#unstructured — Public Fediverse posts

[en] Paper: LLMs can be used to perform at-scale #deanonymization

"With full Internet access, our #agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given #pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator."

"Our results show that the practical #obscurity protecting pseudonymous users online no longer holds and that #threat models for online #privacy need to be reconsidered."

"We demonstrate that LLMs fundamentally change the picture, enabling fully automated deanonymization attacks that operate on #unstructured text at scale."

Note: also check paragraphs "Potential harms" and "Potential benefits".

https://arxiv.org/html/2602.16800

#llm #research

[en] Paper: LLMs can be used to perform at-scale #deanonymization

"With full Internet access, our #agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given #pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator."

"Our results show that the practical #obscurity protecting pseudonymous users online no longer holds and that #threat models for online #privacy need to be reconsidered."

"We demonstrate that LLMs fundamentally change the picture, enabling fully automated deanonymization attacks that operate on #unstructured text at scale."

Note: also check paragraphs "Potential harms" and "Potential benefits".

https://arxiv.org/html/2602.16800

#llm #research

[en] Paper: LLMs can be used to perform at-scale #deanonymization

"With full Internet access, our #agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given #pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator."

"Our results show that the practical #obscurity protecting pseudonymous users online no longer holds and that #threat models for online #privacy need to be reconsidered."

"We demonstrate that LLMs fundamentally change the picture, enabling fully automated deanonymization attacks that operate on #unstructured text at scale."

Note: also check paragraphs "Potential harms" and "Potential benefits".

https://arxiv.org/html/2602.16800

#llm #research

[en] Paper: LLMs can be used to perform at-scale #deanonymization

"With full Internet access, our #agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given #pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator."

"Our results show that the practical #obscurity protecting pseudonymous users online no longer holds and that #threat models for online #privacy need to be reconsidered."

"We demonstrate that LLMs fundamentally change the picture, enabling fully automated deanonymization attacks that operate on #unstructured text at scale."

Note: also check paragraphs "Potential harms" and "Potential benefits".

https://arxiv.org/html/2602.16800

#llm #research

[en] Paper: LLMs can be used to perform at-scale #deanonymization

"With full Internet access, our #agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given #pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator."

"Our results show that the practical #obscurity protecting pseudonymous users online no longer holds and that #threat models for online #privacy need to be reconsidered."

"We demonstrate that LLMs fundamentally change the picture, enabling fully automated deanonymization attacks that operate on #unstructured text at scale."

Note: also check paragraphs "Potential harms" and "Potential benefits".

https://arxiv.org/html/2602.16800

#llm #research

I’m starting to think that this is the easiest way to describe #Web3. #Structureddata uses a spreadsheet-like format that machines read with speed and confidence. #Unstructured data is everything else and needs formatting before machines can search, group, and analyze it.

#BackToSchool #Recess #GetOutside #Play #Unstructured

In California, "EC Section 49056 prohibits school staff members from restricting a student’s recess unless there is an immediate threat to the physical safety of the student or the physical safety of one or more of the student’s peers. If a student’s recess period is denied, school staff members shall make all reasonable efforts to resolve such threats and minimize exclusion from recess (EC Section 49056(a)(4)).” https://www.cde.ca.gov/fg//it/sb291letter.asp / https://www.kpbs.org/news/health/2024/01/30/for-the-first-time-california-law-will-protect-students-right-to-recess

Other states also address, e.g., in Texas, "Senate Bill 25 bars schools from taking away recess as punishment for younger students.” https://www.houstonpublicmedia.org/articles/news/texas/2025/08/31/529807/more-than-830-new-texas-laws-take-effect-sept-1-heres-whats-changing/

I'm reading, "Assisted Serendipity, Random Coffee and the power of the unstructured meeting" https://emilywebber.co.uk/assisted-serendipity-random-coffee-and-the-power-of-the-unstructured-meeting/ by @ewebber from March 2020.

#meetings
#serendipity
#unstructured
#randomcoffee

I've been wanting to try out #maturin for awhile now, and with some of the #LLM tinkering I've done at work, I finally had an excellent use case for it.
Its an opinionated #rust implementation of splitting #langchain documents as well as some #unstructured post processors. For cleaning and splitting, I've clocked it at between 40 and 75x faster than the python implementation, and on my machine it can clean and split 25,000 documents in a second.

Check it out at https://github.com/cam-barts/rs_document

And this is also interesting for the upcoming #textplusplenary @Textplus regarding the integration of #unstructured #data

The usually interesting Paul Graham has written a longgg piece (I guess c. 8k-10k words) which meanders a bit; it is a bit unstructured, and is a gathering of thoughts around getting ‘great work done’. However - that’s the point: it’s a really interesting and worthwhile read, with lots of nuggets of wisdom throughout. The journey is the point!

http://paulgraham.com/greatwork.html
#paulgraham #unstructured #work #wisdom

me learning mastodon is screaming #data lake vibes.... #unstructured af