Clémentine Fourrier 🍊 is off till Dec 2026 (🪂)

Hey twitter!

I'm releasing the LLM Evaluation Guidebook v2!
Updated, nicer to read, interactive graphics, etc!
huggingface.co/spaces/OpenEva…

After this, I'm off: I'm taking a sabbatical to go hike with my dogs :D
(back @huggingface in Dec *2026*)

See you all next year!

Was in my first PhD jury this week, for @KovacGrgur's thesis!

My favorite results:
- Model "values" are incredibly unstable from prompt to prompt even if the convo is about stg neutral 🧵
- Data collapse won't happen uniformly but in bubbles across the web

Know anyone who need some help to get started with ML, open source and 🤗?

We partnered with @TechToTheRescue, a tech for good incubator, & answered all AI questions their non profits had to create an FAQ!

github.com/huggingface/faq

Come add your Q/As, it's collaborative! 🔥

Did you know the @MinistereCC (French gov) has started an LLM arena for 🇫🇷 about one year ago?

They got 182K votes, and are now publishing the results!

Congrats to @MistralAI for having the best French speaking model overall, and @deepseek_ai for the best open source one!

Good dive by @xeophon on OS world: what the tasks are, where they are broken, and yet another way evals can be unreproducible through time ^^

ht to @eliebakouch for the pointer!