#NAACL24 Come see the work of Domenic Anthony Rosati and David in the second interpretability session (11:00 am). If you are not around, see the recording here: #syntacticprobing: youtube.com/watch?v=J6kyFd… #modelediting: youtube.com/watch?v=8J3zlL…

The next conferences promotional videos presented at #NAACL2024 seem like a marketing pitch. Do we "scientists" need Hollywood-style videos to convince ourselves to submit a paper to a top NLP venue? Can we use this money to lower the barrier for underrepresented groups? It was

🇵🇸People are starving again in Northern Gaza, kids are dying as a result of malnutrition. In the video a father, helplessly, weeping for his kid. This is a man made catastrophe, it can be easily stopped; 🛑STOP babysitting Israel 🛑STOP babysitting the Israeli radicals and

It's hard to ignore! For folks who think this is politics and we should keep it away from science, I disagree. This is humanity and one of the goals of science is to improve and protect human life...

Despite having 100s of evaluation metrics to measure the progress of LLMs, they are still brittle to small changes in the input. I am excited to share SUGERCREPE++ benchmark that evaluates LLMs sensitivity to semantic and lexical alternations. My take home from the findings:

It’s disgusting that my colleagues would support legislation to prohibit U.S. officials from even citing the Palestinian death toll. They want to erase the Palestinians who are living, and now they are trying to erase the Palestinians who are dead. This is genocide denial.

5/n overall architecture of Chameleon

Flying to one the largest NLP conferences #ACL2024 with a heavy heart! While we fear AI to be unsafe for humans, we constantly lose the battle against humans keeping other humans safe. A few questions to ponder on: can we build AI that prevents humans from hurting other humans?

It was a pleasure giving a keynote at #Repl4nlp #ACL2024 on Latent Space Exploration for Safe and Trustworthy AI. I look forward to seeing works targeting various challenges in understanding deep learning models' complex latent space and their applications to developing better

Here is our pape: arxiv.org/abs/2408.07445

Thanks a lot, Preslav Nakov (Preslav Nakov) for hosting me at MBZUAI and for the opportunity to talk about some of the work we have been doing at Dalhousie University. It was a productive visit with many useful discussions to follow up on. #NLP

I am excited to share our two papers on Safe and Trustworthy AI accepted at EMNLP 2024 #EMNLP2024. Thanks to my awesome students and collaborators. Latent Concept-based Explanation of NLP Models arxiv.org/pdf/2404.12545 Immunization against harmful fine-tuning attacks