N1 2022-07 — audit report

68 questions (問題1–13, the non-listening half). Generated from the OCR and correction audit trails.

OCR confidence

Source: retypeset-digital — quality 3/5: Community recall version (回忆版) compiled by the Vietnamese 'Tôi Yêu Ngoại Ngữ / Yuuki Bùi' group, fully retypeset as clean digital text — everything is legible and most pages are crisp. Marked down for recall fidelity rather than legibility: two written-half questions were not recalled (shifting all numbering after Q40), several retype typos (い2・3・4, 3月81日, 追奈, 注6 bracket), kana substitutions in listening options, and the group's branding injected into actual question content (YUUKI遊園地 in Q32, YUUKI社 in Q40, a Facebook URL inside the 問題13 stimulus, YUUKIBUI cell in the listening 問題4 grid). Every page carries a diagonal watermark and a Telegram QR code; page 21 is pure advertising.

Every section was transcribed twice independently (a full-page pass and a double-resolution half-page pass) and reconciled. Of 13 sections, 4 agreed on all content between the two passes (reconciled automatically, differing only in transcriber notes) and 9 had at least one content difference resolved by a third adjudication pass that re-read the page images. 121 leaf-level differences were examined in total.

Residual OCR risks (30):

• 問題1: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題2: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題3: Q15 answer option 3 「がっかりして」 and option 4 「ほっとして」 read clearly at high res, but the heavy retypeset font makes っ vs つ distinctions low-stakes here; no anomaly observed.
• 問題3: The page is a recall reconstruction, so wording could contain recall typos beyond the flagged ones; transcription preserves what is printed rather than the official exam text.
• 問題3: Q18 (手分け→bare nouns) and Q19 (mixed adverbial/dictionary-form options) are structurally unusual for a real N1 問題3 and are most likely recall-version artifacts, but they are faithfully transcribed.
• 問題4: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題5: The 人 vs 入 glyph distinction relies on the absence of the top diagonal stroke in a watermarked retypeset render; rendering confirms 人 consistently in both 問題5 and 問題6 instructions, so confidence is high but not absolute.
• 問題5: Source is a community recall reconstruction (回忆版); wording may differ from the official exam regardless of OCR fidelity (e.g. the し omission in q33 option 4 may be a recall error rather than a printing one). Transcribed as printed per convention.
• 問題6: The source is a recall reconstruction, so q40 option 2 「コンサートで」 and the YUUKI社 / YUUKI遊園地 branding may diverge from the genuine 2022-07 exam wording; transcribed verbatim as printed per instructions.
• 問題6: The instruction line is heavily garbled in the retype (misplaced に, い for 1・, 人る for 入る); 人る vs 入る rests on a kaisho-glyph judgment from the render — high confidence it is 人, but the underlying intended character is certainly 入る.
• 問題7: The 注 colon gap was kept as one full-width space because a deliberate ~1-em gap is clearly printed and 'transcribe exactly as printed' governs; however the convention's canonical gloss example shows no space after the colon, so a stricter reading could drop it.
• 問題7: 「低齢児」 is preceded by a single typed half-width space (kept); whether the original intended 低年齢児 with no space cannot be recovered from the recall reconstruction.
• 問題7: Source is a community 回忆版 retype with many embedded typos; values were transcribed faithfully but cannot be cross-checked against an official paper.
• 問題8: Q48 option 3 ・: the glyph is unambiguously a printed mid-line dot in both double-resolution halves, but whether the original source intended an interpunct vs. a stray retype artifact cannot be known; transcribed as printed (・) per convention.
• 問題8: Several passages contain probable recall/retype typos (か for が; missing で/では; doubled な; 一万 for 一方; missing を). All are transcribed verbatim and flagged; the underlying community 回忆版 may differ from the genuine exam.
• 問題8: No diagonal-watermark or branding injection (e.g. YUUKI遊園地, facebook URLs) was found within 問題8 content; those known injections belong to other sections/questions and do not affect mondai8.
• 問題9: Source is a community recall reconstruction (回忆版) with multiple recall typos (決めたれた, 役に立っ, 考こ, ことか大切た, 探し出すへきだ, 型身につけ, サラグ油, パンコン×2). These are transcribed verbatim per convention; the underlying authentic exam wording cannot be recovered from this image and may differ.
• 問題9: No injected branding (YUUKI遊園地 / YUUKI社 / facebook URL) appears within 問題9 itself — those belong to other sections (q32/q40/問題13); none was found inside this section's content, so none was transcribed here.
• 問題9: Faint dakuten/handakuten and small kana on the diagonally watermarked retype are inherently low-contrast; the two corrected glyphs (立っ, へきだ) were each confirmed on the double-resolution half-page renders, but other unflagged dakuten could in principle be marginal.
• 問題10: The passage carries multiple genuine source-retype typos (自間/学間/入/捉えらななる/ゆかめて); these are deliberately preserved as printed and were each glyph-verified, but a downstream consumer expecting clean kanji should be aware.
• 問題10: Watermarks and QR codes overlap the passage on the source renders; readings were taken where glyphs were unobstructed and cross-checked across page halves, but a few strokes sit under faint watermark diagonals.
• 問題10: 注6 missing closing parenthesis is reproduced as printed; if the project later decides to auto-close such parens this entry would change.
• 問題11: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題12: DNA spacing: I read the passage occurrence as tight ('DNAが') but the answer options as spaced ('DNA が'/'DNA に'). This passage-vs-answer inconsistency is plausible for a hand-retyped source but rests on a fine spacing judgement at the available resolution; a higher-res scan could overturn it.
• 問題12: The ②　カイコ gap was encoded as exactly one full-width space; the printed gap is roughly one cell but its exact width is a judgement call.
• 問題12: 乍 vs 作 in q66 option 1 relies on the absence of the 亻 radical at the rendered resolution; confident but not certain.
• 問題12: Numerous source recall typos (こくこく希に, ところが。, ・。, 渡い, ①のを) are preserved verbatim per convention and were not corrected.
• 問題13: Boldness of the two title lines is a visual judgment from the render; they are clearly heavier than body text but a faint anti-aliased retype could in principle be just centered-not-bold. Marked bold per the stronger reading.
• 問題13: The 81 in 3月81日 and グウンロード and the missing opening paren in item ④ are preserved as printed source typos; if these are render artifacts rather than the source's own errors, they would need correction, but all three were verified at high zoom by both transcribers and match the images.
• 問題13: The single full-width gap in 同年度内　に is a judgment call about whether the printed gap is a real space or auto-spacing; both transcribers and the final treat it as one full-width space.

Source-text quirks the OCR transcribed verbatim (printed typos/recall artifacts; the ones judged to be retype errors were corrected in the layer below) — 86 noted at OCR time:

• 問題1: Question 6 underlines the conjugated form 阻まれて, while all four answer options are dictionary forms (かこむ／からむ／こばむ／はばむ) rather than 〜まれて forms; transcribed as printed.
• 問題1: Question 6: the underlined word is the inflected form 阻まれて, but all four answer options are printed in dictionary form (かこむ／からむ／こばむ／はばむ) rather than the expected 〜まれて readings; transcribed as printed (likely a recall-reconstruction simplification).
• 問題2: Q12 stem ends in polite form 「になります。」 while all other stems in this section use plain form — possibly a recall/retype artifact; transcribed as printed.
• 問題2: The printed instruction and question blanks use a narrow （　） in this retypeset; normalized to （　　） per convention.
• 問題2: Question 9 option 3 verified as 干渉 (干, not 千).
• 問題3: Source typo in the 問題3 instruction line: it prints 「…ものを、い2・3・4から一つ選びなさい。」 — hiragana 「い」 where the standard wording would have 「1・」 (the digit 1 was apparently retyped as い). Verified against the double-resolution render and transcribed as printed (い2・3・4).
• 問題3: Q15: the printed underline covers only 「閉口して」; いる is outside the underline. Verified at high resolution (the underline terminates under て, before い).
• 問題3: Q18: the underlined headword is 手分け in 「手分けした」 (した is outside the underline), while the four answer options are bare nouns (分割／分別／分担／分類). Likely a recall-version simplification; transcribed as printed.
• 問題3: Q19: the printed underline starts at the particle を, covering 「をてきぱきと」 rather than just 「てきぱきと」 (ends at と, before 処理). Transcribed as printed.
• 問題3: Q19: answer options 3 (張り切る) and 4 (嫌がる) are dictionary-form verbs while options 1 (早く正確に) and 2 (時間をかけて丁寧に) are adverbial phrases — likely recall-version paraphrasing; transcribed as printed.
• 問題3: This page is a community recall reconstruction (回忆版) by the Vietnamese study group Tôi Yêu Ngoại Ngữ / Yuuki Bùi; diagonal watermarks ("Tôi Yêu Ngoại Ngữ Group", "Yuuki Bùi"), the header "2022 年 7 月 N1 YUUKI BUI", the section tab, the QR/Telegram badge, and the page number were ignored. No injected branding appeared inside the 問題3 question content itself.
• 問題4: q21 answer 2: the personal name reads 「符由氏」 — an implausible Japanese surname, likely a recall/retype corruption of the original name; transcribed as printed.
• 問題4: q22 answer 2 begins 「機の点検のため」 — likely a dropped character in the retype (presumably 機械); transcribed as printed.
• 問題4: q21 option 2: the source prints the name 「符由氏」 (verified at high zoom); this looks like a recall-reconstruction invention or typo for some original name — transcribed as printed.
• 問題4: q22 option 2 begins 「機の点検のため」 (verified at high zoom); likely a recall typo for 「機械の点検」 — transcribed as printed.
• 問題4: q24: underline extents vary by option as printed: ぎこちなく / ぎこちない / ぎこちない / ぎこちなくて.
• 問題5: The printed 問題5 instruction reads 「に人れるのに」 with 人 instead of 入 — apparent retype typo in the source, transcribed as printed. The adjacent 問題6 instruction has the same 人/入 substitution (人る for 入る), confirming a consistent source misprint.
• 問題5: q32 contains the study group's injected branding 「YUUKI遊園地」 in the stem; transcribed as printed.
• 問題5: q33 option 4 is printed as 「てはいられない」 without a leading し (the parallel options 1-3 all start with する/して — してはいられない would be the expected form); likely a recall/retype omission, transcribed as printed.
• 問題5: q34 B's line 「え、私も思い出せない」 has no sentence-final 。 in print; transcribed as printed.
• 問題6: The printed 問題6 instruction line is garbled in this retype: it reads 「次の文のに ★ 人る最もよいものを、い2・3・4から一つ選びなさい。」 — the に is misplaced before the starred blank instead of after it, 入る is mistyped as 人る (verified against the double-resolution p03-b render: the glyph is the two-stroke 人, not 入), and the leading 1・ of 1・2・3・4 is mistyped as い. Transcribed exactly as printed.
• 問題6: Q40 option 4 contains the study group's injected branding 「YUUKI社」 (expected per known branding injections); transcribed as printed.
• 問題6: Q40 option 2 reads 「コンサートで」, which is semantically odd in a sentence about a ものづくりコンテスト — possibly a recall/retype error, but transcribed as printed.
• 問題7: Instruction line contains a recall typo 「中に人る」 (人, person, not 入); verified character-by-character at high zoom against the genuine 入 glyphs in the q43 answer choices. Transcribed as printed.
• 問題7: Instruction prints 「文章全体の内容を考えて」 where official papers use 「文章全体の趣旨を踏まえて」; transcribed as printed.
• 問題7: Numerous suspected retype typos verified at high zoom and kept as printed: 「使われていることはだと思うのですが」 (for ことばだ), 「0威から五成（就字前まで）」 (for 0歳から五歳（就学前まで）), 「五成児までいる」 (for 五歳児), 「目分は小さい子」 (目 for 自; the paired 「自分は一番大きい子」 in the same sentence does use 自分), 「環境て過ごす」 (て without dakuten for で — verified at 14x zoom; the earlier 「環境で生活をする」 in the same sentence does use で), 「思つのす。」 (for 思うのです。), 「進字したとき」 (for 進学した), 「年長児（注1）たったときは」 (たった for だった), 「少さなお友だち」 (少 for 小), 「面側を見てもらう」 (for 面倒を見てもらう).
• 問題7: Cloze blanks 41/42/43 are each printed boxed and immediately followed by a comma — including the ungrammatical 「もあります。【41】、に行って驚くのは」; transcribed as printed. 44 has no comma (「なんとも【44】。」).
• 問題7: A clear typed half-width space is printed between 「二歳児までの」 and 「低齢児」 (likely a dropped 年 from 低年齢児); kept as a single half-width space and the 低齢児 spelling as printed.
• 問題7: Paragraph 8 ends with three ideographic full stops 「かわいいのですが。。。」; kept as printed (they are 。 glyphs, not the … ellipsis), so the … normalization was not applied.
• 問題7: 「ぼくはわたしは大きくなったのに！」 has no separator between ぼくは and わたしは, as printed.
• 問題8: Passage (1): printed 「メッセージかシンプルな番組のほうが」 — か where が is expected; transcribed as printed (likely recall/retype typo).
• 問題8: Passage (3): printed 「探求への意識関心がないところ発生することもありません」 — reads oddly (apparently missing で/では after ところ, and 意識関心 run together); transcribed as printed.
• 問題8: Question 47, option 4: printed 「問題への探求心持ち続けていなければ」 — apparently missing を after 探求心; transcribed as printed.
• 問題8: Question stems for 46 and 47 (筆者の考えに合うのはどれか) are printed without a trailing 。, while 45 and 48 end with 。; transcribed as printed.
• 問題8: Passage (4): printed 「嫌な不愉快なな「思い出」」 — doubled な confirmed at high zoom; transcribed as printed (likely recall/retype typo).
• 問題8: Passage (4): printed 「一万の楽しい愉快な「思い出」」 — the glyph is clearly 万, likely a typo for 一方の; transcribed as printed.
• 問題8: No source citations are printed for any of the four passages (the recall retype omits them).
• 問題9: Passage (1): 「サラグ油」 printed (suspected retype typo for サラダ油); confirmed at high zoom, dakuten clearly on グ.
• 問題9: Passage (1): a run of four middle dots 「・・・・」 follows 「三、溶き卵を入れる。」; kept as printed since it is a nakaguro run, not 3-dot ellipsis groups.
• 問題9: q49 option 3: 「決めたれた理由」 printed (suspected typo for 決められた); transcribed as printed.
• 問題9: q50 option 3: 「役に立っとは言えない」 printed with a small っ (suspected typo for 役に立つとは); confirmed against p08-t/p08-b — A had 立つ, corrected to B's 立っ.
• 問題9: q51 stem: 「筆者の考こ合うのはどれか」 printed (suspected typo for 筆者の考えに合うのはどれか); transcribed as printed.
• 問題9: q51 option 1: 「発見することか大切た。」 — か and た printed without dakuten (likely ことが大切だ; q51 option 3 prints ことが大切だ correctly); transcribed as printed.
• 問題9: q51 option 3: 「まず型身につけ」 — appears to be missing を after 型 (likely まず型を身につけ); transcribed as printed.
• 問題9: q54 stem ends with 。 while all other stems in this section (49-53, 55-57) are printed without a trailing 。; kept as printed.
• 問題9: q53's この魔物 is underlined in both the passage and the stem, but q56's referent この考え is underlined in neither the passage nor the stem; transcribed as printed without <u>.
• 問題9: Passage (3) paragraph 3 ends 「デジタルだ。（注1）」 with the note reference printed after the sentence-ending period rather than attached to the word デジタル; transcribed as printed.
• 問題9: Passage (3): 「パンコン」 printed twice — paragraph 4 and the （注2） gloss (suspected typo for パソコン; the ン glyph matches the trailing ン, not ソ); transcribed as printed.
• 問題10: Passage paragraph 3 reads 「先生の言葉に私は感銘し」, an odd phrasing (standard would be 感銘を受け); transcribed as printed.
• 問題10: Passage paragraph 1 has odd punctuation 「…読書です」、と仰った。 (a comma after the closing 」); transcribed as printed.
• 問題10: Passage paragraph 4: 自間するようになった — 間 where 問 is expected (自問); verified at high zoom on p11-t (the gate radical encloses 日, not 口); transcribed as printed.
• 問題10: Passage paragraph 5: ほかの入やほかの物 — 入 where 人 is expected; verified at high zoom on p11-t (top-left crossing stroke of 入); transcribed as printed.
• 問題10: Passage paragraph 6: 正しい学間だとされてきた — 学間 where 学問 is expected (q60 option 1 prints 学問 correctly on p12-t); verified at high zoom on p11-t; transcribed as printed.
• 問題10: The 注6 gloss line is printed without its closing parenthesis and without the ） before 至高: prints 「（注 6 至高： 最高」; under the （注N） normalization this becomes 「（注6至高：最高」, the missing ） preserved as printed.
• 問題10: Q58 stem and Q60 stem are printed without a trailing 。 while Q59's stem ends with 。; transcribed as printed.
• 問題10: Q58 option 3 「自身の知識を増やすため」 ends without 。 while the other options end with 。; transcribed as printed.
• 問題10: Q60 option 2: 「個別性や独自性でしか捉えらななる。」 — garbled retype (clearly ら-な-な-る at high zoom on p12-t), suspected for 捉えられなくなる; transcribed as printed.
• 問題10: Q60 option 4: 「ゆかめてしまう」 — suspected typo for ゆがめてしまう; transcribed as printed.
• 問題11: Passage A: '具体的で明快な例を決める人もいるし' — '例を決める' is unusual Japanese and is likely a recall typo (e.g. for 挙げる/示す); transcribed as printed (決める).
• 問題11: Passage A paragraph 2: a 読点 '、' is printed after 'ならない' ('…身につけておかなければならない、中身のある話が…') where a sentence break or different connector would be expected; transcribed as printed.
• 問題11: Passage B: 'キャッチポール' is likely a recall typo for 'キャッチボール'; transcribed as printed.
• 問題11: Passage B final sentence '…間違いなく好印象を与えます' ends with no closing '。'; transcribed as printed (no terminal period).
• 問題11: Q62 option 4: '相手に合わて情報量を調整する' — '合わて' is likely a recall typo for '合わせて'; transcribed as printed.
• 問題11: Passage A, first paragraph: 「具体的で明快な例を決める人もいるし」 — 決める is suspicious (standard phrasing would be 挙げる); transcribed as printed.
• 問題11: Passage A, second paragraph: 「マナーは身につけておかなければならない、中身のある話が…」 — a comma where a 。 would be expected; transcribed as printed.
• 問題11: Passage B, last paragraph: 「キャッチポール」 printed with handakuten (ポ), almost certainly a recall typo for キャッチボール; transcribed as printed.
• 問題11: Passage B ends 「…間違いなく好印象を与えます」 with no closing 。, as printed.
• 問題11: Question 62, option 4: 「Bは相手に合わて情報量を調整する」 — 合わて looks like a typo for 合わせて; transcribed as printed.
• 問題11: Neither passage has a source citation line; presumably omitted by the retypeset.
• 問題11: Question stems 61 and 62 end in か with no 。, as printed.
• 問題12: 問題12 has NO printed instruction sentence in this recall version: the heading line reads only "問題12" and the passage begins immediately, so the instruction field is left empty.
• 問題12: Passage circled reference marks are printed AFTER their underlined span, opposite to the question stems (q64 prints ①<u>ホワイトレグホン</u> with the mark before). Transcribed as printed: passage has <u>ホワイトレグホン</u>① and <u>カイコ</u>（注6）②.
• 問題12: Suspected recall typo in passage para 5: '<u>ホワイトレグホン</u>①のを選びとった' — the stray 'の' before を is ungrammatical (expected 'ホワイトレグホンを選びとった'); transcribed as printed.
• 問題12: Suspected recall typo in passage para 2: 'こくこく希に' (furigana まれ over 希) where standard text would be 'ごくごく稀に' — no dakuten visible on こくこく; transcribed as printed.
• 問題12: Suspected recall typo in passage para 3: 'ところが。' ends with a full-width 。 immediately after ところが where a comma ('ところが、') would be expected; transcribed as printed.
• 問題12: Suspected recall typo in q63 option 4: 'たまたま自然環境に合ったから・。' contains a stray middle dot (・) before the full stop; transcribed as printed.
• 問題12: Suspected recall typo in 注7 gloss: '球状の渡い' (渡い appears to be a typo for 覆い); transcribed as printed.
• 問題12: Recall typo in q66 option 1: the verb is printed with 乍 ('乍りかえた'), a variant/typo for 作 ('作りかえた'); transcribed as printed.
• 問題12: In q63 options 3 and 4 the embedded half-width 'DNA' is followed by a visible space in the source ('DNA が', 'DNA に'); kept as printed. The passage occurrence (para 2) is tight ('DNAが') with no space; kept as printed.
• 問題12: Question-stem punctuation is inconsistent in the source (q63 '残せたか' and q64 '述べているか' have no period; q65 '述べているか。' and q66 '何か。' do); transcribed as printed.
• 問題13: Injected group branding: the last line inside the document box, under 提出・問い合わせ先：人事課平川（内線3210）, is the URL https://www.facebook.com/bui.yuuki.9 — clearly the recall group's own Facebook link injected into the 問題13 stimulus; transcribed as printed.
• 問題13: Suspected source typo in item ①: printed 「人事課ウェブページよりグウンロード」 (verified at high zoom) — presumably ダウンロード in the original.
• 問題13: Suspected source typo in item ④: printed 「講座の内容が書かれた資料パンフレット等）」 with a closing paren but no opening one — presumably 資料（パンフレット等） in the original; transcribed as printed.
• 問題13: Suspected source typo in the boxed note at the bottom of the document: 「補助金は、3月81日に振り込みます。」 — 81 is clearly printed (verified at high zoom); presumably 31.
• 問題13: Inconsistent thousands separators in the source: the q67 stem and the document table print ASCII commas (9,000 / 27,000 / 30,000 / 25,000 / 10,000), but all four q67 options print an ideographic comma (9、000円 / 10、000円 / 27、000円); transcribed as printed.
• 問題13: The two document title lines (2022年度 / 外部講座の受講料補助について) are printed bold (heavier weight than body text, verified against the double-resolution p16-t render) and were marked with <b>.

Corrections applied (49)

Evidence-based reconstructions of the original exam text. This section is the canonical correction log for the sitting.

• 問3 q19 question: 「残った仕事<u>をてきぱきと</u>処理」 → 「残った仕事を<u>てきぱきと</u>処理」
  • The answer choices are adverbial replacements for てきぱきと, so the object particle を must remain with 残った仕事 outside the tested span. Including it drops the particle when the correct 早く正確に is substituted.
  • evidence: structural: 問題3 choices replace the underlined span.
• 問4 q22 answers[1]: 「機の点検のため」 → 「機械の点検のため」
  • Dropped character: 機の点検 is unidiomatic in a factory (工場の稼働) context; the natural term is 機械の点検.
  • evidence: Internal context: the sentence concerns 工場の稼働 (factory operation), where machinery inspection is 機械の点検. (Cited jlpt-freq witness does not actually contain this distractor sentence, but the idiomatic/contextual reconstruction is sound.) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問5 q33 answers[3]: 「てはいられない」 → 「してはいられない」
  • Dropped leading し; the grammatical form is してはいられない.
  • evidence: Confirmed in jlpt-freq witness: 「のんびり(してはいられない)」 appears as the Japanese answer text. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問6 q40 answers[1]: 「コンサートで」 → 「コンテストで」
  • Semantically wrong in a sentence about a ものづくりコンテスト; recall error for コンテスト.
  • evidence: Confirmed in jlpt-freq witness: 「…始めたコンテストで、今年で20回目を迎える」. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q41 passage: 「使われていることはだと思う」 → 「使われていることばだと思う」
  • ことはだ is a non-word; should be ことばだ（言葉だ）.
  • evidence: Internal: 言葉だと思う is the only sensible reading; ことはだ is not Japanese. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q41 passage: 「0威から五成（就字前まで）」 → 「0歳から五歳（就学前まで）」
  • Retype glyph errors: 威→歳, 成→歳, 字→学.
  • evidence: 威/成/字 are non-words here; ages and 就学前 are intended. jlpt-freq Chinese gloss reads 「0岁到5岁(到就学前)」 confirming the meaning. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q41 passage: 「五成児までいる」 → 「五歳児までいる」
  • Retype glyph error 成→歳; should be 五歳児.
  • evidence: Internal: matches the corrected 五歳 age range above; 五成児 is a non-word. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q42 passage: 「「目分は小さい子」」 → 「「自分は小さい子」」
  • 目分 is a non-word; should be 自分.
  • evidence: Internal: the paired clause 「自分は一番大きい子」 in the same passage uses 自分. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q42 passage: 「という環境て過ごす」 → 「という環境で過ごす」
  • Missing dakuten: て→で (環境で).
  • evidence: Internal: the earlier 「環境で生活をする」 in the same passage uses で. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q42 passage: 「違いが出てくると思つのす。」 → 「違いが出てくると思うのです。」
  • 思つのす is garbled; should be 思うのです.
  • evidence: Internal: 思つのす is not Japanese; 思うのです is the obvious intended form. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q43 passage: 「小学校一年生へと進字した」 → 「小学校一年生へと進学した」
  • Retype glyph error 字→学; should be 進学.
  • evidence: Internal: 進字 is a non-word; 進学 is intended. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q43 passage: 「年長児（注1）たったときは」 → 「年長児（注1）だったときは」
  • Missing dakuten: たった→だった.
  • evidence: Internal: たった is ungrammatical after 年長児; だった is intended. (Note: actual text uses half-width （注1）, corrected here.) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q43 passage: 「少さなお友だち」 → 「小さなお友だち」
  • Retype glyph error 少→小; should be 小さな.
  • evidence: Internal: 少さな is a non-word; 小さな is intended. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q43 passage: 「一年生は面側を見てもらう」 → 「一年生は面倒を見てもらう」
  • 面側 is a non-word; the idiom is 面倒を見てもらう.
  • evidence: Internal: 面側を見る is not Japanese; 面倒を見てもらう is the only valid idiom in context. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q45 passage: 「メッセージかシンプルな番組」 → 「メッセージがシンプルな番組」
  • Particle error: か should be が (subject of シンプルだ).
  • evidence: Internal: メッセージがシンプルな番組 is the grammatical reading. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q47 passage: 「関心がないところ発生する」 → 「関心がないところに発生する」
  • Dropped obligatory particle; ところ発生する is ungrammatical, and に発生する is the standard collocation.
  • evidence: Internal: full context 「意識関心がないところ発生することもありません」 is missing a particle; に発生する is standard. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q47 answers[3]: 「探求心持ち続けて」 → 「探求心を持ち続けて」
  • Dropped particle を after 探求心.
  • evidence: Internal: 探求心を持ち続ける requires を. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q48 passage: 「嫌な不愉快なな「思い出」」 → 「嫌な不愉快な「思い出」」
  • Doubled な; should be a single な.
  • evidence: Internal: 不愉快なな is a clear duplication typo. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q48 passage: 「一万の楽しい愉快な」 → 「一方の楽しい愉快な」
  • 一万 is a retype error for 一方 (contrast with the preceding 嫌な不愉快な clause).
  • evidence: Internal: parallel contrast structure 一方の楽しい愉快な「思い出」 vs the unpleasant memories; 一万 is meaningless here. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q48 answers[2]: 「異なっていくこ・とが多い」 → 「異なっていくことが多い」
  • Stray middle dot inside こと.
  • evidence: Internal: こ・と is a clear stray-dot artifact; ことが多い is intended. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q49 passage: 「大さじ二杯のサラグ油」 → 「大さじ二杯のサラダ油」
  • サラグ油 is a non-word; the cooking oil is サラダ油.
  • evidence: Internal: サラダ油 is the only valid term in a recipe context. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q49 answers[2]: 「手順が決めたれた理由」 → 「手順が決められた理由」
  • 決めたれた is a non-word; should be the passive 決められた.
  • evidence: Internal: 決めたれた is not a valid conjugation; 決められた is intended. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q50 answers[2]: 「役に立っとは言えない」 → 「役に立つとは言えない」
  • 立っと is wrong; should be 立つと (dictionary form + と).
  • evidence: Internal: 役に立つとは言えない is the grammatical reading. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q51 question: 「筆者の考こ合う」 → 「筆者の考えに合う」
  • 考こ is garbled; should be 考えに.
  • evidence: Internal: 筆者の考えに合うのはどれか is the standard stem; 考こ合う is not Japanese. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q51 answers[0]: 「発見することか大切た。」 → 「発見することが大切だ。」
  • Missing dakuten: か→が and た→だ (ことが大切だ).
  • evidence: Internal: option 3 in the same question prints 「ことが大切だ」 correctly. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q51 answers[1]: 「探し出すへきだ」 → 「探し出すべきだ」
  • Missing dakuten: へ→べ (すべきだ).
  • evidence: Internal: すべきだ is the intended modal; へきだ is a non-word. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q51 answers[2]: 「まず型身につけ」 → 「まず型を身につけ」
  • Dropped particle を after 型.
  • evidence: Internal: 型を身につける requires を. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q55 passage: 「パンコンやスマートフォン」 → 「パソコンやスマートフォン」
  • パンコン is a non-word; should be パソコン.
  • evidence: Internal: パソコン is the only valid term. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q55 passage: 「スマートフォン：パンコンの機能」 → 「スマートフォン：パソコンの機能」
  • パンコン is a non-word; should be パソコン (注2 gloss).
  • evidence: Internal: パソコン is the only valid term. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q58 passage: 「自間する」 → 「自問する」
  • 自間 is a non-word; should be 自問 (じもん, to ask oneself).
  • evidence: Internal: actual text reads 「読書でいいのだろうかと自間するように」; 自問する is the valid word. (Tightened from the original proposal, whose surrounding span 読むので did not match the file's 読書で.) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q59 passage: 「ほかの入やほかの物」 → 「ほかの人やほかの物」
  • 入 is wrong; should be 人 (ほかの人).
  • evidence: Internal: ほかの入 is not Japanese; ほかの人 is intended. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q60 passage: 「正しい学間だとされて」 → 「正しい学問だとされて」
  • 学間 is a non-word; should be 学問.
  • evidence: Internal: q60 option 1 prints 「学問としての正しい在り方」 correctly. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q60 answers[1]: 「独自性でしか捉えらななる。」 → 「独自性でしか捉えられなくなる。」
  • Garbled retype; should be 捉えられなくなる.
  • evidence: Internal: 捉えらななる is not a valid conjugation; 捉えられなくなる is intended. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q60 answers[3]: 「力を加えて、ゆかめてしまう」 → 「力を加えて、ゆがめてしまう」
  • Missing dakuten: ゆかめて→ゆがめて (ゆがめる, to distort).
  • evidence: Internal: ゆがめてしまう is the intended verb. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問11 q62 passageB: 「キャッチポール」 → 「キャッチボール」
  • キャッチポール does not fit; the idiom is キャッチボール.
  • evidence: Internal: context 「言葉と心のキャッチボールができる人」 requires キャッチボール. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問11 q62 answers[3]: 「相手に合わて情報量」 → 「相手に合わせて情報量」
  • 合わて is a non-word; should be 合わせて.
  • evidence: Internal: 合わせて is the valid conjugation of 合わせる. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q63 passage: 「ホワイトレグホン</u>①のを選びとった」 → 「ホワイトレグホン</u>①を選びとった」
  • Stray の before を; should be ①を選びとった.
  • evidence: Internal: ①のを選びとった is ungrammatical; ①を選びとった is intended. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q63 answers[3]: 「自然環境に合ったから・。」 → 「自然環境に合ったから。」
  • Stray middle dot before the period.
  • evidence: Internal: から・。 is a clear stray-dot artifact. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q63 passage: 「球状の渡い」 → 「球状の覆い」
  • 渡い is a non-word in this gloss; should be 覆い (おおい, covering).
  • evidence: Internal: 「自身を包むための球状の覆い」; 渡い is not Japanese. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q66 answers[0]: 「人間が乍りかえた生物」 → 「人間が作りかえた生物」
  • 乍 is a retype error for 作; should be 作りかえた.
  • evidence: Internal: 乍 is not used here; 作りかえた is intended. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問13 q67 passage: 「人事課ウェブページよりグウンロード」 → 「人事課ウェブページよりダウンロード」
  • グウンロード is a non-word; should be ダウンロード.
  • evidence: Internal: ダウンロード is the only valid term. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問13 q67 passage: 「資料パンフレット等）」 → 「資料（パンフレット等）」
  • Closing paren with no opening paren; matches list-item formatting.
  • evidence: Internal: sibling list item ① uses full parens 「（人事課ウェブページより…）」. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問13 q68 passage: 「補助金は、3月81日に振り込みます。」 → 「補助金は、3月31日に振り込みます。」
  • 81 (81) is an impossible date; should be 31, matching the 3月31日 deadline.
  • evidence: Confirmed in jlpt-freq witness: 「3月31日前提交2~4到人事部」 references the same 3月31日 deadline. (Note: actual text uses half-width 3月81日, corrected here.) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q58 passage: 「（注6至高：最高」 → 「（注6）至高：最高」
  • The closing full-width parenthesis ） after the note number was dropped; every other gloss in this passage uses the pattern （注N）word：…, so this should read （注6）至高：最高.
  • evidence: Category dropped-char, confidence high. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問12 q63 passage: 「こくこく<ruby>希<rt>まれ</rt></ruby>に」 → 「ごくごく<ruby>稀<rt>まれ</rt></ruby>に」
  • The set phrase is ごくごく稀に ('very rarely'). こくこく lacks the dakuten of ごくごく (a non-word as printed), and 希 is the wrong homophone kanji for the reading まれ, whose standard kanji is 稀; the furigana まれ is unaffected.
  • evidence: Category homophone, confidence medium. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問8 q0 passage: 「（注1）されて縮」 → 「されて（注1）縮」
  • （注N）note marker repositioned to follow its glossed term (the retype placed the marker before the word).
  • evidence: Marker-only move; the marker-stripped text is byte-identical. [note-marker relocation pass, 2026-06-20]
• 問12 q0 passage: 「（注2）にお」 → 「に（注2）お」
  • （注N）note marker repositioned to follow its glossed term (the retype placed the marker before the word).
  • evidence: Marker-only move; the marker-stripped text is byte-identical. [note-marker relocation pass, 2026-06-20]
• 問12 q0 passage: 「（注4）する。」 → 「する（注4）。」
  • （注N）note marker repositioned to follow its glossed term (the retype placed the marker before the word).
  • evidence: Marker-only move; the marker-stripped text is byte-identical. [note-marker relocation pass, 2026-06-20]
• 問9 q55 passage: 「デジタルだ。（注1）」 → 「デジタル（注1）だ。」
  • （注N）note marker repositioned to follow its glossed term (the retype placed the marker before the word).
  • evidence: Marker-only move; the marker-stripped text is byte-identical. [note-marker relocation pass, 2026-06-20]