With Dutch parliamentary elections next week, explore this interactive AI-powered analysis of every vote since December 2023. See parties’ stances by topic, impact and beneficiaries (NB in Dutch). datascience.web.rug.nl/parliamentar...

The model is the product. New blog post on what the latest research trends mean for the next commercial cycle: specialized models behaving like an integrated systems, model providers moving up to application layer, training or being trained on. vintagedata.org/blog/posts/m...

I'll get straight to the point. We trained 2 new models. Like BERT, but modern. ModernBERT. Not some hypey GenAI thing, but a proper workhorse model, for retrieval, classification, etc. Real practical stuff. It's much faster, more accurate, longer context, and more useful. 🧵

“They said it could not be done”. We’re releasing Pleias 1.0, the first suite of models trained on open data (either permissibly licensed or uncopyrighted): Pleias-3b, Pleias-1b and Pleias-350m, all based on the two trillion tokens set from Common Corpus.

🪼 Welcome JUST-OS! JUST-OS is an exciting initiative by researchers at the University of Groningen and FORRT (@forrt.bsky.social; forrt.org). We’re developing an AI-based chatbot to simplify navigating Open Science resources.

Try it out: openscholar.allen.ai Read more: allenai.org/blog/opensch... Paper: openscholar.allen.ai/paper

Releasing two trillion tokens in the open. huggingface.co/blog/Pclangl...

Introducing Early American HistoriChat, a chatbot trained on the EvansTCP corpus, ~5,000 American printed texts from 1640 to 1800: eahc.mhvdr.nl Inspired by the work of @dorialexander.bsky.social; designed & built by @michielree.bsky.social in collaboration with MLT & the H-GEAR project

liking 10 @vickiboykis.com posts to bootstrap this bsky algorithm