Exciting opportunity! 🎉 Joining SPY Lab has been one of the best decisions I've ever made🙈

What happened recently in AI/ML safety research (1/8) 🧵:

LLMs may be copying training data in everyday conversations with users! In our latest work, we study how often this happens compared to humans. 👇🧵

We looked into "Ensemble Everything Everywhere", an adversarial examples defense that caused some excitement. But Jie Zhang broke the current version: arxiv.org/abs/2411.14834 Good time to announce you can also find me somewhere over the rainbow: 🦋 bsky.app/profile/floria…

We are excited that this work has been accepted by SaTML Conference! We’ve put together a fun blog post, check it out here: spylab.ai/blog/mia_posit…

Adversarial ML research is evolving, but not necessarily for the better. In our new paper, we argue that LLMs have made problems harder to solve, and even tougher to evaluate. Here’s why another decade of work might still leave us without meaningful progress. 👇

At SpyLab we not only do great research but also have great fun 🏔️

I’ll be mentoring MATS for the first time this summer, together with Daniel Paleka! Link below to apply

Congrats, your jailbreak bypassed an LLM’s safety by making it pretend to be your grandma! But did the model actually give a useful answer? In our new paper we introduce the jailbreak tax — a metric to measure the utility drop due to jailbreaks.

The oral presentation of the jailbreak tax is tomorrow at 4:20pm in Hall 4 #6. The poster is up from 5pm. See you at ICLR Building Trust in LLMs Workshop. ICLR 2026

The Jailbreak Tax got a Spotlight award ICML Conference see you in Vancouver!

It’s been a wonderful time working, studying, and hanging out together 😭. Wishing you all the best in this exciting new chapter! 🙉

🎉 Announcing our ICML2025 Spotlight paper: Learning Safety Constraints for Large Language Models We introduce SaP (Safety Polytope) - a geometric approach to LLM safety that learns and enforces safety constraints in LLM's representation space, with interpretable insights. 🧵

How well can LLMs predict future events? Recent studies suggest LLMs approach human performance. But evaluating forecasters presents unique challenges compared to standard LLM evaluations. We identify key issues with forecasting evaluations 🧵 (1/7)