brainsteam.co.uk/brainsteam/content/annotations/2023/01/29/1674988556.md

date	hypothesis-meta	in-reply-to	tags	type	url
2023-01-29T10:35:56	created document flagged group hidden id links permissions tags target text updated uri user user_info 2023-01-29T10:35:56.649264+00:00 title 2301.11305.pdf false __world__ false tr0lTp_AEe2k81d5ilJ0Xw html incontext json https://hypothes.is/a/tr0lTp_AEe2k81d5ilJ0Xw https://hyp.is/tr0lTp_AEe2k81d5ilJ0Xw/arxiv.org/pdf/2301.11305.pdf https://hypothes.is/api/annotations/tr0lTp_AEe2k81d5ilJ0Xw admin delete read update acct:ravenscroftj@hypothes.is acct:ravenscroftj@hypothes.is group:__world__ acct:ravenscroftj@hypothes.is chatgpt detecting gpt selector source end start type 1096 756 TextPositionSelector exact prefix suffix type his approach, which we call DetectGPT,does not require training a separate classifier, col-lecting a dataset of real or generated passages, orexplicitly watermarking generated text. It usesonly log probabilities computed by the model ofinterest and random perturbations of the passagefrom another generic pre-trained language model(e.g, T5) is generated from a givenLLM. T . We find DetectGPT is more disc TextQuoteSelector https://arxiv.org/pdf/2301.11305.pdf The novelty of this approach is that it is cheap to set up as long as you have the log probabilities generated by the model of interest. 2023-01-29T10:35:56.649264+00:00 https://arxiv.org/pdf/2301.11305.pdf acct:ravenscroftj@hypothes.is display_name James Ravenscroft	https://arxiv.org/pdf/2301.11305.pdf	chatgpt detecting gpt hypothesis	annotation	/annotations/2023/01/29/1674988556

his approach, which we call DetectGPT,does not require training a separate classifier, col-lecting a dataset of real or generated passages, orexplicitly watermarking generated text. It usesonly log probabilities computed by the model ofinterest and random perturbations of the passagefrom another generic pre-trained language model(e.g, T5)

The novelty of this approach is that it is cheap to set up as long as you have the log probabilities generated by the model of interest.