ravenscroftj
/
brainsteam.co.uk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
brainsteam.co.uk/brainsteam/content/annotations/2023/01/29/1674989184.md

2.4 KiB
Raw Blame History

date hypothesis-meta in-reply-to tags type url
2023-01-29T10:46:24
created document flagged group hidden id links permissions tags target text updated uri user user_info
2023-01-29T10:46:24.271948+00:00
title
2301.11305.pdf
false __world__ false LNKuap_CEe2NNLuZfhdxTA
html incontext json
https://hypothes.is/a/LNKuap_CEe2NNLuZfhdxTA https://hyp.is/LNKuap_CEe2NNLuZfhdxTA/arxiv.org/pdf/2301.11305.pdf https://hypothes.is/api/annotations/LNKuap_CEe2NNLuZfhdxTA
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
31791 31366 TextPositionSelector
exact prefix suffix type
Figure 5. We simulate human edits to machine-generated text byreplacing varying fractions of model samples with T5-3B gener-ated text (masking out random five word spans until r% of text ismasked to simulate human edits to machine-generated text). Thefour top-performing methods all generally degrade in performancewith heavier revision, but DetectGPT is consistently most accurate.Experiment is conducted on the XSum dataset etectGPTLogRankLikelihoodEntropy .XSum SQuAD WritingPromptsMethod TextQuoteSelector
https://arxiv.org/pdf/2301.11305.pdf
DetectGPT shows 95% AUROC for texts that have been modified by about 10% and this drops off to about 85% when text is changed up to 24%. 2023-01-29T10:46:24.271948+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/2023/01/29/1674989184
Figure 5. We simulate human edits to machine-generated text byreplacing varying fractions of model samples with T5-3B gener-ated text (masking out random five word spans until r% of text ismasked to simulate human edits to machine-generated text). Thefour top-performing methods all generally degrade in performancewith heavier revision, but DetectGPT is consistently most accurate.Experiment is conducted on the XSum dataset
DetectGPT shows 95% AUROC for texts that have been modified by about 10% and this drops off to about 85% when text is changed up to 24%.