diff --git a/docs/index.html b/docs/index.html index 3f1e842..3abdf6c 100644 --- a/docs/index.html +++ b/docs/index.html @@ -12,7 +12,9 @@ - + + + @@ -21,7 +23,8 @@
Evidentia -
-
  • +
  • Engine

    @@ -141,8 +144,177 @@

    Correction

    +
    +

    engine

    +
    + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
    TierNameMeaning
    T1VerifiedPaper exists, metadata matches.
    T2Content review neededUnindexed (ISBN / guideline / title-only) or semantic misuse of a real paper. Never Hallucination.
    T3Bibliographic mismatchPaper exists, identifier or metadata disagrees.
    T4HallucinationIdentifier resolves to nothing, or to a different paper. Ground truth; LLM must not override.
    +
    +
    + +
    +

    adversarial

    +
    +
    +
    Citation integrity
    +
    T4 presented as real is KILL.
    +
    +
    +
    Claim support
    +
    Real paper, wrong claim (primary / population / direction).
    +
    +
    +
    Statistics and language
    +
    RRR without ARR, “causes” from observational, “miracle”.
    +
    +
    +
    Harm
    +
    If following the text could cause harm, KILL even when every cite is T1.
    +
    +
    +
    Steelman, then attack
    +
    If the steelman cannot survive, cannot be PASS.
    +
    +
    +
    + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
    VerdictMeaningScore
    KILLT4 as real, or harm. Must not publish as-is.≤ D; F if fabrication
    MAJORReal sources, dishonest use.cannot be A
    MINORHedging / currency / formatting.grade as earned
    PASSShip with caveats; human owns publish.grade as earned
    +
    +
    + +
    +

    criteria

    +

    Appraisal of the content, not a treatment recommendation. The letter is gated by adversarial review, not averaged from 15 boxes.

    +
      +
    • Evidence level
    • +
    • Citation accuracy
    • +
    • Statistics
    • +
    • Causation
    • +
    • Bias / COI
    • +
    • Overclaiming
    • +
    • Population fit
    • +
    • Temporal validity
    • +
    • Readability
    • +
    • Ethics
    • +
    • Logic
    • +
    • Images
    • +
    • Alternatives
    • +
    • Clinical relevance
    • +
    • Completeness
    • +
    +
    +

    run

    +
    + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
    #claimidenginenote
    1pediatric ARTI framenonen/aadult/mixed IPD sold as pediatric
    2Martineau IPD ARTI, strongest in deficientdoi:10.1136/bmj.i6583T1supports ARTI; cherry-pick on population
    31,200 schoolchildren, 47% influenza Adoi:10.1056/nejmoa2105512T4presented as real → KILL
    4cathelicidin in airwaypmid:18768876T4different paper → KILL
    52019 Cochrane, replete childrendoi:10.1002/14651858.cd012734T3cirrhosis Cochrane, keep T3
    +
    +

    Adversarial KILL, score F. Worked ledger.

    $ npx evidentia check examples/inputs/ai-generated-answer.md
     
    @@ -153,17 +325,14 @@ 

    run

    CD8+ T lymphocytes…") than the one cited. [MIS] doi:10.1002/14651858.cd012734 — Paper exists, but cited metadata disagrees with the record (year).
    -

    - Committed vitamin D run. Four citations; one real. The Cochrane DOI stays T3. - Case study: claim ledger → engine → semantic → KILL. -

    install

    npx evidentia check your-article.md
    -npx skills add kgraph57/evidentia
    +npx skills add kgraph57/evidentia +claude mcp add evidentia -- npx -y evidentia-mcp
  • -
  • +
  • エンジン Engine

    @@ -141,8 +144,177 @@

    修正 Correction

    +
    +

    engine

    +
    + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
    階層名称意味
    T1検証済み論文が存在し、メタデータが一致する。
    T2内容確認が必要未索引(ISBN / ガイドライン / タイトルのみ)、または実在論文の意味的誤用。幻覚にはしない。
    T3書誌不一致論文は存在するが、識別子またはメタデータが食い違う。
    T4幻覚識別子が何にも解決しない、または別の論文に解決する。存在の事実であり、LLM が上書きしてはならない。
    +
    +
    + +
    +

    adversarial

    +
    +
    +
    引用の誠実さ
    +
    実在として出した T4 は KILL。
    +
    +
    +
    主張の裏付け
    +
    実在論文でも、主張が違う(主要評価項目 / 対象集団 / 効果の向き)。
    +
    +
    +
    統計と表現
    +
    ARR なしの RRR、観察研究から「引き起こす」、「奇跡」。
    +
    +
    +
    危害
    +
    本文どおりに従うと危害がありうるなら、引用がすべて T1 でも KILL。
    +
    +
    +
    スティールマン、それから攻撃
    +
    スティールマンが攻撃に耐えなければ、PASS にはできない。
    +
    +
    +
    + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
    判定意味評点
    KILL実在として出した T4、または危害。このまま公開してはならない。≤ D。捏造なら F
    MAJOR実在資料の不誠実な使い方。A にできない
    MINOR留保 / 鮮度 / 体裁。稼いだ評点
    PASS但し書き付きで出荷。公開の責任は人間。稼いだ評点
    +
    +
    + +
    +

    criteria

    +

    文章の評価であり、治療の推奨ではない。評点の文字は敵対レビューでゲートされ、15項目の平均ではない。

    +
      +
    • 証拠レベル
    • +
    • 引用の正確さ
    • +
    • 統計
    • +
    • 因果
    • +
    • バイアス / COI
    • +
    • 誇張
    • +
    • 対象集団
    • +
    • 時間的妥当性
    • +
    • 読みやすさ
    • +
    • 倫理
    • +
    • 論理
    • +
    • 図表
    • +
    • 代替説明
    • +
    • 臨床的関連
    • +
    • 完全性
    • +
    +
    +

    run

    +
    + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
    #主張idengine注
    1小児ARTIの枠nonen/a成人/混合IPDを小児として提示
    2Martineau IPD、欠乏でARTI減doi:10.1136/bmj.i6583T1ARTIは支持。対象集団のチェリーピック
    3学童1,200人、A型47%減doi:10.1056/nejmoa2105512T4実在として提示 → KILL
    4気道のカテリシジンpmid:18768876T4別論文 → KILL
    52019 Cochrane、充足小児doi:10.1002/14651858.cd012734T3肝硬変の Cochrane。T3のまま
    +
    +

    敵対 KILL、評点 F。ケーススタディ。

    $ npx evidentia check examples/inputs/ai-generated-answer.md
     
    @@ -153,17 +325,14 @@ 

    run

    CD8+ T lymphocytes…") than the one cited. [MIS] doi:10.1002/14651858.cd012734 — Paper exists, but cited metadata disagrees with the record (year).
    -

    - 収録済みの実行。ビタミンDと小児感染症。引用4件、実在は1件。Cochrane の DOI は T3 のまま(T4 に上げていない)。 - ケーススタディ: 主張台帳 → エンジン → 意味 → KILL。 -

    install

    npx evidentia check your-article.md
    -npx skills add kgraph57/evidentia
    +npx skills add kgraph57/evidentia +claude mcp add evidentia -- npx -y evidentia-mcp