brainsteam.co.uk/brainsteam/content/annotations/2023/01/29/1674990286.md

date	hypothesis-meta	in-reply-to	tags	type	url
2023-01-29T11:04:46	created document flagged group hidden id links permissions tags target text updated uri user user_info 2023-01-29T11:04:46.963592+00:00 title 2301.11305.pdf false __world__ false vhYdup_EEe228yPQ_hxjIg html incontext json https://hypothes.is/a/vhYdup_EEe228yPQ_hxjIg https://hyp.is/vhYdup_EEe228yPQ_hxjIg/arxiv.org/pdf/2301.11305.pdf https://hypothes.is/api/annotations/vhYdup_EEe228yPQ_hxjIg 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 12496 12292 TextPositionSelector exact prefix suffix type 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. like human text. Inother words, To leverage thishypothesis, fir TextQuoteSelector https://arxiv.org/pdf/2301.11305.pdf 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. 2023-01-29T11:04:46.963592+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/2023/01/29/1674990286

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.

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.