Arianna Bisazza
@AriannaBisazza
AssociateProfessor #NLProc @univgroningen | Multilingualism | Intrepretability | Language Learning in Humans vs NeuralNets | Mum^2
ID:882521793239691264
http://www.cs.rug.nl/~bisazza 05-07-2017 08:49:26
336 Tweets
1,0K Followers
244 Following
Highlight of my Japan tour 🤩 Giving a talk at YoheiOseki’s lab 東京大学 | UTokyo, enjoying group lunch & meeting “language acquisition sisters” Miyu Oba Akiyo Fukatsu Akari Haga to find out we share so many research interests! Thanks for your hospitality, I hope we meet again soon!
![[1/4] Introducing “A Primer on the Inner Workings of Transformer-based Language Models”, a comprehensive survey on interpretability methods and the findings into the functioning of language models they have led to. ArXiv: arxiv.org/pdf/2405.00208](https://pbs.twimg.com/media/GMpDM85WcAELzoF.jpg "Javier Ferrando (@javifer_96) on Twitter photo 2024-05-03 08:49:36 [1/4] Introducing “A Primer on the Inner Workings of Transformer-based Language Models”, a comprehensive survey on interpretability methods and the findings into the functioning of language models they have led to. ArXiv: arxiv.org/pdf/2405.00208")
I had lots of fun collaborating with Javier Ferrando on this!
Apart from methods and findings, we also have a comprehensive survey of LM interpretability tools and annotated equations using a unified notation 🙂 check it out!
arxiv.org/abs/2405.00208
Feeling fortunate to be part of this collaboration on the encoding of tone in speech models, just accepted to #NAACL2024 !
Speech is the next frontier for interpretability. I learned a lot in this process, and not only about models!
Check out Gaofei Shen’s thread ⬇️
Arianna Bisazza is presenting 'Inner workings and societal impacts of AI tools for text generation' at the 1871st KNG lecture at the Groningen Forum 🔍
Thrilled to announce Aya 🌿, a massively multilingual instruction-tuned LLM, featuring 101 languages and the largest collection of multilingual instruction datasets. Over half of these languages are under-resourced. A monumental effort from Cohere For AI and Aya team 🚀
I'm happy to announce that my work 'Are Character-level Translations Worth the Wait? Comparing ByT5 and mT5 for Machine Translation' with Gabriele Sarti, Antonio Toral, gertjanvannoord, and Arianna Bisazza was accepted to #TACL ! I'll be presenting it at #EACL2024 !
arxiv.org/abs/2302.14220
Interpretability research is popular these days, but public attention is still mostly focused on improving task performance.
To reverse this trend, I started curating a Hugging Face collection of daily picks in interpretability & analysis of LMs! 🔍 huggingface.co/collections/gs…