N1 2022-12 — audit report

66 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: Clean third-party digital retype (recall version compiled by 'Yuuki Bui' of the Vietnamese 'Tôi yêu Ngoại ngữ' group; printed to PDF via Foxit PhantomPDF, March 2023). Crisp uniform type, fully legible throughout; the light diagonal watermark on every page never impairs reading. Loses points for fidelity: it is a recall reconstruction rather than the official booklet, with no furigana, no covers/instructions, omitted 問題6 example, garbled instruction lines, scattered typos, and a listening section reduced to printed options plus an empty answer grid.

Every section was transcribed twice independently (a full-page pass and a double-resolution half-page pass) and reconciled. Of 13 sections, 2 agreed on all content between the two passes (reconciled automatically, differing only in transcriber notes) and 11 had at least one content difference resolved by a third adjudication pass that re-read the page images. 117 leaf-level differences were examined in total.

Residual OCR risks (33):

• 問題1: Q6 underline extent: 如実 is clearly underlined; whether the trailing に is inside the underline is slightly ambiguous in the render. Both transcribers chose <u>如実に</u>; kept for consistency with the full reading-target convention.
• 問題1: All other leaves were identical between A and B and matched the images, so correlated-error risk is low but non-zero for an OCR source that is itself a retypeset recall reconstruction.
• 問題2: The 入/人 distinction in the instruction line relies on glyph shape in a low-contrast retypeset font under a diagonal watermark; read as 人 with high confidence but not absolute certainty since no higher-zoom crop tool (ImageMagick/PIL) was available in this environment.
• 問題2: All seven question stems and answer arrays were identical between A and B with no structural disagreements, so correlated errors beyond the spot-checked items cannot be fully excluded, though the spot-checks found none.
• 問題3: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 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 instruction-line 人/入 distinction relies on the absence of a top stroke at this render resolution; it is consistent with the official instruction normally reading 入れる, so the 人れる is a genuine source typo rather than an OCR artifact, but a higher-DPI source could in principle reveal a faint stroke.
• 問題5: Q34 北市 and kana いったん are retained as printed; they are plausibly recall-version artifacts (likely intended 〇〇町 and 行ったん) but cannot be corrected without a cleaner source.
• 問題6: The instruction's underlined-★ rendering as <u>★</u> is a judgment call: the print is a single underlined blank carrying the ★, distinct from the ＿＿＿ low-line instruction-blank convention. Chose the underline convention; a strict reading could differ.
• 問題6: The garbled option-list 「い 2・3・4」 is transcribed literally; whether the leading glyph is a recall-corrupted 1 or genuinely い cannot be resolved from the image, but it is rendered as printed (い).
• 問題6: Suspected recall typos (旧白本小学校, 始めて見た, 初めての, 真夏の戻った, 人る) are intentionally left uncorrected per conventions; these are source artifacts, not OCR errors.
• 問題7: The internal gaps in メニュ ー and ネー ミング are encoded as half-width spaces (following B). They could instead be pure font side-bearing of the long-vowel mark ー rather than intentional spaces; if so they should be removed. Gaps are clearly visible at high zoom, so encoding-as-printed was chosen, but this is the least certain call.
• 問題7: OOパスタ is read as two Latin capital O; the source font for letters vs the digit 0 is distinguishable here (the O are notably rounder/wider) but a retype using 〇〇 (kanji zero) cannot be 100% excluded.
• 問題7: バッケージ is read with dakuten (バ); the watermark grazes this character, but the two diagonal strokes are clearly distinct from a handakuten circle at 10x zoom.
• 問題8: Passage (4) first column (科学とは) shows a ~20px larger top offset than body columns; this is well under a full-character indent and is read as the natural top-bearing of 科 / passage-start padding, but it is the only column whose top differs, so a faint paragraph-start indent there cannot be 100% excluded. It does not affect the body being a single paragraph.
• 問題8: 質間 (passage 3) and 知護 (passage 4) render small/vertical; both transcribers independently read them as typos and the renders support that, but at this resolution a 質問/知識 reading is not fully excludable. Transcribed as printed per convention regardless.
• 問題8: Watermark overlap is heavy on every page; no place was found where it changed a glyph, but low-contrast strokes under the diagonal watermark carry minor residual uncertainty.
• 問題9: Q54 answer 4: confirmed 疑開 as a deliberate source typo via zoomed p11-b crop; the glyph's right component is 開 not 問. Very high confidence, but it is an unusual typo, so flagged.
• 問題9: Justification spaces were collapsed throughout (環境適応性, 食物新奇性恐怖, 対峙する, コアラ); if downstream consumers want raw spacing preserved this differs, but it follows the stated convention.
• 問題9: 筆/筈 in passage (3) '質問を促す力になる筈なのだ' is printed as 筈 (both transcribers agree); the watermark crosses this region but the glyph is legible as 筈.
• 問題9: The instruction and a few passage parentheses around digits print full-width and were kept full-width; only Latin letters/digits were forced to ASCII.
• 問題10: Instruction first separator: both transcribers read 1. (half-width period) rather than a nakaguro, and both report a space before から. At full-page resolution the period vs nakaguro distinction is near the limit of legibility; kept the agreed 1. 2・3・4 から per the as-printed rule, but a sharper scan could conceivably show a nakaguro.
• 問題10: No image-cropping/zoom tooling (PIL/ImageMagick) was available in the environment, so judgments rely on the supplied full-page and double-resolution half-page renders without further magnification. The 注-line spacing was legible enough in p13-b to confirm B's raw observation, which was then normalized per convention.
• 問題10: The 注 normalization (（注1） vs （注 1)) follows the documented convention example; if the project later decides to preserve raw retype spacing verbatim, this passage would need to switch to B's form.
• 問題11: Colon spacing in the 注 gloss: the printed gap after the full-width colon could be read as a typed half-width space; resolved to no space for cross-section determinism (consistent with mondai8.B), but A read it as a space — low-impact single-character ambiguity.
• 問題11: Gloss attachment: the gloss is physically printed beneath passage B but glosses a marker in passage A; attached to A on the reference-marker principle. A consumer that strictly follows 'transcribe where printed' might prefer B.
• 問題11: 前伺き is recorded as printed (recall typo); not corrected, so downstream search for 前向き will miss it.
• 問題12: The watermark diagonals overlap parts of the passage on p16-t/p16-b; the long passage was only fully re-read where A and B disagreed plus the spot-checked stems, so a correlated single-character misread shared by both transcribers in an unspotted passage region cannot be fully excluded.
• 問題12: Instruction-line normalization 1.2・3・4 follows the pilot convention for dropping half-width spaces around the garbled run; if the project later decides to preserve the literal spacing, this line would need revisiting.
• 問題13: Email line: the wide phone↔email gap is a tab-like whitespace of indeterminate width; rendered as a single half-width space per the project's autospacing norm. Reasonable transcribers could render it as a full-width or double space.
• 問題13: Eメール internal katakana spacing: the メ ー ル run is evenly kerned; I treated it as a single token with no internal space (matching ダウンロード handling), but B read a space after メ. Low-stakes either way.
• 問題13: The two leader-dot styles (・・・ vs ...) before the 指導教員 and 申し込み lines were both normalized to a single … per the 3-dot-group rule; the underlying glyphs differ in print and the normalization is a convention choice.
• 問題13: Whether the nested inner box around the A/B/C list should be represented at all is a structural judgment; minimal HTML cannot express the nesting, so it is flattened to <p> lines.

Source-text quirks the OCR transcribed verbatim (printed typos/recall artifacts; the ones judged to be retype errors were corrected in the layer below) — 78 noted at OCR time:

• 問題1: The instruction line prints a half-width space between 1・2・3・4 and から ("1・2・3・4 から"); transcribed as printed.
• 問題1: Page carries a retypeset title header "2022年12月N1 YUUKI BUI" above 言語知識（文字・語彙） plus diagonal "Tôi Yêu Ngoại Ngữ Group / Yuuki Bùi" watermarks; all ignored per conventions.
• 問題1: In question 3 the underline covers only 透けて; the following いる is outside the underline as printed.
• 問題2: The printed 問題2 instruction line is garbled: it reads 「（　）に人れるのに」 with 人れる instead of 入れる (verified at high zoom on both p01-t and p01-b). Transcribed as printed per convention.
• 問題2: Question 10's stem itself uses the correct 「手に入れた」, so the 人れる typo is confined to the instruction line.
• 問題2: The retypeset font renders the digit 1 as a glyph resembling Latin capital I (e.g. 問題 I, I・2・3・4, 20 分のI); normalized to ASCII 1 per convention.
• 問題3: Instruction line is garbled in the source: it reads 「…ものを、い 2・3・4 から一つ選びなさい。」 — the digit 1 appears to have been dropped/corrupted into 「い」 (standard wording would be 1・2・3・4). Transcribed as printed.
• 問題3: Question 15: option 1 is さわやかな but options 2-4 lack the final な (おだやか, こまやか, かろやか) — likely a recall/retype omission; transcribed as printed.
• 問題3: Instruction line is garbled in the source: it reads 「…ものを、い 2・3・4 から一つ選びなさい。」 — the option list is missing 「1・」 and shows hiragana 「い」 in its place (presumably a corruption of 「1・2・3・4」 in the retype). Transcribed as printed, including the half-width spaces around 「2・3・4」.
• 問題3: Question 15: the printed underline covers only 温和 (not the following な), while 性格 is printed in bold without an underline — apparently misplaced emphasis in the retype (the intended target is presumably 温和な, given the options). Underline transcribed exactly as printed; the stray bold on 性格 is not marked up.
• 問題3: Question 15: option 1 is さわやかな (with な) but options 2-4 are おだやか／こまやか／かろやか (without な) — as printed.
• 問題4: 21-1 prints 「新聞を名家庭に譲る仕事」 — 「名家庭」 is a suspected retype typo for 「各家庭」; transcribed as printed.
• 問題4: 22-1 prints 「毎日体を運動しているので」 — odd phrasing (体を運動する), likely a recall/retype artifact; transcribed as printed.
• 問題4: 22-3 prints 「動物の言菫が分かる」 — the character is clearly 菫 (verified at 24x zoom), a typo for 葉 in 言葉; transcribed as printed.
• 問題4: 24-4 prints 「口川さん」 — unusual surname, suspected typo (perhaps for 山川/小川); transcribed as printed.
• 問題4: 21 option 1: printed as 「新聞を名家庭に譲る」 — 名家庭 is almost certainly a retype error for 各家庭; transcribed as printed (verified at high zoom).
• 問題4: 22 option 1: printed as 「毎日体を運動しているので」 — odd phrasing, possibly a recall/retype artifact; transcribed as printed.
• 問題4: 22 option 3: printed glyph is 言菫 (草かんむり over 堇), almost certainly a retype error for 言葉; transcribed as printed (verified at high zoom).
• 問題5: The 問題5 instruction line is printed with the garbled 人れる instead of 入れる (verified at high zoom on the double-resolution render: the glyph is 人, with no top horizontal stroke); transcribed exactly as printed. The question bodies themselves (e.g. 29 入学/入りたい/入った, 33 入れた…) use 入 correctly.
• 問題5: Question 34: the place name is printed as 北市 (unusual for a JLPT item, where invented town names like 北町 are typical); possibly a recall/retype artifact, but transcribed exactly as printed. The verb いったん is written in kana (likely 行ったん); transcribed as printed.
• 問題5: Question 35: options 3 and 4 print 乗って欲しがった with 欲 in kanji, contrasting with kana-only 乗りたがった in options 1 and 2; transcribed as printed.
• 問題6: Instruction line is garbled in the source. It prints (minus the ignored 問題6 prefix) as 「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。」 — 人る is printed where 入る should be; the ＿＿＿ slot is absent and the blank (an underlined ★) sits after 「のに」 instead of between 「の」 and 「に」 (expected 「次の文の＿★＿に入る」); and the option list reads 「い 2・3・4」 where 「1・2・3・4」 is expected (the digit 1 appears dropped/garbled to い). Transcribed as printed. The ★ sits on a single short underline (one underlined blank, not full-width ＿ low-line characters), rendered as <u>★</u> per the underline convention; the inter-character spaces around the ★ and the option list are typesetting kerning and were not transcribed as literal spaces.
• 問題6: Q36: 「秋の初めてのこの時期」 and 「真夏の戻った」 are suspected recall typos (likely 「秋の初めの」 and 「真夏に戻った」); transcribed as printed. The ★ is the second of the four slots for this question only, and its four answer options are laid out on a single line, unlike the two-column layout of the other questions.
• 問題6: Q37: school name prints as 「旧白本小学校」 (白本 is an unusual name, possibly a recall typo for 日本 or another school name), and 「始めて見た」 prints where 「初めて見た」 is expected; transcribed as printed.
• 問題7: Instruction line is garbled in the source: it prints 「の中に人る」 with 人る (U+4EBA) where 入る (U+5165) is expected (verified at high zoom); transcribed as printed per task guidance. The 41/44 in the instruction are plain digits, not boxed.
• 問題7: Instruction wording 「文章全体の内容を考えて」 differs from the official phrasing 「文章全体の趣旨を踏まえて」; transcribed as printed.
• 問題7: Placeholder marks: 「OOパスタ」 uses two Latin capital O (round, clearly wider than the narrow digit 0 used elsewhere and not 〇); 「XX県産」 Latin X; 「△△」 full-width triangles. Kept as printed.
• 問題7: Suspected retype typo in paragraph 3: 「商品のバッケージ」 prints バ with dakuten (two diagonal strokes, verified at high zoom) where パ (handakuten) is expected; transcribed as printed.
• 問題7: A full-width period 。 follows the boxed 【42】 (「するのも【42】。消費社会とは」); included as printed.
• 問題7: メニューis printed with an internal gap both times (「メニュ ー」), and ネーミング likewise (「ネー ミング」) in the passage and in the （注2）gloss; encoded with a half-width space as printed (see residual risk in audit re: possible font side-bearing).
• 問題7: In-passage cloze blanks 41-44 are boxed and rendered 【41】〜【44】; option-block numbers below the passage print as 「41)」〜「44)」. Q44 options laid out two per row (1-2, 3-4), order verified.
• 問題8: Source is the Tôi Yêu Ngoại Ngữ / Yuuki Bùi recall retype; diagonal watermarks (Tôi Yêu Ngoại Ngữ Group / Yuuki Bùi) and the page header 2022年12月N1 YuukiBui overlay the text on every page and were ignored. No group branding was injected into the question content of this section.
• 問題8: Instruction line is garbled as printed: 最もよいものを is printed with a stray space as 最もよ いものを; transcribed as printed. The (1)/(4) in the instruction are printed with narrow (half-width style) parentheses and were kept half-width.
• 問題8: Passage (1), second paragraph: パワー is printed with a stray space as パワ ー; transcribed as printed.
• 問題8: Passage (1), first paragraph ends 大事な機能でもあります、何より根絶できません。 — the 、 after あります reads oddly (a 。, あり、 or ありますし、 might be expected) and is suspected recall typo, but transcribed as printed.
• 問題8: Question 45 option 3: 感清 is printed (suspected recall typo for 感情); transcribed as printed.
• 問題8: Passage (3): そうした質間を is printed with 間 (suspected recall typo for 質問); transcribed as printed. The same passage prints 質問 correctly twice earlier.
• 問題8: Passage (4): 知護 is printed (suspected recall typo for 知識); transcribed as printed; the later occurrence その知識 is printed correctly.
• 問題8: Question 48 stem is printed without a trailing 。 (筆者の考えに合うのはどれか); transcribed as printed.
• 問題9: Stems printed WITHOUT a trailing 。: Q49 ("とんでもないことになりますとあるが、どうなるのか"), Q50 ("筆者が言いたいことは何か"), Q54 ("どのようにすれば質問の量が増えるか"). Q51-53, 55-56 stems DO end with 。. Transcribed as printed.
• 問題9: Q49 reference phrase 'とんでもないことになります' is NOT underlined in this retypeset (JLPT originals normally underline it); transcribed without <u>.
• 問題9: Q54 answer 4 is printed with the typo 疑開 (門-radical glyph), not 疑問: '自分の既存の知識に、常に疑開を持つ。' — verified on the double-resolution p11-b crop; transcribed as printed.
• 問題9: Passage (2): 原型 is printed wrapped in straight ASCII double quotes ("原型"), not 「」 or 『』; transcribed as printed.
• 問題9: Transcriber B introduced numerous garbled readings in passages (1)-(4) (e.g. 'もったまま', '温存をシフト', '相手盾', 'Bは社資本金', dropped negation '汚染であるという'); the full-page renders and double-resolution halves confirm the coherent A readings, which were adopted. The source text itself is a clean retype here, so these were OCR errors by B, not source typos — except Q54's 疑開, which IS a genuine source typo (B captured it, A over-corrected to 疑問).
• 問題10: Source is the Tôi Yêu Ngoại Ngữ / Yuuki Bùi recall retype; the page header 2022年12月N1 YuukiBui, page numbers, and diagonal watermarks were ignored.
• 問題10: Instruction line is garbled as printed: the option list reads 1. 2・3・4 — a half-width period and space after 1 instead of the first nakaguro, and a visible space before から. Both transcribers read 1. (period) at high zoom; transcribed as printed.
• 問題10: Lead-in line prints ボランティア活軌をしている — 活軌 is a likely recall typo for 活動; verified at high zoom (the glyph is 軌, radical 車 + 九, not 重 + 力); transcribed as printed.
• 問題10: Stray half-width spaces appear mid-word in the passage and were kept as printed: 「人間のためにならない」自 然は (paragraph 5); 人間 のためになろうとも and 生存権をも っており (paragraph 6); 立場にたつこ ともできない (paragraph 9). Likely retype artifacts, not justification.
• 問題10: Paragraph 9 first occurrence prints 立場にたっことも with a small っ (likely typo for たつことも); the second occurrence in the same sentence prints regular つ (たつこ とも). Both kept as printed.
• 問題10: Paragraph 10 prints この間いに対して with 間 (likely recall typo for 問い); the instruction line and paragraph 4 (問われれば) use the correct 問. Verified at high zoom; kept as printed.
• 問題10: Underline span mismatch between passage and question, both kept as printed: the passage underlines 「人間のための」論 (closing bracket before 論, は not underlined), while question 57 underlines 「人間のための論」 (closing bracket after 論).
• 問題10: Question 59 option 2 prints 見付け出す (unusual okurigana for 見つけ出す); transcribed as printed.
• 問題10: No source citation line (（…による）) is printed after the passage; the recall version omits it. The passage ends with the two 注 gloss lines.
• 問題11: The retype puts half-width spaces immediately around every half-width Latin letter (e.g. 「次の A と B の」, 「A も B も」); these letter-adjacent spaces are treated as automatic inter-script spacing and dropped (matching the pilot N1_2023-12 normalization), but stray spaces inside Japanese text runs are kept as printed and itemized below.
• 問題11: Instruction line has stray spaces: 「最もよい ものを」 and 「から 一つ選びなさい」 — kept as printed.
• 問題11: Passage A, second paragraph: stray space in 「誰にでも 調子の良い時と」 — kept as printed.
• 問題11: Passage B, second paragraph: 「前伺きに物事を捉える」 — 伺 is almost certainly a retype/recall typo for 向 (前向きに); the person radical is clearly visible at high zoom, identical in both page-half renders; transcribed as printed.
• 問題11: Passage B, third paragraph: stray space in 「出てきませ ん。」 — kept as printed.
• 問題11: Question 61, option 3: stray space in 「やる気や 自信を」; option 4: stray spaces in 「Bは 現状を認めて」 (extra gap between は and 現状, beyond letter-adjacent spacing) and 「今後どうす るべきか」 — all kept as printed.
• 問題12: Instruction line is garbled as printed: the options read 「1. 2・3・4」 — the separator between 1 and 2 is a baseline period instead of a 中黒 (middle dot). Transcribed as 1.2・3・4 (half-width spaces around the half-width run dropped as typesetting artifacts, matching the pilot normalization).
• 問題12: Passage paragraph beginning それでも、青年期に向けて: printed 挫折しり — almost certainly a recall/retype typo for 挫折したり (question 62 option 2 prints 挫折したり correctly); transcribed as printed.
• 問題12: Same paragraph: printed 個「生的であることから — a stray opening corner bracket 「 inside what should be 個性的; transcribed as printed.
• 問題12: Same paragraph: printed 自分にとって個性とはと、考えるのはよいことだ — reads as if quotation brackets around 自分にとって個性とは were dropped in the retype; transcribed as printed. Only the paragraph's first よいことだ is underlined (the question-62 reference); this second occurrence is not.
• 問題12: Paragraph beginning もちろん、人間はまわりの社会を: the page prints 無理につっぱって with handakuten (ぱ), consistent with つっぱったり/つっぱり later; verified against the double-resolution p16-b render (transcriber A's つっばって with dakuten was an OCR misread).
• 問題13: Source is the Tôi Yêu Ngoại Ngữ / Yuuki Bùi recall retype; the page header 「2022年 12月 N1 YuukiBui」 and diagonal watermarks (Tôi Yêu Ngoại Ngữ Group / Yuuki Bùi) were ignored per conventions.
• 問題13: Study-group branding is injected INTO the document content: the 問い合わせ先 line ends 「…/research. html yuukibui」 — the trailing 「yuukibui」 is the group's branding appended to the university URL; transcribed exactly as printed.
• 問題13: The contact URL is printed with literal ASCII-to-ASCII spaces: 「https : // www.hoshikawa-u.ac. jp/research. html」 (spaces around the colon, after //, after 「ac.」, and after 「research.」). These are real keystrokes, not CJK/Latin autospacing, and were kept as printed; the leading 研究推進課↔https boundary space was dropped as autospacing.
• 問題13: The Eメール address is printed as graduate-research@hoshikawa.ac.jp (no 「-u」), inconsistent with the website domain hoshikawa-u.ac.jp on the line above; possibly a recall typo — transcribed as printed.
• 問題13: The email line uses a half-width ASCII colon before the address (Eメール:graduate-research…), unlike the full-width ： used for 問い合わせ先 and 電話; verified at high zoom and kept as printed.
• 問題13: Half-width spaces immediately at CJK↔Latin/digit boundaries (e.g. 「4 月 26 日」, 「研究推進課 https」) were dropped as retype autospacing artifacts, matching the normalization used for the other N1_2022-12 sections; stray spaces between CJK/katakana characters were kept as printed and itemized below.
• 問題13: Document item 2 ② prints 「渡航期間は間わない」 — 間わない is a typo for 問わない; verified at high zoom (the character is 間, not 問); transcribed as printed.
• 問題13: The C-item note line prints 「幕集期間後」 — 幕集 is a typo for 募集 (elsewhere the document correctly prints 募集期間); verified at high zoom (bottom radical is 巾, not 力); transcribed as printed.
• 問題13: Instruction line has stray spaces as printed: 「下の問いに対 する答えとして」 and 「から 一つ選びなさい」.
• 問題13: Question 65 stem has stray spaces as printed: 「思っ ている」, a space after the opening bracket in 「「 グロ ーバル」, and a space before the ー in 「グロ ー」 (the printed line wraps between グロ ー and バル).
• 問題13: Question 66 option 3 prints 「取りに行かなけれはならない」 — は without dakuten, a retype typo for ば; transcribed as printed.
• 問題13: First ● bullet prints 「ダウンロ ー ド」 with stray spaces around the long vowel mark; second ● bullet prints 「②で発表 が確定していない」 with a stray space between 発表 and が — kept as printed.
• 問題13: The A/B/C list lines of 応募手続き are printed with a half-width space after each label letter (「A 申請書」, 「B 推薦書」, 「C 発表…」); verified at high zoom and kept. These lines sit in a nested inner box within the page-wide document box; the nesting is not representable in minimal HTML, so they are plain <p>s and the two ● bullets that follow are outside that inner box.
• 問題13: Table cells in question 65 wrap mid-phrase across two printed lines (e.g. 海外のシンポジウムで／講演を聞く, 2022年8月24日／から3日間); joined without <br> since the wraps are layout-only. The 渡航期間 dates and the 採択歴 column (なし／なし／2020年度にあり／2021年度にあり) were verified at high zoom.

Corrections applied (28)

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

• 問3 q17 question: 「<u>拮抗</u>している」 → 「<u>拮抗している</u>」
  • The choices are complete predicates, including the correct 差がない. Leaving している outside the tested span produces 差がないしている; the full target is 拮抗している.
  • evidence: structural: 問題3 choices replace the underlined span.
• 問1 q6 question: 「<u>如実に</u>」 → 「<u>如実</u>に」
  • Reading target is 如実 (じょじつ); the printed underline ends at 実 and does not extend under the particle に
  • evidence: underline-boundary-vs-image
• 問3 q15 answers[1]: 「おだやか」 → 「おだやかな」
  • 問題3 synonym option must be parallel to the な-adjective target 温和な and to option 1 さわやかな; clean jlpt-freq witness prints all four options with trailing な
  • evidence: jlpt-freq witness line 63: 1.さわやかな 2.おだ定かな 3.こまやかな 4.かろうやかな — all four end in な (OCR garbles the stems but the な is present on each) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問3 q15 answers[2]: 「こまやか」 → 「こまやかな」
  • 問題3 synonym option must be parallel to the な-adjective target 温和な; clean jlpt-freq witness prints こまやかな with trailing な
  • evidence: jlpt-freq witness line 63: 3.こまやかな [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問3 q15 answers[3]: 「かろやか」 → 「かろやかな」
  • 問題3 synonym option must be parallel to the な-adjective target 温和な; clean jlpt-freq witness prints かろ(う)やかな with trailing な
  • evidence: jlpt-freq witness line 63: 4.かろうやかな (trailing な present) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問4 q21 answers[0]: 「新聞を名家庭に」 → 「新聞を各家庭に」
  • 名家庭 is a non-word retype error for 各家庭 (deliver newspapers to each household)
  • evidence: clean jlpt-freq witness: 新聞を各家庭に譲る仕事をしている [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問4 q22 answers[0]: 「毎日体を運動しているので」 → 「毎日体を鍛えているので」
  • 体を運動する is ungrammatical; the original verb is 鍛える, a recall/retype artifact
  • evidence: clean jlpt-freq witness: 毎日体を鍛えているので、腕の底力なら誰にも負けない [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問4 q22 answers[2]: 「動物の言菫が分かる」 → 「動物の言葉が分かる」
  • 言菫 is a non-word retype error for 言葉
  • evidence: clean jlpt-freq witness: 動物の言葉が分かる特別な底力 [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問4 q24 answers[3]: 「口川さんは」 → 「今川さんは」
  • 口川 is an implausible surname; the source typo is for 今川
  • evidence: clean jlpt-freq witness: 今川さんは、ほかの人より手痛いスピードで作業した [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問6 q36 question: 「秋の初めてのこの時期は」 → 「秋の初めのこの時期は」
  • 初めての is a recall/retype artifact; the original reads 秋の初めの (early autumn)
  • evidence: clean jlpt-freq witness: 秋の初めのこの時期は [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問6 q36 question: 「真夏の戻った」 → 「真夏に戻った」
  • 真夏の戻った is ungrammatical; the particle should be に (returned to midsummer)
  • evidence: clean jlpt-freq witness: 真夏に戻った [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問6 q37 question: 「旧白本小学校」 → 「旧白木小学校」
  • 白本 is an unusual school name; the original is 白木 (しらき), a one-stroke 本/木 retype error
  • evidence: clean jlpt-freq witness: 旧白木小学校の校舎 [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問6 q37 question: 「校舎を始めて見たが」 → 「校舎を初めて見たが」
  • 始めて is a retype error for 初めて (saw for the first time)
  • evidence: clean jlpt-freq witness: 校舎を初めて見たが [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q42 passage: 「商品のバッケージを」 → 「商品のパッケージを」
  • バッケージ is a non-word; the original is パッケージ (package) with handakuten
  • evidence: clear non-word; only パッケージ is a real word (both retype witnesses share the dakuten typo, but no alternative reading exists) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q45 answers[2]: 「その感清を」 → 「その感情を」
  • 感清 is a non-word retype error for 感情
  • evidence: clean jlpt-freq witness: その感情を抑えることができる [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q47 passage: 「そうした質間を」 → 「そうした質問を」
  • 質間 is a non-word retype error for 質問; the same passage prints 質問 correctly twice earlier
  • evidence: clean jlpt-freq witness: そうした質問を真正面から投げかけられたとき [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q48 passage: 「得た知護を」 → 「得た知識を」
  • 知護 is a non-word retype error for 知識; the same passage prints その知識 correctly later
  • evidence: clear non-word; 知識 used correctly later in same passage (注 references 知る) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q54 answers[3]: 「常に疑開を持つ」 → 「常に疑問を持つ」
  • 疑開 is a non-word retype error (門-radical glyph) for 疑問
  • evidence: clean jlpt-freq witness: 自分の既存の知識に、常に疑問を持つ [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q57 passage: 「ボランティア活軌をしている」 → 「ボランティア活動をしている」
  • 活軌 is a non-word retype error for 活動
  • evidence: clean jlpt-freq witness: 森林を守るボランティア活動をしている [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q57 passage: 「この間いに対して」 → 「この問いに対して」
  • 間い is a non-word retype error for 問い; the passage uses 問 correctly elsewhere
  • evidence: clean jlpt-freq witness: この問いに対して、私たちは [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q57 passage: 「立場にたっことも」 → 「立場にたつことも」
  • たっこと (small っ) is ungrammatical for 立つこと; the second occurrence in the same sentence prints たつ correctly
  • evidence: clean jlpt-freq witness: 立場にたつことも、自然中心的な立場にたつこと [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問11 q60 passageB: 「前伺きに物事を捉える」 → 「前向きに物事を捉える」
  • 前伺き is a non-word retype error for 前向き
  • evidence: clean jlpt-freq witness: 前向きに物事を捉えることができる [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q62 passage: 「個性的であろうとして挫折しり」 → 「個性的であろうとして挫折したり」
  • 挫折しり is a retype truncation of 挫折したり (question 62 option 2 prints 挫折したり correctly)
  • evidence: q62 option 2 in same item prints 挫折したり; clear truncation (both retype witnesses share the typo but only 挫折したり is grammatical) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q62 passage: 「個「生的であることから」 → 「個性的であることから」
  • 個「生 contains a stray corner bracket and is a retype garbling of 個性的
  • evidence: clean jlpt-freq witness: 個性的であることから逃避する [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問13 q65 passage: 「渡航期間は間わない」 → 「渡航期間は問わない」
  • 間わない is a non-word retype error for 問わない
  • evidence: clear non-word; only 問わない is grammatical (both retype witnesses share the typo, no alternative reading) [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問13 q65 passage: 「幕集期間後になる」 → 「募集期間後になる」
  • 幕集 is a retype error for 募集; the document prints 募集期間 correctly elsewhere
  • evidence: clean jlpt-freq witness: 申し込みの開始が募集期間後になる場合は [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問3 q15 question: 「<u>温和</u>な」 → 「<u>温和な</u>」
  • 問題3 underline boundary: synonym options are adnominal, so the trailing な belongs inside the underlined span (same pattern as N1_2020-12 q16 架空の). Without it the answer leaves a dangling な.
  • evidence: structural: 問題3 options replace the underlined adnominal span.
• 問13 q66 answers[2]: 「研究推進課窓口に取りに行かなけれはならない」 → 「研究推進課窓口に取りに行かなければならない」
  • なけれは is an impossible form; the obligation construction is 〜なければならない. The は lost its dakuten (は→ば), a classic dropped-dakuten retype artifact. This is a reading-comprehension distractor (not a 問題4 deliberate-misuse sentence), so it must be grammatical Japanese as printed in the real exam.
  • evidence: Category typo, confidence high. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]