Very excited to be joining AI at Meta GenAI as a Visiting Researcher starting this June in New York City!🗽 I’ll be continuing my work on studying memorization and safety in language models. If you’re in NYC and would like to hang out, please message me :)

📢 ACL 2025 notifications have been sent out, making this the perfect time to finalize your commitment. Don't miss the opportunity to be part of the workshop! 🔗 Commit here: openreview.net/group?id=aclwe… 🗓️ Deadline: May 20, 2025 (AoE) #ACL2025 #NLProc

the closest thing i've seen to actual "physics for LMs" was probably this (single-author!) paper from neurips 2024: Understanding Transformers via N-Gram Statistics this is how we used to think about LMs; not sure why we stopped.

Following on Andrej Karpathy's vision of software 2.0, we've been thinking about *malware 2.0*: malicious programs augmented with LLMs. In a new paper, we study malware 2.0 from one particular angle: how could LLMs change the way in which hackers monetize exploits?

We (w Zachary Novack Jaechul Roh et al.) are working on #memorization in #audio models & are conducting a human study on generated #music similarity. Please help us out by taking our short listening test (available in English, Mandarin & Cantonese). You can do more than one! Link ⬇️

Excited to announce we'll be hosting the 2nd U&Me workshop at #ICCV2025 2025 in October. If you work in unlearning or model editing you should consider submitting something or participating in the challenge ( top winners will be co-authors on workshop challenge paper )

Signal boosting this awesome opportunity to join the SAM Team at AI at Meta FAIR! Please apply using the link in the post! Several folks from the SAM team will be at CVPR in a couple of weeks, reach out if you want to chat! 👋🏾

The privacy-utility tradeoff for RAG is much worse than you think. Our work detailing one such (stealthy) MIA, now accepted to CCS'25 ACM CCS 2024 !

For trying to understanding LMs deeply, EleutherAI’s Pythia has been an invaluable resource: 16 LMs (70M to 12B parameters) trained on the same data (The Pile) in the same order, with intermediate checkpoints. It’s been two years and it’s time for a refresh.

Our new Google DeepMind paper, "Lessons from Defending Gemini Against Indirect Prompt Injections," details our framework for evaluating and improving robustness to prompt injection attacks.

Two years in the making, we finally have 8 TB of openly licensed data with document-level metadata for authorship attribution, licensing details, links to original copies, and more. Hugely proud of the entire team.

🪄We made a 1B Llama BEAT GPT-4o by... making it MORE private?! LoCoMo results: 🔓GPT-4o: 80.6% 🔐1B Llama + GPT-4o (privacy): 87.7% (+7.1!⏫) 💡How? GPT-4o provides reasoning ("If X then Y"), the local model fills in the blanks with your private data to get the answer!

now a part of Meta superintelligence labs! exciting times!

🍿 the place to be to meet some v cool interpretability folks (incl my phd friends) :)

some really cool ICML Conference trustworthy ml papers here!

Introducing SmolLM3: a strong, smol reasoner! > SoTA 3B model > dual mode reasoning (think/no_think) > long context, up to 128k > multilingual: en, fr, es, de, it, pt > fully open source (data, code, recipes) huggingface.co/blog/smollm3

We are raising $20k (which amounts to $800 per person), to cover their travel and lodging to Kigali, Rwanda in August, from either Nigeria or Ghana. Donate what you can here! donorbox.org/mlc-nigeria-de…

Are you a researcher, trying to build a small GPU cluster? Did you already build one, and it sucks? I manage USC NLP’s GPU cluster and I’m happy to offer my expertise. I hope I can save you some headaches and make some friends. Please reach out!