям - Л в 1D, У в 2D, К в 3D - и это сужение не «в среднем», а по конкретным когортам: слова на Л (9), на У (16), на К (4). Одна трёхбуквенная связка потратила 90-95% свободы трёх соседних записей - посев работает как принудительный первый шаг следа. Чему учит скорость:...
fill_grid
ru
good
mini
4x4
intermediate
original
yes
cw_241
Mid-fill audit. The autofiller has paused with three entries placed in this corner mask, and a constructor reviewing the state wants to type ACE into BOTH 4D and 5D (they never share a cell, so the crossings do not object). Review the proposal against the no-duplicate-words rule, explain how the autofiller's used-words...
MASK: #...# ..... ..... ..... #...# CURRENT STATE: #...# ARENA CIVIC EERIE #...# Placed: 4A ARENA, 6A CIVIC, 7A EERIE. Proposed: 4D ACE and 5D ACE. WORD LIST: 3-letter: ACE, ACT, ADD, AGE, AGO, AID, AIL, AIR, ALE, ALP, ANT, APE, APT, ARC, ARE, ARM, ART, ASH, ASK, AUK, AVE, AWE, AXE, AYE, BAD, BAN, BAR, BAT, BAY, B...
First verify the proposals are crossing-legal, because the duplicate question only matters for placements that would otherwise stand: 4D reads down column 0 - A from ARENA's first letter, C from CIVIC's first, E from EERIE's first - spelling ACE, and every crossing agrees. 5D reads down column 4 the same way: ARENA's l...
fill_grid
en
good
mini
5x5
intermediate
original
no
cw_042
(0,0) पर केवल न या ब बैठ सकता है — यह दावा अब तक टोली-गणित से चला। अब भराई से तौलिए: चारों वैध ग्रिड दिखाइए और बताइए कि तीसरा कोई व्यंजन क्यों नहीं उभर सकता।
MASK: MASK: ..# ... #.. मशीन-जनगणना (147 दो-अक्षर शब्दों को एक-एक कर 1A पर बीज बनाकर पूरी खोज): वैध भराव ठीक चार भराव 1: 1A बल, 1D बस, 2D लड़का, 3A सड़क, 4D कल, 5A काल भराव 2: 1A नल, 1D नस, शेष वही भराव 3 (पार-संस्करण): 1A नस, 1D नल, 2D सड़क, 3A लड़का, 4D काल, 5A कल भराव 4: 1A बस, 1D बल, शेष पार-संस्करण वही
चारों भरावों को ख़ाली ढाँचे पर एक-एक कर बिठाइए — चारों चलते हैं, मशीन हर एक को पूरी कड़ाई से पास करती है: सब प्रविष्टियाँ पूल में, साझा खानों पर गवाहियाँ मेल, कोई दुहराव नहीं। दो 'सीधे' भराव हैं (बल-बस और नल-नस), दो 'पार' वाले (नस-नल और बस-बल) जिनमें लड़का आड़ी चलकर सड़क खड़ी हो जाती है। और चारों में (0,0) पर केवल दो च...
fill_grid
hi
good
mini
3x3
advanced
original
yes
cw_334
The trace's first line for this grid is '4D (468 candidates in play) = ADD'. Why did the solver start there? Work out what every entry's candidate count was at the root, and why the opening move looks mild rather than bold.
MASK: #.... #.... ..... ....# ....# FIRST DECISION: 1. 4D (468 candidates in play) = ADD: 1A narrowed 1,552 -> 17; 5A narrowed 1,552 -> 113; 6A narrowed 1,372 -> 86 WORD LIST: 3-letter: ACE, ACT, ADD, AGE, AGO, AID, AIL, AIR, ALE, ALP, ANT, APE, APT, ARC, ARE, ARM, ART, ASH, ASK, AUK, AVE, AWE, AXE, AYE, BAD, BAN, ...
At the root no letter is placed, so every entry's candidate set is simply its length-pool, countable from the list: the two 3-letter entries (4D, 6D) hold 468 words each; the four 4-letter entries (1A, 5A, 7A, 8A) hold 1,552; the four 5-letter entries (1D, 2D, 3D, 6A) hold 1,372. MRV - always touch the most-constrained...
fill_trace
en
good
mini
5x5
intermediate
original
no
cw_056
भराई की सारी कला एक सवाल में बँटी है: 'इस अक्षर से शुरू/इस अक्षर पर ख़त्म होने वाले शब्द कौन-कौन?' नीचे की टोली-ड्रिल पूरी कीजिए और समझाइए कि आगे-पीछे दोनों सूचकांक क्यों ज़रूरी हैं।
पूल-प्रश्न: ल पर ख़त्म होने वाले दो-अक्षर शब्द = ? स पर ख़त्म होने वाले दो-अक्षर शब्द = ? ल से शुरू होने वाले तीन-अक्षर शब्द = ? बीच में ड़ रखने वाले तीन-अक्षर शब्द = ? का से शुरू होने वाले दो-अक्षर शब्द = ? क से शुरू होने वाले दो-अक्षर शब्द = ?
जवाब, मशीन की गिनती के मुताबिक़। ल पर ख़त्म: 21 — कल, काल, कील, जल, जाल, झील, तिल, तेल, दाल, दिल, नल, नील, पल, पुल, फल, फूल, बल, बाल, रेल, साल, स्कूल। स पर ख़त्म: 6 — ओस, नस, पास, बस, भैंस, रस। ल से शुरू (तीन-अक्षर): केवल लड़का। बीच में ड़: खिड़की, लड़का, सड़क। का से शुरू (दो-अक्षर): कान, काल। क से शुरू (दो-अक्षर): कक्...
fill_grid
hi
good
mini
3x3
intermediate
original
yes
cw_335
नीचे खाली ढाँचे की एक वास्तविक मशीन-खोज है। हर क़दम पर पूछिए: मशीन ने यह प्रविष्टि क्यों चुनी, उम्मीदवार कैसे सिकुड़े, और आख़िरी प्रविष्टि किसने पूरी की।
MASK: MASK: ..# ... #.. मशीन-खोज (बिना बीज, 19 नोड): क़दम1: 2D, उम्मीदवार 57, चुना लड़का; असर: 1A 147→21, 3A 56→2, 5A 147→2 क़दम2: 3A, उम्मीदवार 2, चुना सड़क; असर: 1D 147→6, 4D 147→3 क़दम3: 5A, उम्मीदवार 2, चुना काल क़दम4: 1D, उम्मीदवार 6, चुना नस; 1A छाया-पूर्ण हुआ — नल — मशीन ने जाँचा परिणाम: words: 6, blocks: 2, unc...
क़दम 1: सब प्रविष्टियाँ खुली हैं। सबसे लंबी 2D और 3A (तीन-तीन खाने) की उम्मीदवार-टोलियाँ सबसे छोटी दिखती हैं — 2D के 57 (सारे तीन-अक्षर शब्द) बनाम दो-अक्षर प्रविष्टियों के 147। MRV वही चुनता है: सबसे पतली टोली पहले। लड़का गिरते ही तीन चौराहे एक साथ सिकुड़ते हैं: 1A को अब ल से ख़त्म होना है — 21 बचे (147 में से); 3A के ...
fill_grid
hi
good
mini
3x3
beginner
original
yes
cw_331
A 7x7 midi's autofill ran overnight and returned nothing. The constructor switches to partial placement: lock the theme entry HARVEST at 6A, verify exactly which crossing cells that fixes, then compute what every affected down entry can still be. Rank the resulting candidate sets and state the fill's next move - and th...
MASK: ##...## #.....# ....... ...#... ....... #.....# ##...## Committed placement: 6A = HARVEST. WORD LIST: 3-letter: ACE, ACT, ADD, AGE, AGO, AID, AIL, AIR, ALE, ALP, ANT, APE, APT, ARC, ARE, ARM, ART, ASH, ASK, AUK, AVE, AWE, AXE, AYE, BAD, BAN, BAR, BAT, BAY, BED, BEE, BET, BIG, BIN, BIT, BOA, BOG, BOO, BOW, BOX...
Placement verification first - always partial-works discipline: - 6A spans (2,0)-(2,6), all seven cells open in the mask, entry length 7 = HARVEST's seven letters. Legal. - The placement fixes exactly one cell per crossing entry: (2,0) H for 6D, (2,1) A for 4D, (2,2) R for 1D, (2,3) V for 2D, (2,4) E for 3D, (2,5) S fo...
fill_grid
en
good
daily
7x7
advanced
original
no
cw_046
नीचे दो छोटी लेजर हैं — एक जीत की, एक मौत की। दोनों में मौत के दो अलग तरीक़े छिपे हैं: सामने से शून्य-उम्मीदवार, और पास से छाया-हत्या। दोनों को अलग करके पहचानिए — यही निदान की बुनियाद है।
MASK: MASK: ..# ... #.. लेजर 1 (बीज 1A=नल, 6 नोड, जीत): नोड1: 2D (उम्मीदवार 1) → लड़का नोड2: 1D (उम्मीदवार 2: नदी, नस) — नदी पहले: नोड3 पर 3A के उम्मीदवार शून्य (दीवार वा-मध्य अकेला, 2D में जा चुका), मौत; वापसी नोड4: नस → असर: 3A 2→1; नोड5: सड़क (मजबूरी); नोड6 में 5A काल — कान पहले मरा (4D कन शब्द नहीं) लेजर 2 (बीज 1A=...
मौत की पहली किताब: शून्य उम्मीदवार। नदी-शाखा में मशीन ने कुछ गिराया ही नहीं — 1D ने नदी चुना, और अगले नोड में 3A की टोली देखी: वा-मध्य शब्द पूल में दीवार अकेला है, और दीवार 2D में जा चुका है — छोड़ दो उम्मीदवार कहाँ, एक भी नहीं। नोड बनते ही मौत। इस मौत की पहचान सरल है: टोली देखिए — अकेली टोली का शब्द जहाँ कहीं बैठ चुका...
fill_trace
hi
good
mini
3x3
advanced
original
yes
cw_341
Demonstrate why 'fill the widest-open entry first' fails. In this two-block mini, a constructor opens on the bottom row - 8A, the entry with the largest candidate pool - and picks WOKE because it sounds good. Run the propagation honestly, identify what kills the fill, and prove no completion exists.
MASK: #.... ..... ..... ..... ....# Opening choice forced by the constructor: 8A = WOKE ('widest-open entry first'). WORD LIST: 4-letter: ABET, ABLE, ACED, ACID, ACNE, ACRE, AIDE, AJAR, ALOE, ALTO, AMEN, AMID, AMPS, ANTE, AQUA, ARID, ATOM, AUNT, AVID, AVOW, AWAY, AWRY, BABE, BABY, BACK, BAIL, BAIT, BAKE, BALD, BALE...
The heuristic feels right - start where you have the most freedom - but freedom is exactly what you should not spend first. 8A is the widest-open entry on the board, 1,552 candidates before any letter is placed. The constructor spends that freedom on taste: WOKE, a perfectly good word. One placement, four letters - and...
fill_grid
en
good
mini
5x5
advanced
original
no
cw_044
From the corner mini's first four decisions (AGO at 1A, ONION at 3D, AHEAD at 1D, DEN at 8A), reconstruct the exact grid state: which cells are filled, what each contains, and what the state reveals about the solver's next move.
MASK: #...# ..... ..... ..... #...# TRACE STEPS 1-4: 1. 1A (468 candidates in play) = AGO: 1D narrowed 1372 -> 45; 2D narrowed 1372 -> 72; 3D narrowed 1372 -> 23 2. 3D (23 candidates in play) = ONION: 4A narrowed 1371 -> 117; 6A narrowed 1371 -> 101; 7A narrowed 1371 -> 81; 8A narrowed 467 -> 46 3. 1D (45 candidates i...
Placing each word on its entry's cells - 1A runs (0,1)-(0,3), 3D runs (0,3)-(4,3), 1D runs (0,1)-(4,1), 8A runs (4,1)-(4,3): #AGO# .H.N. .E.I. .A.O. #DEN# Row by row: row 0 is fully filled (#AGO#). Row 1 holds AHEAD's H at (1,1) and ONION's N at (1,3). Row 2 holds AHEAD's E at (2,1) and ONION's I at (2,3). Row 3 hold...
fill_trace
en
good
mini
5x5
intermediate
original
yes
cw_069
Dissection drill, harder: this clue stacks two mechanisms, and the enumeration is the only length hint. Parse it completely and justify the indicator assignments - one of the two transformation words could plausibly belong to either mechanism; say why the parse reads it as it does.
Clue: 'Nowhere, going back, holds the present (3)' Answer: NOW
The parse: 'Nowhere' -> the fodder text (a real word) 'going back' -> REVERSAL: read the fodder backward: EREHWON 'holds' -> HIDDEN: a consecutive run inside EREHWON: positions 4-6 spell W-O-N 'the present' -> DEFINITION: NOW Assembly: WON is NOW read backward; the hidden run holds the answer's reversal. Lett...
cryptic_clue
en
good
cryptic
na
advanced
cryptic-guide
yes
cw_160
State and defend the identity prohibition: a clue may not contain its answer, its answer's obvious stem, or a definition circular with the answer - and the surface may not merely re-ask 'what is the word for ___'. Write the borderline cases for PHONE and CIVIC that show where the line is.
Answers: PHONE (5), CIVIC (5). The leak case from the flawed row: 'The burden of an ONUS, so to speak'.
The prohibition, precisely: the clue's surface must not contain the answer string ('an ONUS, so to speak'), the answer's stem with decoration ('A phone-y device' for PHONE), or a circular definition ('Civic word' for CIVIC - defining the word with itself). The deep version: a clue must reach its answer through MEANING,...
write_clues
en
good
general
na
intermediate
original
no
cw_140
Assemble the indicator lexicon: for each of the mechanisms, list the standard indicator families and the grammatical shapes they take - then state the meta-rule that keeps them all honest.
The mechanisms: anagram, charade, container, deletion (head/tail/heart), double definition, homophone, hidden, reversal.
ANAGRAM - upheaval words: scrambled, mixed, reformed, broken, wild, mad, upset, out of sorts, strange, peculiar, engineered. Shape: [definition], [indicator] [fodder]. CHARADE - NO indicator: juxtaposition does the work. Shape: [definition]: [part-clue], [part-clue]. CONTAINER - containment words: in, inside, within, h...
cryptic_clue
en
good
cryptic
na
intermediate
cryptic-guide
no
cw_154
日本語盤のマスに入る単位はモーラ (拍) です — 文字コードでも漢字数でもない。ワカメが 3 マス、アサが 2 マスな理由を、クロスワードの表記慣習 (カタカナ・一拍一格) とともに説明してください。
MASK: MASK: ..# ... #.. FILL: FILL: アサ# ワカメ #ナシ 単位の事実: アサ = [ア, サ] 2 マス; サカナ = [サ, カ, ナ] 3 マス; ワカメ = [ワ, カ, メ] 3 マス; 語はすべてカタカナ表記。
日本語クロスワードの掟は昔から単純です: 答えはカタカナ、一マスには一モーラ (一拍)。アサ (朝) は二拍なので二マス、サカナ (魚) は三拍なので三マス。仮名一文字がほぼ一拍に対応するので「マス目 = 仮名数」が成り立ちます — ここが、辞書に音節単位のインターフェイスを持つインド系文字や、字母と発音の距離が遠い英語とは決定的に違う点です。 なぜカタカナだけなのか。第一に、解答の表記を一意に固定するためです。朝・麻・痷 — 漢字なら書き方が何通りもある。ところが答えをカタカナに限れば「アサ」の三拍は「アサ」の一綴りだけ。第二に、ひらがなとの区別です。パズル誌の盤面は「カタカナの世界」であり、同じ音の ひらがな 表記は別の文字列...
design_grid
ja
good
mini
3x3
beginner
original
yes
cw_375
あるモデルが盤面を提出し、「ナシはなし、サカナはさかな — 同じ語、意味も読みも同じ、検証は済み」と報告しました。グリッドと報告書の双方を検死し、この「文字が同じだから語も同じ」という弁明が盤面の契約でどう扱われるかを説明してください。
MASK: MASK: ..# ... #.. モデル提出の盤: アサ# ワカメ #なし モデルの主張: 「3 つの項目をひらがなで書いた。サカナ = さかな、ナシ = なし。音も意味も同一。検証完了。」 機械の評決: unknown_words: 2D サカな, 4D メし, 5A なし。
まず盤そのものを再導出します。真ん中の行は ワ・カ・メ — ここは無事。 タテ 2D は上から サ (アサ)・カ (ワカメ)・な (最終行) = サカな。タテ 4D は メ・し = メし。ヨコ 5A は なし。機械は三つの項目を unknown と裁きます — な・し・までがひらがなの世界の住人で、この盤の契約書 (カタカナ・プール照合) に署名していないからです。 モデルの弁明を審きます。「音も意味も同一」— 言語学の世界では正しい。 かな表記の違いは書体の違いに近い。しかし盤面の契約書は言語学の世界に置かれていません。語プールは [サ, カ, ナ] というカタカナの分解列で語を保管しており、照合は分解列の完全一致で行われます。...
fill_trace
ja
bad
mini
3x3
advanced
original
yes
cw_390
Историческая справка: как сканворд завоевал русскоязычную прессу и чем его победа объясняется? Публичные факты жанра: массовое распространение в газетах конца 1980-х - 1990-х, несимметричные сетки, стрелки, имена собственные.
(без артефакта - вопрос по публичной истории жанра)
Хронология: к концу 1980-х - началу 1990-х сканворды становятся массовым жанром русскоязычной периодики - от газетных разворотов до толстых сборников; потоковый формат «две сетки на страницу, десятки загадок в номер» стал обычным явлением киоска. Чем объясняется победа: жанр перенастроил параметры дисциплины под массо...
historical_note
ru
good
general
na
intermediate
historical
no
cw_263
Teach the REVERSAL mechanism: the answer is a real word read backward (or forward from backward fodder). Parse the worked example, give the direction vocabulary for across versus down clues, and state the fairness requirement on the fodder.
Worked clue: 'Ned turned around, finding the lair (3)' Answer: DEN
The parse: DEN reversed = NED. - 'Ned' -> the FODDER: a name, read as a word. - 'turned around' -> the REVERSAL indicator: read it backward. N-E-D becomes D-E-N. - 'the lair' -> the DEFINITION: a den is a lair. Direction vocabulary: in ACROSS clues, 'back, backward, turned around, reversed, going west, left' signal ri...
cryptic_clue
en
good
cryptic
na
beginner
cryptic-guide
yes
cw_152
下のトレースは 237 ノードの戦場を生き抜いた成功経路の切符です。各欄の読み方 — 特に「影響」欄の 79→5 のような数字がどの帳簿から来るのか — と、最後に名前のない 5A がどう完成したかを説明してください。
MASK: MASK: ..# ... #.. トレース (無シード、ノード台帳 237): 1A (2 マス, 候補 80) → アサ | 影響: 1D 79→5, 2D 100→6 1D (2 マス, 候補 5) → アワ | 影響: 3A 100→2 3A (3 マス, 候補 2) → ワカメ | 影響: 2D 6→1, 4D 78→1 2D (3 マス, 候補 1) → サカナ | 影響: 5A 78→2 4D (2 マス, 候補 1) → メシ | 影響: なし
欄の読み方から。各行は一つの「決定」です: 選ばれた項目番号、マス数、その瞬間の候補数、置いた語、そしてこの語が交差相手の候補をどう潰したか。一行目の 79 はこういう帳簿です: 80 語のうちアサが今し方「使用済み」に登録されたので、1D の世界の在庫は 79。そこへ (0,0) のアが降りて 「アで始まる語」だけが残る — 5 候補。2D には (0,1) のサが降りて 「サで始まる 3 拍語」の 6 候補。つまり影響欄は「置く前の在庫 → 置いた後の在庫」の二段括弧です。 二行目、1D の 5 候補からアワ。影響欄に注目 — 相手は 3A のみで 100→2。 アワ が (1,0) にワを置いたので、3A は「ワで始まる 3...
fill_trace
ja
good
mini
3x3
intermediate
original
yes
cw_387
Write the historical note on the crossword's origin: who created the first one, where and when, and what its creator would recognize in today's grids?
(no artifact - public-domain history question)
The first crossword appeared in the New York World on December 21, 1913 - the 'Word-Cross' of Arthur Wynne, a British-born journalist who folded it into the paper's fun supplement. Its shape was a diamond with a hollow center, derived from the ancient word square, and it came loaded with features we would now audit awa...
historical_note
en
good
general
na
beginner
historical
no
cw_226
A language model was asked for 'a cryptic clue for PLEA' and produced the clue below. Audit it against the fairness triplet, name the failure mechanism, and derive the correct approach for machine-assisted cryptic writing.
THE MODEL'S CLUE: 'Angry LEAP mixed around the court's calm ocean of pearl (4)' Intended answer: PLEA
The triplet audit: 1. Honest indicators: 'mixed' is a real anagram word - attached to LEAP, the operation is performable and yields PLEA. One honest indicator exists. 2. Letter accounting: the clue spends LEAP - and then spends 'the court's calm ocean of pearl': words that account for nothing. Is 'pearl' fodder? PEARL ...
cryptic_clue
en
bad
cryptic
na
advanced
original
no
cw_168
Teach the ANAGRAM mechanism completely: the three parts (definition, fodder, indicator), what the indicator promises, and how the solver verifies. Use this worked example and show the full parse.
Worked clue: 'Court appeal from a scrambled LEAP (4)' Answer: PLEA
The three parts: - DEFINITION: 'Court appeal' - the straight-meaning end that tells you what the answer MEANS. Here it opens the clue. - FODDER: 'LEAP' - the raw material; its letters will be rearranged into the answer. - INDICATOR: 'scrambled' - the word that PROMISES the transformation. Anagram indicators are always ...
cryptic_clue
en
good
cryptic
na
beginner
cryptic-guide
yes
cw_145
Языковую модель попросили заполнить русскую 4×4; её ответ перед вами. Разберите отказ на уровне механизма - почему появились именно ЭТИ буквы - и выведите правильный подход из самого провала.
МАСКА: МАСКА: #... .... .... ...# ВЫВОД МОДЕЛИ: ВЫВОД МОДЕЛИ: #ЛУК ЗЛОК АТОМ РОВ# WORD LIST: 3-letter: АГА, АИР, АКТ, АРА, АУЛ, БИТ, БОР, БЫК, ВАЛ, ВЕК, ВИД, ВОР, ГАД, ГАЗ, ГАМ, ГОД, ДАР, ДОМ, ДУБ, ДЫМ, ЕДА, ЖАР, ЖИР, ЖУК, ЗАЛ, ЗУБ, ИВА, КИТ, КОМ, КОТ, КУБ, ЛЁД, ЛАК, ЛЕВ, ЛЕС, ЛОБ, ЛОМ, ЛУК, ЛУЧ, МЁД, МАК, МИГ, МИР...
Пере-выводим записи из клеток: 1A ЛУК ✓, 5A АТОМ ✓, 6A РОВ ✓ - модель сохранила три настоящих слова; но 4A читается З-Л-О-К = «ЗЛОК» - не слово, собранное из частотных русских морфем, похожее на слово ровно настолько, чтобы обмануть беглый взгляд. Локальность поражения - вся механика урока. ЗЛОК встал в ДЕЛОвскую стро...
fill_grid
ru
bad
mini
4x4
intermediate
original
no
cw_255
Step back from the mechanics: what is a solver trace FOR? Cover the four roles traces play in crossword construction - including the parallel to how professional autofill tools actually work - and state the discipline a trace imposes on everyone who reads one.
You have now read traces of three minis: openings, cascades, forced endings, side effects, a seeded route, a death log. Synthesize: what is a trace for?
A trace is the construction process's flight recorder, and it serves four distinct roles: 1. DEBUGGING. When a fill fails, the trace (or its terminal variant, the death log) says WHERE it died, at what depth, and which entry ran out of candidates - which converts 'the grid won't fill' into 'the V at (2,3) leaves 2D ex...
fill_trace
en
good
general
na
intermediate
original
no
cw_078
Why does the discipline treat 'the model wrote something fluent' as a DEFECT signal rather than a feature?
(no artifact - pure methodology question)
Because the project's documented failure family is exactly that: fluency without operations. HEPT, ONUX and WERS (letter-salad in word-shaped costume), PLAE and ERSAE (traces with real numbers and fake words), 'Church official' for PEW (definition-shaped prose), the machinery-soup cryptic, the verdict-first verificatio...
methodology_qa
en
good
general
na
advanced
original
no
cw_219
ナシ (梨/無し) は日本語盤の常連の同音語です。同じ二拍に二つの意味を住まわせたカギの梯子 — 易から難へ — を三段作り、各段の妥当性を置換テストで示してください。
目標語: ナシ (2 マス) プール内の同音近隣: ナベ (2 マス) — ただし意味は全く別 本盤での立場: 5A — (2,1) は 2D の三拍目、(2,2) は 4D の二拍目
第一段 (易、実義カギ): 「果物。旬は夏から秋。甘みとざくざくした歯触り」。置換テスト: ナシ (梨) — 成立。このカギは簡単ですが、競合が いないことを確認する必要があります — 2 拍の果物はプールに他にある?カキ (柿)? プールを調べ、いなければカギは単独で答えを固定します。易カギは「プール内の競合不在」で成り立っているのであって、辞書世界での競合不在ではありません。 第二段 (中、反義カギ): 「あるの反対 — 『元手が○○で商いを始める』の ○○」。置換テスト: 元手がなしで…成立。今度はナシ (無し) の顔を使っています。同音の二つの意味が同じマスに住んでいるので、カギは どちらの顔でも到着できる — それを逆手...
write_clues
ja
good
mini
3x3
intermediate
original
yes
cw_392
(0,0) पर कौन-सा व्यंजन बैठ सकता है? मशीन का जवाब है: ठीक दो — ब या न। इस 'दो-व्यंजन सिद्धांत' को टोली-गणित से साबित कीजिए और बताइए कि बीज चुनते समय कौन-सी दो शर्तें एक साथ पूरी होनी चाहिएं।
टोली-तथ्य (मशीन-गणना): ख में ख़त्म होने वाले 2-अक्षर शब्द — कोई नहीं; ओ में ख़त्म: कोई नहीं ल में ख़त्म होने वाले: 21 शब्द (कल, काल, जल, जाल, बल, नल, पल, फल, फूल, बाल ...) स में ख़त्म होने वाले: 6 शब्द (ओस, नस, पास, बस, भैंस, रस) दोनों सूचियों का प्रथम-व्यंजन प्रतिच्छेदन: {ब, न}
(0,0) एक ही साथ 1A (आड़ी) और 1D (खड़ी) का पहला खाना है। इस पाठ्यक्रम के संदर्भ-ढाँचे में 1A का दूसरा अक्षर ल बनना है (ग्रिड की साझा गवाही के हिसाब से) और 1D का दूसरा अक्षर स। तो (0,0) पर बैठने वाले व्यंजन X को दो टोलियाँ एक साथ चाहिए: Xल पूल में हो और Xस पूल में हो। मशीन ने पूरी पूल पर यह प्रतिच्छेदन निकाला: Xल-टोली के...
design_grid
hi
good
mini
3x3
advanced
original
yes
cw_330
「コードポイント数 = マス数」は日本語盤では必ず崩れます。拗音 (ゃゅょ)、促音 (っ)、長音符 (ー) の三法を、語族の実例で説明してください。
実測値 (機械検証): シャツ = [シャ, ツ] — 3 コードポイント / 2 マス ヨット = [ヨッ, ト] — 3 コードポイント / 2 マス サッカー = [サッ, カ, ー] — 4 コードポイント / 3 マス ワイシャツ = [ワ, イ, シャ, ツ] — 5 コードポイント / 4 マス カヌー = [カ, ヌ, ー] — 3 コードポイント / 3 マス 長音符を含む語は全プールで 29 語。ただし一語も ー で始まらない。
第一の法、拗音。ャ・ュ・ョ は直前の仮名と合体して一拍になります: シャツ は シ+ヤ+ツ の 3 文字ですが、盤面では シャ が一マス、ツ が一マスの 2 マス語。ワイシャツ の 5 文字は 4 マスに収まり、コードポイントとマス数の差が最大になります。カボチャ の チャ、キャベツ の キャ も同じ — 小さい文字は「半分の幅で前の拍にくっつく」記号であって、独立したマスではありません。 第二の法、促音。っ は前の拍と一体化します: ヨット は ヨッ+ト の 2 マス、ラッパ は ラッ+パ、サッカー は サッ+カ+ー。モロッコ は モ+ロッ+コ。小さい っ は単独で一拍を占めることもありますが、このプールの慣行では前に合流します...
design_grid
ja
good
mini
3x3
intermediate
original
yes
cw_376
ग्रिड ने अक्षर से इक़रार किया है, वर्ण से नहीं — यह बात इतनी बार दोहराई गई कि अब उसका पूरा हिसाब माँगना ज़ायज़ है। पाँच सबूतों की मेज़ लगाइए — हर सबूत पूल की गिनती से — जो दिखाए कि खाने की ईकाई अक्षर ही हो सकती है।
गवाहों की सूची: काल [का, ल] बनाम कल [क, ल] — एक ही ढाँचे में अलग घर कुत्ता [कु, त्ता] — दो खाने, छह वर्ण ड़-मध्य तीन-अक्षर टोली: खिड़की, लड़का, सड़क — तीन शब्द, एक मोड़; ड़-आदि टोली: शून्य
पहला सबूत: मात्रा की सत्ता। का और क के बीच की दूरी एक कोडपॉइंट है — पर ग्रिड में यह दूरी कल और काल के बीच की दूरी है। इसी ढाँचे में दोनों भाई अलग-अलग घरों में रहते हैं और दोनों के घर क़ानूनी — यह तभी संभव है जब घर की गिनती मात्रा सहित हो। वर्ण-गिनती में काल को तीन खाने मिलते, कल को दो — दोनों एक ही प्रविष्टि-दौड़ में फ...
methodology_qa
hi
good
general
na
advanced
original
yes
cw_368
Dissection drill: parse this multi-mechanism clue completely, in writing - every fodder word, every indicator, the assembly order, and the fairness checks.
Clue: 'Quiet expression of disgust in church (3)' Answer: PEW
The parse: 'Quiet' -> P (fixed music abbreviation: piano = quiet, marked p) 'expression of disgust' -> EW (interjection) 'in church' -> the DEFINITION: a pew Assembly: P + EW = PEW (1 + 2 = 3 = enumeration) The fairness checks: (1) P from 'quiet' is a conventional equivalence (the stave's dynamics), not a priv...
cryptic_clue
en
good
cryptic
na
intermediate
cryptic-guide
yes
cw_159
Why audit a SEED's crossing letters before running a fill? Isn't that what the search will discover anyway?
(no artifact - pure methodology question)
The search discovers it at search-speed; the audit knows it at counting-speed. The pre-flight profile - for each of the seed's letters, the size of the position-cohort it demands - is seven list counts, seconds of work, and it predicted everything the probes later proved: CURTAIN's floor of 15 (safe), KARAOKE's floor o...
methodology_qa
en
good
general
na
intermediate
original
no
cw_210
Профиль длин записей - это «прейскурант» сетки: какие когорты словаря она потребует. Выведите профиль для русской 4×4 и посчитайте, глубоки ли соответствующие когорты в словаре проекта (90 трёхбуквенных, 192 четырёхбуквенных слова).
МАСКА: MASK: #... .... .... ...# Словарные когорты проекта: 3 буквы - 90 слов, 4 буквы - 192 слова (проверено при импорте).
Профиль: по горизонтали - 3, 4, 4, 3; по вертикали - 4, 4, 3, 3. Итого сетке нужно четыре трёхбуквенных и четыре четырёхбуквенных слова. Сравнение с когортами: четырёхбуквенных слов в словаре 192 - на каждое место есть 192 кандидата в пустой сетке; трёхбуквенных - 90. Уже на этом уровне видно правило экономики заполне...
design_grid
ru
good
mini
4x4
intermediate
original
yes
cw_232
Explain the historical and mechanical partnership between cryptic clues and BARRED grids: why the wordplay tradition grew up on bars rather than blocks, and what the pairing does to solve difficulty.
Facts: British cryptic crosswords traditionally use barred grids - few or no black squares, long entries, heavy checking; American-style blocked grids use black squares and short entries. The cryptic clue already delivers the answer through wordplay; the standard-clue grid must let crossings carry inference.
The mechanical fit: a cryptic clue is SELF-VERIFYING - once parsed, the answer is certain, so the solver needs crossings far less for CONFIRMATION than a standard-clue solver does. Bars deliver exactly that economy: long entries, fewer of them, each crossing checked but sparse overall - the grid can afford light crossi...
cryptic_clue
en
good
cryptic
na
intermediate
cryptic-guide
no
cw_163
Two more entries have landed since the last audit. Recompute what 5A can be and explain what changed in its candidate set. Then finish the grid mentally and verify your expectation against the full fill.
MASK: #.... #.... ..... ....# ....# CURRENT STATE ('.' = empty): #PLEA #.ERD NOMAD ..US# ..RE# Placed so far: 4D = ADD, 1A = PLEA, 3D = ERASE, 6A = NOMAD, 2D = LEMUR WORD LIST: 4-letter: ABET, ABLE, ACED, ACID, ACNE, ACRE, AIDE, AJAR, ALOE, ALTO, AMEN, AMID, AMPS, ANTE, AQUA, ARID, ATOM, AUNT, AVID, AVOW, AWAY, AW...
Reading 5A's cells against the current state gives its pattern; filtering the word list through the crossing letters leaves exactly 1 candidate: HERD. The set has collapsed to a single word: the fill forces HERD here. No choice remains - this is the solver's 'last entry falls out' moment, where the grid completes itse...
fill_grid
en
good
mini
5x5
intermediate
original
yes
cw_035
Run the complete publish pipeline on this finished mini - every hard gate, every warning, and the constructor's own audit - and produce the go/no-go verdict with the full evidence trail, as if submitting to a publishing tool for the first time.
SUBMISSION: MASK: #...# ..... ..... ..... #...# FILL: #AGO# PHONY EERIE WAGON #DEN# Title: 'Corner Store' Clues: all ten entries clued (audited earlier in this project: definitions, grammar, and enumerations verified per the clues module's four-point audit).
HARD GATES: - Grid filled: 21 open cells, all holding letters, re-derived into ten entries (AGO, PHONY, EERIE, WAGON, DEN, AHEAD, GORGE, ONION, PEW, YEN). PASS. - Every entry clued: ten clues attached to the ten entry keys the mask numbers (1A, 1D, 2D, 3D, 4A, 4D, 5D, 6A, 7A, 8A). PASS. - Title present: 'Corner Store'....
validate_puzzle
en
good
mini
5x5
intermediate
original
yes
cw_178
यह रिपोर्ट 'सब पास' की मुहर लेकर आई है। साथ का ग्रिड उसकी कहानी से मेल नहीं खाता। रिपोर्ट के निरीक्षण का निरीक्षण कीजिए: रिपोर्ट क्या कहती है, ग्रिड क्या कहता है, दोनों एक साथ सच हो सकते हैं?
रिपोर्ट (मॉडल-निर्मित): 'छह प्रविष्टियाँ सत्यापित: 1A बल, 1D बस, 2D लड़का, 3A सड़क, 4D कल, 5A काल; साझा खाने मेल; दुहराव नहीं; निर्णय: उत्तीर्ण।' साथ का ग्रिड: बल# सडक #काल
पहले ग्रिड की बात सुनिए — क्योंकि काग़ज़ पर मुहर जितनी भारी हो, खाने उतने ही भोले हैं। इस ग्रिड की प्रविष्टियाँ: 1A बल, 1D बस, 3A सडक (बिना नुक़्ता), 2D लडका (बिना नुक़्ता), 4D कल, 5A काल। मशीन इस ग्रिड को अस्वीकार करती है: unknown_words — सडक, लडका। यानी ग्रिड जिसे 'उत्तीर्ण' बताया जा रहा है, वह दो प्रविष्टियों में ही...
validate_puzzle
hi
bad
mini
3x3
advanced
original
yes
cw_356
Audit this clue. A junior setter swears it is 'minimalist cryptic'. Run the contract test, identify what is missing, and repair the clue without changing its wordplay.
Clue: 'Scrambled LEAP (4)' Intended answer: PLEA
The contract test: a cryptic clue owes the solver BOTH a wordplay route to the answer AND a definition of it. This clue delivers the wordplay (LEAP, honestly scrambled, yields PLEA for anyone who tries) and delivers NO definition at all - nothing in the surface says what the answer MEANS. The solver who does not happen...
cryptic_clue
en
flawed
cryptic
na
intermediate
original
no
cw_164
Verify this completed fill against the word list and the mask. Check every entry for list membership, every crossing for letter agreement, and the grid for duplicate words. Report each check as PASS or FAIL with the evidence.
MASK: #.... #.... ..... ....# ....# FILL TO VERIFY: #PLEA #HERD NOMAD ONUS# WERE# WORD LIST: 3-letter: ACE, ACT, ADD, AGE, AGO, AID, AIL, AIR, ALE, ALP, ANT, APE, APT, ARC, ARE, ARM, ART, ASH, ASK, AUK, AVE, AWE, AXE, AYE, BAD, BAN, BAR, BAT, BAY, BED, BEE, BET, BIG, BIN, BIT, BOA, BOG, BOO, BOW, BOX, BOY, BUD, BUG...
Entry-by-entry audit, reading the fill back off the grid: 1. List membership: - 1A reads P-L-E-A = PLEA - in list. PASS - 1D reads P-H-O-N-E = PHONE - in list. PASS - 2D reads L-E-M-U-R = LEMUR - in list. PASS - 3D reads E-R-A-S-E = ERASE - in list. PASS - 4D reads A-D-D = ADD - in list. PASS - 5A reads H-E-R-D = HERD...
fill_grid
en
good
mini
5x5
beginner
original
yes
cw_040
A constructor building a beginner-friendly 5x5 submits this mask, confident that 'beginners love short words, so short is short-friendly'. Review the design against the beginner-experience argument and the structural rules, identify the single worst design decision, and show the standard mask this shape should have bee...
SUBMITTED MASK: #.#.# #.#.# ..... #.#.# #.#.#
Structural audit first: blocks at (0,0),(0,2),(0,4),(1,0),(1,2),(1,4),(3,0),(3,2),(3,4),(4,0),(4,2),(4,4) - twelve blocks, 48 percent of the grid. Entries: row 0: (0,1),(0,3) are single cells - not entries. Row 1: same. Row 2: full 5. Rows 3-4: mirror rows 1-0. Columns: c0: (2,0) alone between blocks - single cell, no...
design_grid
en
flawed
mini
5x5
beginner
original
no
cw_030
A trainee's clue set keeps producing wrong-length solver complaints. Three sampled clues below - find the grammatical mismatch, explain why it manufactures wrong answers, and repair all three.
Intended answers: GLARE (5), RISEN (5), WERE (4). Trainee's clues: 'Angrily stared' (intended: GLARE) 'Getting up, as the sun' (intended: RISEN) 'Existing in the past' (intended: WERE)
The mismatch, clue by clue: 'Angrily stared' is a past-tense verb phrase - a solver obeying clue grammar writes GLARED (6), breaks a crossing, and blames the grid. The base form GLARE needs 'Stare angrily' (present/imperative). 'Getting up, as the sun' is a present participle - it points at RISING (6), not RISEN (5). T...
write_clues
en
flawed
general
na
intermediate
original
no
cw_128
Design a 5x5 mask with exactly two black squares, rotationally symmetric, and prove in writing that no unchecked squares can exist in it - not by inspecting all 23 cells, but by a structural argument about the block placement.
Target: 5x5, exactly 2 blocks, rotational symmetry, zero unches.
With only two blocks and rotational symmetry, the blocks must be a rotated pair. Choose them at opposite corners: (0,0) and (4,4). FINAL MASK: #.... ..... ..... ..... ....# Structural argument: an unchecked square can only appear at a cell where one direction is cut short by a block or edge. So unches can only occur ...
design_grid
en
good
mini
5x5
intermediate
original
yes
cw_005
A solver emails: 'your pinwheel has two different solutions - I filled it completely differently from your answer key and every word is real.' The constructor verifies: the solver is right. Diagnose the situation: is this a defect, what do the gates say, when does multiple completions actually harm, and what should the...
MASK: #.... #.... ..... ....# ....# PUBLISHED FILL: FILL: #PLEA #HERD NOMAD ONUS# WERE# SOLVER'S ALTERNATIVE (machine-verified: every entry a real word, every crossing agreeing, no duplicates): #STEW #HIRE NOMAD EVES# WERE#
First, the facts on the table: the alternative is legal - STEW, HIRE, NOMAD, EVES, WERE across; SHOVE, TIMER, ERASE, WED, NEW down. Every entry is a real word, every crossing agrees, no duplicates. The published fill (PLEA, HERD, NOMAD, ONUS, WERE; PHONE, LEMUR, ERASE, ADD, NOW) is equally legal. The mask admits (at le...
diagnose_failure
en
good
mini
5x5
intermediate
original
yes
cw_097
State and justify the definition-position rule for cryptic clues: where the definition lives, how to find it when the clue will not say, and the practical scanning protocol a solver uses on every unfamiliar clue.
Worked examples to scan: 'Court appeal from a scrambled LEAP (4)' 'Ned turned around, finding the lair (3)'
The rule: the definition sits at ONE END of the clue - the first few words or the last few words - and the wordplay occupies the rest. The middle of a cryptic clue is never the definition's home. Scanning example one: 'Court appeal from a scrambled LEAP' - test the ends. 'Court appeal' as definition? PLEA fits, and th...
cryptic_clue
en
good
cryptic
na
intermediate
cryptic-guide
no
cw_155
This corner-mini trace's final line shows no effects, and the trace records eight decisions for a grid with ten entries. Reconcile both facts: where do the last two words come from, and what does the asymmetry say about what a trace can and cannot be trusted to show?
MASK: #...# ..... ..... ..... #...# SOLVER TRACE: 1. 1A (468 candidates in play) = AGO: 1D narrowed 1372 -> 45; 2D narrowed 1372 -> 72; 3D narrowed 1372 -> 23 2. 3D (23 candidates in play) = ONION: 4A narrowed 1371 -> 117; 6A narrowed 1371 -> 101; 7A narrowed 1371 -> 81; 8A narrowed 467 -> 46 3. 1D (45 candidates in p...
The last decision is 4A (3 candidates) = PHONY, with no effects: nothing open was left to narrow. But eight decisions cannot fill ten entries. The two missing entries are 4D and 5D, the 3-letter corner downs, and they never received a decision because PHONY completed them from above: PHONY's P at (1,0) is 4D's first ce...
fill_trace
en
good
mini
5x5
intermediate
original
yes
cw_060
एक मॉडल ने जवाब सही सोचे — और ग्रिड में BALA, SADAK लिख दिए। भाषा सही, लिपि ग़लत। 'लिपि की परत' की यह चूक ग्रिड के तीन नियमों के ख़िलाफ़ क्यों जाती है, तीनों गिनाइए।
MASK: MASK: ..# ... #.. मॉडल का ग्रिड: BAL# SADA K#KA मॉडल का बयान: 'जवाब मैंने सही सोचे थे — बल, सड़क, काल — बस लिपि बदल दी।'
पहला नियम-भंग: खाने की गिनती। सड़क के तीन खाने हैं — मॉडल ने चार रोमन अक्षर भेजे (S,A,D,A), और व्यंजनों की तो लंबाई ही किसी और गणित में है। ग्रिड खाने गिनता है — लेखा बाँझ पर जमा नहीं होता। BAL तीन अक्षर लेकर दो खानों में जमा चाहता है — बाक़ी सब गड़बड़ इसी से फैलती है। दूसरा नियम-भंग: विच्छेदन का टकराव। ग्रिड की तुलना...
diagnose_failure
hi
good
mini
3x3
intermediate
original
yes
cw_352
Inflected answers demand inflection-honest clues. Write clues for RISEN and GREW (both in the verified two-corner fill), keeping each verb's FORM visible in the clue's grammar, and state the convention for each inflection type.
Answers: RISEN (5 - past participle of rise), GREW (4 - simple past of grow).
RISEN (past participle): 'Come up, as the sun'. The clue's verb 'come up' is itself the participle form ('has come up' - 'has risen'); the forms mirror. A solver reads the clue's shape and knows the answer is a participle, not RISE (infinitive) or ROSE (past). GREW (simple past): 'Got bigger, as a plant'. 'Got' is simp...
write_clues
en
good
mini
5x5
intermediate
original
no
cw_121
Compute 2D's candidate set in this partial fill and identify the candidate that minimizes risk for the two remaining 4-letter entries.
MASK: #.... ..... ..... ..... ....# CURRENT STATE ('.' = empty): #ABET .L.R. .I.E. EVICT .E.T# Placed so far: 1D = ALIVE, 7A = EVICT, 1A = ABET, 3D = ERECT WORD LIST: 5-letter: ABOUT, ABOVE, ACTOR, ACUTE, ADMIT, ADOPT, ADULT, AFTER, AGAIN, AGENT, AGREE, AHEAD, AISLE, ALARM, ALBUM, ALERT, ALIBI, ALIEN, ALIGN, ALIKE...
Reading 2D's cells against the current state gives its pattern; filtering the word list through the crossing letters leaves exactly 7 candidates: BASIC, BASIL, BASIN, BEGIN, BRAID, BRAIN, BROIL. Every one of these agrees with all currently placed letters, and none duplicates an entry already used. The choice among the...
fill_grid
en
good
mini
5x5
intermediate
original
yes
cw_039
Why are word lists scored (5-60) instead of just filtered for validity?
(no artifact - pure methodology question)
Because validity is binary and quality is not. A fill tool with an unscored list grabs whatever fits, which is how the crosswordese economy arose: awkward patterns demand awkward letters, and ERIE/ETUI/EPEE absorb them - so unscored lists produce grids that fill but read like inventory. Scoring converts the fill from s...
methodology_qa
en
good
general
na
intermediate
original
no
cw_204
Review this fill submitted for the pinwheel mini. Check in dependency order: mask alignment first, then entry lengths, then words, then crossings. Identify every distinct rule violation with its cell coordinates, and give the minimal fix.
MASK: #.... #.... ..... ....# ....# FILL: #ORE# SALT# EAVE# APE## AT### WORD LIST: 3-letter: ACE, ACT, ADD, AGE, AGO, AID, AIL, AIR, ALE, ALP, ANT, APE, APT, ARC, ARE, ARM, ART, ASH, ASK, AUK, AVE, AWE, AXE, AYE, BAD, BAN, BAR, BAT, BAY, BED, BEE, BET, BIG, BIN, BIT, BOA, BOG, BOO, BOW, BOX, BOY, BUD, BUG, BUM, BUN...
Check 1 - mask alignment. The mask's open cells are a fixed set; the fill's blocks must coincide with the mask's blocks. They do not, in both directions: - fill blocks on open cells: (0,4), (1,4), (2,4), (3,3), (4,2), (4,3) - six cells of the grid erased. - a letter on a blocked cell: S at (1,0), where the mask has a b...
fill_grid
en
flawed
mini
5x5
beginner
original
no
cw_051
Практикум на русском: соберите за один проход всё, чему научила русская часть - возьмите пустую 4×4, пройдите заполнение вслух (MRV, когорты, каскады), затем проведите аудит и назовите три места, где человек ошибается чаще машины.
МАСКА: MASK: #... .... .... ...# WORD LIST: 3-letter: АГА, АИР, АКТ, АРА, АУЛ, БИТ, БОР, БЫК, ВАЛ, ВЕК, ВИД, ВОР, ГАД, ГАЗ, ГАМ, ГОД, ДАР, ДОМ, ДУБ, ДЫМ, ЕДА, ЖАР, ЖИР, ЖУК, ЗАЛ, ЗУБ, ИВА, КИТ, КОМ, КОТ, КУБ, ЛЁД, ЛАК, ЛЕВ, ЛЕС, ЛОБ, ЛОМ, ЛУК, ЛУЧ, МЁД, МАК, МИГ, МИР, МОХ, МУЛ, НИЗ, НОС, ОДА, ОКО, ОСА, ОЧИ, ПЁС, ПАР...
Заполнение вслух: старт - самая узкая когорта (трёхбуквенная, 90). Пробуем ЛУК: проверяем буквы-риски - Л, У, К с когортами 9, 16, 4 - всё глубоко, ставим. След: 1D на Л (9), 2D на У (16), 3D на К (4). MRV: 3D - четыре кандидата; КОМ ставит О и М в 4A и 5A (12 и 4). MRV: 5A - АТОМ (каскад: 1D 9→1 - ЛЕТО принуждено; 4D ...
methodology_qa
ru
good
mini
4x4
intermediate
original
yes
cw_265
A Japanese mini fails with 'crossing mismatch' errors in a kana-aware tool, and the constructor suspects the words are bad. Using three real entries - チーズ, チェリー, フォーク - diagnose the two encoding rules the fill must respect that are probably being violated, and state both as cell laws.
Three real Japanese answers and how the list decomposes them: チーズ (cheese): [チ, ー, ズ] チェリー (cherry): [チェ, リ, ー] フォーク (fork): [フォ, ー, ク] The tool's complaint: crossings disagree; the constructor insists the words are correct.
The constructor is right about the words and wrong about the encoding rules. Two laws are in play: LAW 1 - the long-vowel mark ー is its own cell. チーズ is four kana characters but three CELLS: [チ, ー, ズ], with ー occupying the middle cell like any kana. A fill that treats ー as a modifier attached to the previous kana will...
diagnose_failure
en
good
japanese
3x3
intermediate
japanese-convention
yes
cw_086
Below are the opening decisions of all three standard minis. Compare the candidate counts and explain what a trace's opening number tells you about the mask that produced it - and how a constructor can use opening counts to price a mask before committing to it.
Three 5x5 masks, three solver traces. PINWHEEL MASK: #.... #.... ..... ....# ....# Opening: 4D (468 candidates in play) = ADD CORNER MASK: #...# ..... ..... ..... #...# Opening: 1A (468 candidates in play) = AGO TWO-CORNER MASK: #.... ..... ..... ..... ....# Opening: 1D (1,372 candidates in play) = ALIVE (Word pool...
The pinwheel and the corner mask both open at 468 - both contain 3-letter entries, and an unconstrained 3-letter entry's candidate set IS the 3-letter pool. The two-corner mask opens at 1,372 because every entry in it is 4 or 5 letters: its smallest pool is the 5-letter one, three times the size. What the opening numb...
fill_trace
en
good
mini
5x5
advanced
original
no
cw_076
You need a 5x5 mini that a solver with a modest vocabulary can complete but that still feels dense. Choose between the three masks below by computing each one's stats and reasoning about fill difficulty and solve experience. State your choice and defend it in terms a puzzle editor would accept.
MASK A: #.... #.... ..... ....# ....# MASK B: #...# ..... ..... ..... #...# MASK C: #.... ..... ..... ..... ....#
Statistics, computed per mask: - A (pinwheel): words: 10, blocks: 4, lengths across 4,4,5,4,4 / down 3,5,5,5,3. Six 5-letter, four 4-letter... precisely: two 3s (down shoulders), four 4s, four 5s. rotationally symmetric: yes. unchecked squares: none. - B (corners): words: 10, blocks: 4, lengths 3,5,5,5,3 both direction...
design_grid
en
good
mini
5x5
intermediate
original
yes
cw_027
Финальная сборка русской 4×4 как публикуемого мини: маска + заполнение + заголовок + восемь подсказок + прогон всех шлюзов. Соберите комплект целиком и дайте вердикт «в печать».
МАСКА и ЗАПОЛНЕНИЕ: MASK: #... .... .... ...# FILL: #ЛУК ДЕЛО АТОМ РОВ# Заголовок-кандидат: «Дары лета». Подсказки: 1A «Овощ, от которого плачут»; 4A «Занятие, работа»; 5A «Частица материи из физики»; 6A «Канава вокруг крепости»; 1D «Время года между весной и осенью»; 2D «Добыча рыбака»; 3D «Комок, например земли»; 4D ...
Комплект по слоям: Клетки: 8 записей пере-выведены, словарь 8/8, пересечения 14/14, повторы нет, длины легальны. ✓ Структура: words: 8, blocks: 2, unchecked squares: none, rotationally symmetric: yes. ✓ Подсказки: восемь на восемь записей (каждая запись снабжена), грамматика сходится, подстановка проходит, утечек нет -...
validate_puzzle
ru
good
mini
4x4
intermediate
original
yes
cw_270
Why do constructors run 'death logs' on failed fills instead of just trying again with different settings?
(no artifact - pure methodology question)
Because retrying without diagnosis is superstition with a compute budget. The death log says WHICH entry hit zero candidates, at what depth, after which placements - which localizes the cause: depth 1 = a seed letter with an empty cohort; depth 2-4 = jointly unsatisfiable letters; a forced chain = a single-word depende...
methodology_qa
en
good
general
na
advanced
original
no
cw_208
Here is the complete recorded trace of the corner mini - eight decisions. Derive the full ten-entry fill from it: which words the trace gives directly, where the remaining two come from, and what the completed grid is.
MASK: #...# ..... ..... ..... #...# SOLVER TRACE: 1. 1A (468 candidates in play) = AGO: 1D narrowed 1372 -> 45; 2D narrowed 1372 -> 72; 3D narrowed 1372 -> 23 2. 3D (23 candidates in play) = ONION: 4A narrowed 1371 -> 117; 6A narrowed 1371 -> 101; 7A narrowed 1371 -> 81; 8A narrowed 467 -> 46 3. 1D (45 candidates in p...
Directly from the trace's decision column: 1A AGO, 3D ONION, 1D AHEAD, 8A DEN, 6A EERIE, 2D GORGE, 7A WAGON, 4A PHONY - eight of ten. The remaining two, 4D and 5D, never received decisions. Reconstruct them from the cells the decisions filled: - 4D runs down column 0: (1,0) holds PHONY's P, (2,0) holds EERIE's E, (3,0...
fill_trace
en
good
mini
5x5
intermediate
original
yes
cw_067
Методологический итог русской части: что русские строки курса наследуют от английского ядра и что в них принципиально своё? Составьте карту «переносится / параметризуется / своё».
Материал: эталонная 4×4 с её следом; случай гласно-начальной инженерии; невозможность 3×3; Ё-параметр; сканвордная традиция.
ПЕРЕНОСИТСЯ без изменений (ядро): клетки и маски; вывод записей; равенство букв на пересечениях; словарные проверки; запрет повторов; MRV-порядок с каскадами принуждений; след с его арифметикой (90-1=89 - родной брат английского 1,552-1=1,551); пере-вывод как единственное направление аудита. ПАРАМЕТРИЗУЕТСЯ: алфавит (...
methodology_qa
ru
good
general
na
advanced
original
no
cw_264
Вопрос читателю курса: зачем в следе русской 4×4 ровно семь решений, если записей восемь? Что проверяет существование «восьмого слова» и какой бухгалтерии это требует? Ответьте, не подглядывая в разбор тени.
СЛЕД: 7 решений; СЕТКА: 8 записей.
Ответ: седьмое решение (ДЕЛО на 4A) поставило букву Д в клетку (1,0), которая была последней открытой клеткой вертикали 4D - и 4D завершилась сама, словом ДАР, как теневой (побочный) эффект. Решений 7, записей 8: разница - тень. Что проверяет её существование: целостность бухгалтерии «решения + тени = записи». Теневое...
methodology_qa
ru
good
mini
4x4
intermediate
original
yes
cw_273
A model filled 'the same' two-block mini as the constructor's reference grid, but its output disagrees with the reference everywhere. Diagnose: did the model see the same problem at all? Explain the failure category and the fix.
MASK: #.... ..... ..... ..... ....# MODEL OUTPUT: ABETX GLARE RISEN EVICT WELTX Reference fill (the constructor's verified grid): FILL: #ABET GLARE RISEN EVICT WELT#
Compare the model output to the mask cell by cell. The mask has blocks at (0,0) and (4,4) - 23 open cells. The model's grid: - Row 0 reads A-B-E-T-X. The A sits ON the block at (0,0); the true 1A lives at (0,1)-(0,4), so the whole word drifted one cell left, and the X lands in the open cell (0,4) that should hold ABET'...
fill_grid
en
bad
mini
5x5
intermediate
original
no
cw_054
Why does clue-writing have a SUBSTITUTION test? Can't a good editor just feel whether a clue works?
(no artifact - pure methodology question)
Feel is exactly what the test exists to replace - because the two documented clue-failure families feel FINE. Fluent lies ('Church official' for PEW) read like good clues; grammar mismatches ('Angrily stared' for GLARE) read like good clues; both strand solvers and both pass any amount of editorial vibing. The substitu...
methodology_qa
en
good
general
na
beginner
original
no
cw_212
Why does the solver keep a used-words set at all? The crossing constraints already prevent most repeats.
(no artifact - pure methodology question)
They do not prevent them: two entries that share no cells can legally hold the same word, and even crossing-constrained entries can converge on it (the ARENA/CIVIC/EERIE corner state forced BOTH corner downs to ACE - crossings not only permit the repeat, they demand it). The used-set is the enforcement of the no-duplic...
methodology_qa
en
good
general
na
intermediate
original
no
cw_202
次の三つのカギはどれも一見正しそうで、どれも偽物です。置換テストで 三つとも裁き、日本語のカギにありがちな「うそ」の二つの発生源を示してください。
カギ1 (答え アサ, 2 マス): 「アサヒビールの略」 カギ2 (答え サカナ, 3 マス): 「寿司の上に乗るもの — 具体例としては マグロやサーモン」 カギ3 (答え メシ, 2 マス): 「イエの中の一室」
カギ1 の検死。「アサヒビールの略」を置換テストにかけます: 答え アサヒ? — それなら 3 マス。答え アサはアサヒの略としては不完全で、 ビール会社の名を 2 マスに縮める慣用もありません。このカギの病は 「長さのつりあい」です — 3 マスの語を暗示するカギで 2 マスの答えを 要求する。枚挙の一致はカギの契約の第一条で、意味が合ってもマスが合わなければ失格です。 カギ2 の検死。「マグロやサーモン」は実例として正しい — しかし置換 テスト: マグロ は 3 マスの答えになれるか? プールを調べると マグロ は在庫にある! するとこのカギはサカナとマグロの二つの正解を抱えて しまい、単独では答えを固定できません。2D の...
write_clues
ja
good
mini
3x3
intermediate
original
yes
cw_393
把一个字钉进角落，格盘还能不能活？答案完全取决于这个字在词典两端的队列。请用真实队列表讲清「枢纽不对称」：哪个字能当种子，哪个字看着体面却一钉就死。
队列表（首字队列 / 末字队列，词池真实数据）: 人: 5 / 4 中: 4 / 1 国: 4 / 2 大: 4 / 1 学: 3 / 7 文: 2 / 1 民: 3 / 2 工: 1 / 3 员: 1 / 2 子: 2 / 8 天: 2 / 5
队列是格盘的燃料。「人」字打头的双字词有 5 个（人们、人员、人工、人民、人生），收尾的双字词有 4 个——两端都富裕，所以「人」是全池最强的枢纽之一，两张目录格盘都从「人」或经它起步。「学」更极端：打头 3 个、收尾 7 个，几乎专为交叉而生。 再看反面。「员」打头的词只有「员工」自己，收尾倒有 2 个（人员、店员）——单向枢纽。把「员工」钉在 1A：格子 (0,0) 变成「员」，纵向 1D 需要一个「员」打头的词，可全池只有「员工」，而它已被种子占用。机器一步就判死：节点数 1，零候选，连一次尝试都没有。「工」同样单薄：打头只剩「工人」。 「中」是个精细的反例。打头 4 个（中国、中学、中文、中间）、收尾仅 1 个，两端不平...
design_grid
zh
good
general
na
advanced
original
yes
cw_282
Complete this corner-block mini. Explain each placement decision in terms of the constraint it removes for the crossing entries, and end with the full fill plus a crossing audit of one entry of your choice.
MASK: #...# ..... ..... ..... #...# WORD LIST: 3-letter: ACE, ACT, ADD, AGE, AGO, AID, AIL, AIR, ALE, ALP, ANT, APE, APT, ARC, ARE, ARM, ART, ASH, ASK, AUK, AVE, AWE, AXE, AYE, BAD, BAN, BAR, BAT, BAY, BED, BEE, BET, BIG, BIN, BIT, BOA, BOG, BOO, BOW, BOX, BOY, BUD, BUG, BUM, BUN, BUS, BUT, BUY, CAB, CAN, CAP, CAT, ...
Decision-by-decision, in most-constrained-first order: 1. 1A (468 candidates in play) = AGO: 1D narrowed 1372 -> 45; 2D narrowed 1372 -> 72; 3D narrowed 1372 -> 23 2. 3D (23 candidates in play) = ONION: 4A narrowed 1371 -> 117; 6A narrowed 1371 -> 101; 7A narrowed 1371 -> 81; 8A narrowed 467 -> 46 3. 1D (45 candidates...
fill_grid
en
good
mini
5x5
intermediate
original
yes
cw_032
A language model was asked to 'reproduce the solver trace of the pinwheel mini' and produced the log below. Audit it: what exactly did the model do right, what did it fabricate, and what does the failure pattern reveal about how it generated the text?
MASK: #.... #.... ..... ....# ....# MODEL-PRODUCED TRACE: 1. 4D (468 candidates in play) = ADD: 1A narrowed 1,552 -> 17 2. 1A (17 candidates in play) = PLAE: 1D narrowed 1,372 -> 89; 2D narrowed 1,372 -> 56 3. 3D (32 candidates in play) = ERSAE: 5A narrowed 113 -> 13; 6A narrowed 86 -> 11 WORD LIST: 4-letter: ABET,...
The numbers are real. The words are not. - Step 2 places PLAE. No such word exists in the list - or in English. It is an anagram-ghost of PLEA, letter-adjacent, count-identical. - Step 3 places ERSAE - a letter-permutation of ERASE, equally void. The giveaway is structural: the narrowings quoted after PLAE (1D to 89 =...
fill_trace
en
bad
mini
5x5
advanced
original
no
cw_073
A model produced this clue set for the pinwheel fill. The surfaces are decent; the counts are disaster. Audit the enumeration layer, connect the failure to the fill module's letter-count-blindness case, and state the process fix.
TRUE cell counts: PLEA (4), HERD (4), NOMAD (5), ONUS (4), WERE (4), PHONE (5), LEMUR (5), ERASE (5), ADD (3), NOW (3). THE MODEL'S SET (with its counts): 1A PLEA: 'Courtroom appeal' (5) 5A HERD: 'Group of cattle' (4) 6A NOMAD: 'Wanderer with no fixed home' (7) 7A ONUS: 'Burden of responsibility' (3) 8A WERE: 'You ___ ...
The enumeration audit, answer by answer: PLEA promised 5, true 4 - WRONG. HERD 4 - right. NOMAD promised 7, true 5 - WRONG. ONUS promised 3, true 4 - WRONG. WERE 4 - right. PHONE promised 4, true 5 - WRONG. LEMUR promised 6, true 5 - WRONG. ERASE 5 - right. ADD 3 - right. NOW promised 2, true 3 - WRONG. Six of ten coun...
write_clues
en
bad
mini
5x5
intermediate
original
no
cw_131
一份待审计的谜题包：线索组属于一张格盘，格盘却是另一张。请执行三向审计——线索↔词条↔格盘——找出所有漂移，并说明哪个方向的重推导最先暴露问题。
格盘（已机器验证）: FILL: 人民 员工 随附线索组: 1A 语言学，方块字写成的母语 (2字) 3A 中国传统学术之学 (2字) 1D 中华人民共和国 (2字) 2D 一种文学体裁，韵文之外皆可称此 (2字)
三向审计从最机械的一向开始：从格盘重推导词条。逐格读出：1A 人民、3A 员工、1D 人员、2D 民工——四条词条与随附线索组逐条对照：线索组里没有一条提到人民、员工、人员、民工。 第二向：线索反推答案。「语言学，方块字写成的母语」指向「中文」；「中国传统学术之学」指向「国学」；「中华人民共和国」指向「中国」；「韵文之外皆可称此」……指向「文学」。四条线索的答案集合是 {中文、国学、中国、文学}——这恰好是词组甲，目录里的另一张格盘。漂移现形了：这份线索组是为词组甲（行「中文/国学」、列「中国/文学」）写的，却被装进了词组乙（人民/员工/人员/民工）的格盘包里。编号也对不上：线索组把「中文」标给 1A、「国学」标给 3A——词组甲...
diagnose_failure
zh
good
mini
2x2
advanced
original
yes
cw_305
Буква Ё в русских сетках: живая традиция часто нормирует Ё→Е, а наш словарь содержит Ё-слова (ЁЖ, ЁЛКА). Изложите это как ПАРАМЕТР дисциплины: что нужно объявить проекту, чтобы выбор был честным, и как машина учитывает Ё при пере-выведении.
Факты словаря курса: ЁЖ (2 буквы), ЁЛКА (4 буквы) - полноценные словарные статьи с клеточным разложением. Факт традиции: во многих русских кроссвордах Ё печатается как Е.
Что нужно объявить: (1) АЛФАВИТ сетки - какие буквы различимы: если Ё=Е, слово ЁЖ занимает клетки Е-Ж и любое ЕЖ-слово пересекается с ним по общим правилам; если Ё - самостоятельная буква, клетка с Ё не может быть Е ни в одной пересекающейся записи. (2) СЛЕДСТВИЕ ДЛЯ КЛЮЧЕЙ: при Ё=Е ответ печатается с Е; при самостояте...
methodology_qa
ru
good
general
na
intermediate
original
no
cw_250
Two constructors each locked a 'great word' into the bottom row (8A) of this grid and the autofill died instantly for both. Diagnose both deaths precisely - which entry died, which letter killed it, and why the deaths took only two search nodes apiece.
MASK: #.... ..... ..... ..... ....# Seeds tried, each fatal within one placement: Attempt 1: 8A = QUIZ Attempt 2: 8A = JUMP Autofill report, both attempts: 'entry 5D has zero candidates' after 2 search nodes. WORD LIST: 4-letter: ABET, ABLE, ACED, ACID, ACNE, ACRE, AIDE, AJAR, ALOE, ALTO, AMEN, AMID, AMPS, ANTE, A...
Which entry died: 5D, the 4-letter down in column 0, whose cells are (1,0), (2,0), (3,0) and (4,0) - the last one being 8A's first cell. Which letter killed it: the FIRST letter of each seed. QUIZ puts Q at (4,0); JUMP puts J at (4,0). So 5D's pattern becomes ???Q in the first run and ???J in the second. Checking the ...
diagnose_failure
en
good
mini
5x5
intermediate
original
yes
cw_089
Below are the final decisions of the three standard minis. Compare how each grid ends: what the trace's last line shows, what it cannot show, and the complete audit an ending requires.
PINWHEEL, final step: 8. 7A (1 candidate in play) = ONUS: 6D narrowed 15 -> 4 9. 8A (3 candidates in play) = WERE CORNER MASK, final step: 7. 7A (1 candidate in play) = WAGON: 4D narrowed 63 -> 8; 5D narrowed 63 -> 8 8. 4A (3 candidates in play) = PHONY TWO-CORNER MASK, final step: 8. 6A (1 candidate in play) = RISEN...
All three endings share a shape: a late forced cascade (a 1-candidate placement right before the end), then a final small-count decision (3, 3, 2 candidates) whose line shows NO effects - and each final placement completed entry or entries the line never mentions: - Pinwheel: WERE's W at (4,0) side-completed 6D = NOW. ...
fill_trace
en
good
mini
5x5
advanced
original
no
cw_077
The side-effect ('shadow') entries deserve their own validation step: entries the trace never shows, completed by the final placement's letters. Demonstrate the shadow audit on the corner mini and the pinwheel - name each grid's shadows and the exact checks they need.
From the trace module's findings: the corner grid's PHONY placement side-completed BOTH 3-letter corner downs (4D, 5D); the pinwheel's final WERE placement side-completed 6D; the two-corner grid's GREW placement side-completed 8A.
The shadow audit, grid by grid: - Corner mini: the shadows are 4D (P-E-W: PHONY's P, EERIE's E, WAGON's W) and 5D (Y-E-N). Checks: each is a list word (PEW, YEN - yes); neither duplicates another entry (no); lengths legal (3, 3 - yes). PASS. - Pinwheel: the shadow is 6D (N-O-W, completed when WERE's W landed at (4,0))....
validate_puzzle
en
good
mini
5x5
intermediate
original
yes
cw_189
Capstone: re-clue the ENTIRE pinwheel at its honest maximum difficulty - and discover where the ceiling is. Write all ten clues, distribute misdirection by the see-saw, flag what you use, and state the distribution verdict the exercise teaches.
Verified pinwheel fill: PLEA, HERD, NOMAD, ONUS, WERE / PHONE, LEMUR, ERASE, ADD, NOW. Every entry fully crossed. Target: as hard as honesty allows.
1A PLEA (4): 'Cry before sentencing?' 5A HERD (4): 'Do shepherd's work?' 6A NOMAD (5): 'Pastoral mover?' 7A ONUS (4): 'It's on you, so to speak?' 8A WERE (4): 'We ___ young (1965 hit)' 1D PHONE (5): 'Cell mate?' 2D LEMUR (5): 'Julien's kind, in Madagascar films' 3D ERASE (5): 'Hit delete on' 4D ADD (3): 'Throw in' 6D N...
write_clues
en
good
mini
5x5
expert
original
no
cw_144
Explain the NYT word-count maxima - 78 words for a themed 15x15, 72 for a themeless 15x15, 140 for a 21x21 Sunday - from the solver's side: what does a higher word count physically mean about entry lengths, and why is a themed grid allowed to be denser than a themeless one?
Question: derive and justify the NYT word-count maxima.
Word count is an inverse proxy for entry length. A 15x15 has 225 cells; at 78 words the average entry spans about 2.7 cells per direction of travel... more usefully: white cells per word. Roughly 186-196 white cells (at 16-18 percent blocks) divided by 78 words is about 2.4 letters per across-and-down share - in practi...
design_grid
en
good
daily
15x15
advanced
nyt-specs
no
cw_011
Two decisions into the two-block mini's fill. Scan all eight open entries, predict the solver's next move, and confirm against the recorded step 3.
MASK: #.... ..... ..... ..... ....# TRACE STEPS 1-2: 1. 1D (1372 candidates in play) = ALIVE: 1A narrowed 1552 -> 22; 5A narrowed 1371 -> 124; 6A narrowed 1371 -> 121; 7A narrowed 1371 -> 5; 8A narrowed 1552 -> 225 2. 7A (5 candidates in play) = EVICT: 5D narrowed 1552 -> 129; 2D narrowed 1370 -> 101; 3D narrowed 1370...
Open entries with used words {ALIVE, EVICT} excluded: - 1A (A???): 22. - 5A (?L???): 124. - 6A (?I???): 121. - 8A (???E): 225. - 2D (???I?): 101. - 3D (???C?): 93. - 4D (???T): 191. - 5D (??E?): 129. The smallest is 1A at 22 - the A that ALIVE planted at (0,1) is doing the steering. The recorded step 3: '1A (22 candid...
fill_trace
en
good
mini
5x5
advanced
original
yes
cw_066
Review this argument. A constructor wants to ship the 4-entry mask below (three 5-letter across rows and one 5-letter down column) and answers the reviewer's unchecked-cell objection as shown. Which part of the argument survives scrutiny, and which part is a category error?
MASK: ..... ##.## ..... ##.## ..... THE CONSTRUCTOR'S ARGUMENT: 'Fourteen unchecked cells? So what. The publishing tool has no hard gate on unchecked cells for small grids - it fills, it clues, it ships. Every word I've written into it is real, and it's MY puzzle: the gates exist to stop mistakes, and this is not a mi...
The part that survives: the gates really do not stop this. Publish gates check that the grid is filled, titled, and fully clued - unchecked cells draw warnings at most, and on small grids even the warnings are gentle. The constructor is factually right about enforcement. The part that is a category error: 'the gates p...
diagnose_failure
en
flawed
mini
5x5
intermediate
original
yes
cw_100
Диагноз неудачи: русская 3×3 с маской, которая работает в английском, здесь НЕ заполняется вовсе. Исчерпывающий перебор - 38 узлов. Объясните структурную причину и назовите направление починки.
МАСКА 3×3: MASK: ..# ... #.. Факты словаря: 11 двухбуквенных слов, все начинаются с гласной (ЁЖ, УЖ, УМ, АД, ЯК, АХ, ОХ, УХ, ЭХ, ОЙ, АЙ); их вторые буквы - только Ж, М, Д, К, Х, Й; трёхбуквенных слов 90. Перебор исчерпан: 38 узлов, решений нет.
Требования маски: шесть записей - четыре двухбуквенных (углы) и два трёхбуквенных (средние ряды). Пересечения требуют, чтобы вторые буквы угловых слов стали первыми буквами трёхбуквенных, а первые буквы углов (гласные!) согласовались с чужими началами. Структурная причина провала: весь двухбуквенный слой русского слов...
diagnose_failure
ru
good
mini
3x3
advanced
original
yes
cw_242
Write the historical note on Margaret Farrar and the founding of the editorial standards: who she was, which rules she set, and why they stuck for a century.
(no artifact - public-domain history question)
Margaret Farrar was the first crossword editor of The New York Times - joining when the paper finally embraced the puzzle in the early 1940s - and before that a defining hand at other New York papers through the crossword craze's first wave. She is the source of the standard this dataset teaches as the American canon: ...
historical_note
en
good
general
na
beginner
historical
no
cw_227
The corner mini's trace, four decisions in. Scan the open entries, predict step 5 - which entry, how many candidates - and note the detail in 4D's and 5D's counts that this exercise is really about.
MASK: #...# ..... ..... ..... #...# TRACE STEPS 1-4: 1. 1A (468 candidates in play) = AGO: 1D narrowed 1372 -> 45; 2D narrowed 1372 -> 72; 3D narrowed 1372 -> 23 2. 3D (23 candidates in play) = ONION: 4A narrowed 1371 -> 117; 6A narrowed 1371 -> 101; 7A narrowed 1371 -> 81; 8A narrowed 467 -> 46 3. 1D (45 candidates i...
Scanning the open entries, with used words {AGO, ONION, AHEAD, DEN} excluded: - 2D (G???E): 23. - 4A (?H?N?): 13. - 6A (??EI?): 11. - 7A (??O??): 15. - 4D (???): 466 - note this is NOT 468: the 3-letter pool minus the used AGO and DEN leaves 466. The used-words rule subtracts across the whole grid, entry by entry. - 5D...
fill_trace
en
good
mini
5x5
advanced
original
yes
cw_065
Лестница трудности на одном ответе: напишите подсказки для ЛУК в трёх градациях - простую, среднюю и сложную - и назовите механизм, который добавляется на каждой ступени. Подсказка: у ЛУК два главных смысла, и это ваш рабочий материал.
Ответ: ЛУК (3). Словарные смыслы: огородное растение, заставляющее плакать при нарезке; метательное оружие стрелка (лук со стрелами).
ПРОСТАЯ: «Овощ, от которого плачут». Один смысл, один шаг, ноль перечитываний. СРЕДНЯЯ: «Стрелок зарядил его стрелой - а в суп пошёл другой». Два смысла соединены в сцену; решатель должен увидеть, что предмет один и тот же по написанию, но не по смыслу. Одно перечитывание, честный переход между значениями. СЛОЖНАЯ: «...
write_clues
ru
good
mini
4x4
intermediate
original
no
cw_246
Audit this clue's indicator honesty. The setter insists 'calm' works as 'a gentle sort of anagram signal'. Test the claim against the letters and against the lexicon.
Clue: 'Calm LEAP for a court appeal (4)' Intended answer: PLEA
The letter test: perform the promised operation. 'Calm' signals... what? If it signals anagram: LEAP scrambled gives PLEA - the letters permit it. If it signals nothing: LEAP is just LEAP, and the clue reduces to a false definition ('LEAP for a court appeal'?). The letters alone cannot convict it - the LEXICON must. T...
cryptic_clue
en
flawed
cryptic
na
intermediate
original
no
cw_165
The mask below keeps every row at least three cells long but still fails a basic publishability check. Identify the failure, then repair the mask by moving the smallest possible number of blocks. Give the repaired mask and its statistics.
MASK: ..#.. ..#.. ..... ..#.. ..#..
The failure is in the columns. Column 2 is blocked at rows 0,1,3,4 - leaving only (2,2) open. A single cell cannot start an entry (its neighbor below is a block), so (2,2) is unchecked, and columns 0,1,3,4 each run the full height: the down slots are 5,5,-,5,5 while across slots are 2,2,5,2,2 - the four 2-letter rows v...
design_grid
en
good
mini
5x5
intermediate
original
yes
cw_007
成语是中文格盘最体面的词条。请为「一举两得」「守株待兔」「画蛇添足」各写一条情境式线索，并解释为什么情境式线索比定义式更适合成语。
目标词条（均在词池中，各 4 字）: 一举两得 / 守株待兔 / 画蛇添足
线索组：一举两得——「做一件事，同时收获两份好处」；守株待兔——「守在树桩旁，等兔子自己撞上来——讽刺不肯努力、指望意外的人」；画蛇添足——「蛇画完了，偏要再添几只脚——比喻多余的举动反而坏事」。 为什么情境式优先？因为成语的语义是寓言整体的函项，不是字义的加总。「守株待兔」四个字逐字看毫无信息（守、株、待、兔——「株」字单独还难倒一片），可寓言一讲，答案自己浮上来。定义式线索（「比喻不主动努力」）反而更难：近义成语一抓一把（坐享其成、缘木求鱼、胶柱鼓瑟），定义宽了收不了口，定义窄了变成百科问答。情境线索的收口机制不同：它给的是故事的指纹——树桩、撞上来的兔子、添上的脚——指纹对上了，答案就唯一了。 成语线索还有两道装配工序。第...
write_clues
zh
good
general
na
advanced
original
yes
cw_298