Kuiz berkaitan panduan · Sederhana Tahap

Mengesan Petikan Palsu daripada Kuiz AI

Kuiz diri anda tentang mengapa chatbots mencipta rujukan, empat jenis petikan buruk dan kaedah langkah demi langkah untuk mengesahkan setiap sumber.

Laluan panduan yang berkaitanMengesan Petikan AI Palsu
soalan 1 daripada 8

Mengapakah chatbot boleh menghasilkan petikan yang tidak wujud?

Pratonton setiap soalan8 soalan

Jawapan dan penjelasan muncul semasa anda mengambil kuiz.

  1. Mengapakah chatbot boleh menghasilkan petikan yang tidak wujud?
    • It copies citations from a secret database of fake papers
    • It predicts likely text, and citations follow a predictable format it can imitate
    • It is programmed to test whether students check sources
    • It only makes up citations when asked about recent events
  2. Jenis petikan buruk yang manakah yang paling sukar dikesan oleh panduan?
    • A completely made-up reference
    • A real author paired with an invented title
    • A reference with the wrong year
    • A real, correctly cited paper that doesn't support the claim
  3. What happened in Mata v. Avianca in 2023?
    • Lawyers were sanctioned for citing cases ChatGPT invented
    • A university banned chatbots in research
    • A publisher sued an AI company over citations
    • A court ruled AI citations are always admissible
  4. Apabila menyemak DOI, apakah yang perlu anda sahkan selepas membukanya melalui doi.org?
    • That the page loads quickly
    • That the DOI is shorter than 20 characters
    • That the paper it opens has the same title and authors as the citation
    • That the journal is free to read
  5. Apakah yang perlu anda lakukan jika anda tidak dapat mengesahkan sumber?
    • Keep it but move it to a footnote
    • Remove it
    • Ask the chatbot to confirm it
    • Replace the DOI with a URL
  6. Why is asking the chatbot 'Is this citation real?' not a reliable check?
    • Chatbots can't read their own earlier messages
    • It takes too long
    • Chatbots always say the citation is fake
    • The model can produce a confident yes as easily as it produced the fake reference
  7. What kind of error remains common in retrieval-augmented tools, according to the guide?
    • Misattribution, where a real linked source doesn't fully support the claim
    • DOIs that are always missing
    • Citations only in foreign languages
    • Refusing to give any sources
  8. Which of these is a red flag listed in the guide?
    • The paper has more than three authors
    • The source is older than five years
    • A journal name that is slightly off from the real journal
    • The article has an abstract