How do we stop model collapse in AI autophagy? In our ACM TIST paper, we show how perplexity-filtering mitigates model collapse by filtering out low-quality synthetic data. 📄 Paper here: dl.acm.org/doi/10.1145/... @danigambit.bsky.social

"Runctitiononal features"? "Medical fymblal"? "1 Tol Line storee"? This gets worse the longer you look at it. But it's got to be good, because it was published in Nature Scientific Reports last week: www.nature.com/articles/s41... h/t @asa.tsbalans.se

Thrilled to announce the Best Poster Awards at CCS2025 go to Simay Atasoy Bingöl, Salvatore Citraro, and Elisa Muratore — congratulations! 🏆✨ #CCS2025

Our new #arXiv paper out! "Learning by Surprise": #Surplexity for Mitigation Model Collapse in Generative AI A new characterization of #modelcollapse in #ai #autophagy and a mitigation inspired by #FreeEnergyPrinciple from cognitive science @css-conference.bsky.social arxiv.org/abs/2410.12341

How to conduct experiments to study the effects of #AI based recommender systems? About our survey on arXiv...

Luca Pappalardo@lucapappalardo.bsky.social · last yr.

How do recommender systems shape our behavior? 📄 We review 144 studies across domains, from social media to GenAI. 🧭 With the EU Digital Services Act, data access is now possible. Our survey offers a guide to studying recommender-driven risks. 🔗 arxiv.org/abs/2407.01630

Preprint v2 is out! On @arxiv-cs-cl.bsky.social our update on #ModelCollapse of #AI #LLMs with new characterizations based on #Semantic #Networks and #NextToken predictions, new insights about the collapse effect and possible mitigation methods. @lucapappalardo.bsky.social arxiv.org/abs/2410.12341

Characterizing Model Collapse in Large Language Models Using Semantic Networks and Next-Token Probability

As synthetic content increasingly infiltrates the web, generative AI models may experience an autophagy process, where they are fine-tuned using their own outputs. This autophagy could lead to a pheno...

arxiv.org

create the llm model collapse you want to see in the world

Everything you always wanted to ask about entropy but didn't know whom by John Baez. arxiv.org/abs/2409.09232

What is Entropy?

This short book is an elementary course on entropy, leading up to a calculation of the entropy of hydrogen gas at standard temperature and pressure. Topics covered include information, Shannon entropy...

arxiv.org

Hey blueskiers! In case you missed it, we just uploaded the arXiv for our latest work! This project has been such an amazing journey, thanks to the brilliant collaboration with @nicopedre.bsky.social @lucapappalardo.bsky.social and Renaud Lambiotte. Give it a look :) arxiv.org/abs/2410.18004

Dynamic models of gentrification

The phenomenon of gentrification of an urban area is characterized by the displacement of lower-income residents due to rising living costs and an influx of wealthier individuals. This study presents ...

arxiv.org

We are hiring a postdoc/researcher at my group (SUNLab) @nunetsi.bsky.social. Interested in working on multidisciplinary problems like segregation, health, or economic growth from the perspective of Social Urban Networks? Contact me if you are interested. Please share it!

Anyone here on #bluesky studying #LLMs? Here's our preprint about #autophagy and linguistic analysis of #ModelCollapse we are working on... #ai #phd #research arxiv.org/abs/2410.12341

A linguistic analysis of undesirable outcomes in the era of generative AI

Recent research has focused on the medium and long-term impacts of generative AI, posing scientific and societal challenges mainly due to the detection and reliability of machine-generated information...

arxiv.org