News for hackers. Stuff that matters.

A comment on Show HN: ArXiv Feed – The easiest way to keep up with AI research. Parent comment

Re:Show HN: ArXiv Feed – The easiest way to keep up with AI ... by subtech on Tuesday January 17, 2023 @04:19PM (#34419344)
GPT-3! Thank god I don't have to hand write them lol.
[ Parent | on HN ]
Re:Show HN: ArXiv Feed – The easiest way to keep up with AI ... by cloudking on Tuesday January 17, 2023 @05:42PM (#34420558)
How are you getting past the token limit?
[ Parent | on HN ]
Re:Show HN: ArXiv Feed – The easiest way to keep up with AI ... by pilotneko on Wednesday January 18, 2023 @08:46AM (#34426528)
Likely only summarizing the abstracts.
[ Parent | on HN ]
Re:Show HN: ArXiv Feed – The easiest way to keep up with AI ... by subtech on Wednesday January 18, 2023 @11:04AM (#34428417)
Yup only abstracts for now, but I think it should be possible to parse the paper PDF to extract just the approach + experiment results w/ GPT-3 (despite 4k token limit). Will add this soon.
[ Parent | on HN ]
Re:Show HN: ArXiv Feed – The easiest way to keep up with AI ... by pilotneko on Thursday January 19, 2023 @08:49AM (#34440199)
I would suggest sentence tokenization followed by clustering to pull the most representative sentence from each cluster. Unless you want to go abstractive, that is.

If you run into issues with the token limit, there are some nice architectures for handling longer inputs (e.g., Longformer, Reformer, etc.).

[ Parent | on HN ]
Re:Show HN: ArXiv Feed – The easiest way to keep up with AI ... by subtech on Friday January 20, 2023 @05:38PM (#34461086)
interesting ideas, thanks!
[ Parent | on HN ]