Edoardo Debenedetti @NeurIPS

I am at NeurIPS 🇨🇦, please reach out if you want to grab a coffee!

SPY Lab is in Vancouver for NeurIPS! Come say hi if you see us around 🕵️

I'm in Vancouver for NeurIPS! Feel free to reach out if you wanna meet to chat about security and privacy, especially in the context of LLM agents!

Come do open AI with us in Zurich!
We're hiring PhD students, postdocs (and faculty!)

Feel free to recommend @javirandor.com more researchers to add to the list!

Apropos of today's Overleaf downtime/slowness: remember to have your files backed up on Github or locally! What if this happened on the day of a conference deadline?

Anyone may be able to compromise LLMs with malicious content posted online. With just a small amount of data, adversaries can backdoor chatbots to become unusable for RAG, or bias their outputs towards specific beliefs. Check our latest work! 👇🧵

Gradient Masking All-at-Once: Ensemble Everything Everywhere Is Not Robust Ensemble everything everywhere is a defense to adversarial examples that was recently proposed to make image classifiers robust. This defense works by ensembling a model's intermediate representations...

Ensemble Everything Everywhere is a defense against adversarial examples that people got quite exited about a few months ago (in particular, the defense causes "perceptually aligned" gradients just like adversarial training)

Unfortunately, we show it's not robust...

arxiv.org/abs/2411.14834