Our department is offering a pre-application review service (PARS) initiative to provide support and mentorship to PhD applicants from historically marginalized groups. See details here: stat.uw.edu/pre-applicatio…

Join us for a virtual conversation and Q&A with current Statistics Ph.D. students from the Departments of Statistics at UW, UC Berkeley, and UMich Ann Arbor on Thu., Nov. 3, 2022 at 5:00PM (PST). See details here: bit.ly/3SIAzkR

🤔How can we align AI systems/LLMs 🤖 to better represent diverse human values and perspectives?💡🌍 We outline a roadmap to pluralistic alignment with concrete definitions for how AI systems and benchmarks can be pluralistic! arxiv.org/abs/2402.05070 First, models can be…

This take on the FineWeb release is one of the most interesting feedback and also a reason FineWeb is very different from even larger datasets like RedPajama-V2 (which is double its size!) Surprisingly, the size of the dataset of 15T tokens is not very important, what is much

🚨 New Paper Alert! 🚨 Introducing 💿 StyleRemix: a novel, interpretable method for authorship obfuscation which perturbs fine-grained *style-elements* in the input text 🎨✍️. 📄 Paper: arxiv.org/abs/2408.15666 🧑‍💻 Code: github.com/jfisher52/Styl… 🚀 Demo: huggingface.co/spaces/hallisk…

I will be at @EMNLP in person to present StyleRemix: arxiv.org/abs/2408.15666 I am also officially on the job market for industry research positions focused on alignment, controllable generations, and AI impact on society. Message me if you'd like to chat about jobs or research!

Had a wonderful time presenting our work StyleRemix with Skyler Hallinan at Qualcomm Research & Technologies yesterday! Thank you Harris Teague for inviting us to come speak, we had a wonderful time meeting the team. arxiv.org/abs/2408.15666

I'm very excited to present my work on how biased language models affect human decision-making at Cohere For AI next Thursday, December 19th, at 10 am PST. Feel free to drop in. It will be open to the public!

Excited to be speaking at Stanford this Friday at the Designing Trustworthy AI in a Polarized World symposium! I’ll be around SF/Stanford for a few days—happy to grab a coffee and chat!

Want to know what training data has been memorized by models like GPT-4? We propose information-guided probes, a method to uncover memorization evidence in *completely black-box* models, without requiring access to 🙅‍♀️ Model weights 🙅‍♀️ Training data 🙅‍♀️ Token probabilities 🧵1/5

📢 Excited to announce the first workshop on ✨NLP for Democracy✨ (NLP4Democracy) at Conference on Language Modeling! 🗓️ Submissions (extended abstracts) due on June 19, 11:59pm AoE 🔗 Website: sites.google.com/andrew.cmu.edu… #COLM2025 #LLMs #NLP #NLProc #ComputationalSocialScience