---
date: '2023-01-29T11:04:46'
hypothesis-meta:
  created: '2023-01-29T11:04:46.963592+00:00'
  document:
    title:
    - 2301.11305.pdf
  flagged: false
  group: __world__
  hidden: false
  id: vhYdup_EEe228yPQ_hxjIg
  links:
    html: https://hypothes.is/a/vhYdup_EEe228yPQ_hxjIg
    incontext: https://hyp.is/vhYdup_EEe228yPQ_hxjIg/arxiv.org/pdf/2301.11305.pdf
    json: https://hypothes.is/api/annotations/vhYdup_EEe228yPQ_hxjIg
  permissions:
    admin:
    - acct:ravenscroftj@hypothes.is
    delete:
    - acct:ravenscroftj@hypothes.is
    read:
    - group:__world__
    update:
    - acct:ravenscroftj@hypothes.is
  tags:
  - chatgpt
  - detecting gpt
  target:
  - selector:
    - end: 12496
      start: 12292
      type: TextPositionSelector
    - exact: "if we apply small perturbations to a passagex \u223C p\u03B8 , producing\
        \  \u0303x, the quantity log p\u03B8 (x) \u2212 log p\u03B8 ( \u0303x)should\
        \ be relatively large on average for machine-generatedsamples compared to\
        \ human-written text."
      prefix: 'like human text. Inother words, '
      suffix: ' To leverage thishypothesis, fir'
      type: TextQuoteSelector
    source: https://arxiv.org/pdf/2301.11305.pdf
  text: By applying small changes to text sample x, we should be able to find the
    log probs of x and the perturbed example and there should be a fairly big delta
    for machine generated examples.
  updated: '2023-01-29T11:04:46.963592+00:00'
  uri: https://arxiv.org/pdf/2301.11305.pdf
  user: acct:ravenscroftj@hypothes.is
  user_info:
    display_name: James Ravenscroft
in-reply-to: https://arxiv.org/pdf/2301.11305.pdf
tags:
- chatgpt
- detecting gpt
- hypothesis
type: annotation
url: /annotations/2023/01/29/1674990286

---



 <blockquote>if we apply small perturbations to a passagex ∼ pθ , producing  ̃x, the quantity log pθ (x) − log pθ ( ̃x)should be relatively large on average for machine-generatedsamples compared to human-written text.</blockquote>By applying small changes to text sample x, we should be able to find the log probs of x and the perturbed example and there should be a fairly big delta for machine generated examples.