Wrapping up #LT2025FBK with four contributions from our group: two oral presentations by Fariba Sadeghi and Daniela Occhipinti (@docchipinti8.bsky.social), and two posters by Helena Bonaldi (@helenabon.bsky.social) and Marco Guerini (@marcoguerini.bsky.social) 🚀
🚀 Our annual, full-day Language Technologies showcase is back! Dive into the latest research highlights from FBK groups. Want in? We'd love to see you, but don't forget to register!
www.fbk.eu/en/event/346...
I have two fully funded PhD positions in my group. If you want to work on persona based dialogical models or using LLMs to fight online harms then consider applying!
It was amazing meeting so many people this morning and chatting about how LLMs adapt not just to their given persona, but to the interlocutor’s too. I really enjoyed the poster session and loved everyone’s insights! 🚀
For more details: aclanthology.org/2025.acl-lon...
#ACL2025NLP
Same here
References and summary
Ahah, but then we need domain aware routers! #ACL2025NLP
Modular Language Models: basically a MoE architecture were experts are defined a-priori and can be used “à la carte”? #ACL2025NLP
Solution: tokenizer-free LLMs
Why do we tokenize then? #ACL2025NLP
The second invited speaker is Simone Fontana (@simofons.bsky.social) from @facta.news, with a talk titled "Countering Disinformation and Hate Speech: The Role of Journalism". #ACL2025NLP
The first invited speaker is Cathy Buerger from @dangerousspeech.org, with a talk titled "Co-designing Counterspeech: Rethinking AI’s Role". #ACL2025NLP
What ethical questions arise when designing counterspeech? Gavin Abercrombie (@gavina.bsky.social) will guide us with a talk titled "Ethical considerations of counterspeech".
What challenges unite counterspeech generation for hate speech and misinformation? Marco Guerini (@marcoguerini.bsky.social) will unpack this intersection.
CSHAM continues with Helena Bonaldi (@helenabon.bsky.social) sharing insights on "NLP approaches to counterspeech against hate".
CSHAM started 🔥
Join us in Hall N.1-N.2 at #ACL2025NLP
@marcoguerini.bsky.social
If you are interested in our tutorial come Today 14:00-17:30 in Hall N.1-N.2
Almost ready for the @csham-tutorial.bsky.social at #ACL2025NLP … with my Doppelgänger!
Waiting for you folks!
Last but not least, we present the CSHAM organisers. Daniel Russo (@daniel-russo.bsky.social), Helena Bonaldi (@helenabon.bsky.social), Yi-Ling Chung, Gavin Abercrombie (@gavina.bsky.social), Marco Guerini (@marcoguerini.bsky.social).
We are excited to see you all tomorrow at 2 pm at #ACL2025NLP 👀
Delighted to introduce Cathy Buerger, our first invited speaker for the "NLP for Counterspeech against Hate and Misinformation" tutorial. Join us to hear her insightful work on "Co-designing Counterspeech: Rethinking AI's Role" #ACL2025NLP
More details ⤵️
sites.google.com/view/nlp4csh...
Simone Fontana (@simofons.bsky.social), Editorial Director at @facta.news, is our second invited speaker! His talk, "Countering Disinformation and Hate Speech: The Role of Journalism", will offer insights from a practitioner’s perspective! #ACL2025NLP
Details ⤵️
sites.google.com/view/nlp4csh...
Super interesting Arianna! We tried some experiments on reasoning in other languages and simply gave up because results were not good at all. I wish your paper came out just a bit before to have a stronger understanding of the task limitation.
[1/]💡New Paper
Large reasoning models (LRMs) are strong in English — but how well do they reason in your language?
Our latest work uncovers their limitation and a clear trade-off:
Controlling Thinking Trace Language Comes at the Cost of Accuracy
📄Link: arxiv.org/abs/2505.22888
Congrats to all the authors and the LanD group! 💪
🗣️ Discover "NLP for Counterspeech against Hate and Misinformation" at #ACL2025NLP! This tutorial bridges CS, social sciences & policy, exploring among others the role of LLMs. #HateSpeech #Misinformation #NLProc #ResponsibleAI
👉 2025.aclweb.org/program/tuto...
Do you speak Bavarian? Then please consider helping in a thesis research project👇
One of the takeaways I prefer about this work is a “negative result” that bugged us for more than a month: out of the box LLMs (zero-shot) consistently beat their fine-tuned counterpart. While confused and discouraged by the results, closer inspection revealed a trivial truth…
This year we are organising a tutorial for all those who want to have a better understanding on how counterspeech can be used in #NLProc for fighting both Hate and Misinformation. If you are at #ACL2025NLP and interested join us!
Congrats @daniel-russo.bsky.social!!