Peter Romov

AI Security & Privacy Researcher
PhD @imperialcollege

[Peter Romov @romovpa]

Autoresearch can discover SOTA white-box adversarial attacks on LLMs. We gave Claude 30+ existing GCG-like algorithms and access to a compute cluster, and it quickly learned to combine them into new methods that outperform all existing ones. Here’s what that looks like:

[Nataša Krčo @povrcka]

🎙️ New preprint: You Are the Trigger: Speaker-Activated Backdoors
We introduce a backdoor attack on speech models where the trigger is a target speaker’s voice. The model behaves normally for everyone else, but switches to attacker-chosen behavior when it hears the target. 1/

[Alexander Panfilov @kotekjedi_ml]

We can finally talk about it:

We found a way to extract hidden reasoning of frontier models using a vulnerability in the APIs of every frontier AI company.

We verified that our reasoning token count matches billed API thinking tokens 1:1 for most of the prompts we queried.

[Daria Soboleva @dmsobol]

If you've wondered where I've been recently, I gave an interview. This is a very difficult time for me. I'm trying to bring my mom home, and I could use your help spreading the word. Please watch, and please share.

interview: https://t.co/rloipKmoWW

website: https://t.co/Z7pubzoP0A

[Alexander Panfilov @kotekjedi_ml]

New paper: We deploy Claude Code in an autoresearch loop to discover novel jailbreaking algorithms – and it works. It beats 30+ existing GCG-like attacks (with AutoML hyperparameter tuning)

This is a strong sign that incremental safety and security research can now be automated.

[Peter Romov @romovpa]

Finally, the paper: https://t.co/2tJ8Df3WdW

[Peter Romov @romovpa]

Autoresearch can discover SOTA white-box adversarial attacks on LLMs. We gave Claude 30+ existing GCG-like algorithms and access to a compute cluster, and it quickly learned to combine them into new methods that outperform all existing ones. Here’s what that looks like:

[Peter Romov @romovpa]

Huge thanks to my amazing collaborators @kotekjedi_ml @_igorshilov @yvesalexandre @jonasgeiping @maksym_andr! This is my first release as a PhD student at @imperialcollege , which I joined recently.

[Peter Romov @romovpa]

@Dota_Science Приходи на секцию про eSports https://t.co/8OmT5i51U2

[Peter Romov @romovpa]

@wylsacom Добро пожаловать в клуб) главное никаких документов в тачке не оставлять

[ODS.ai @ods_ai]

Congrats to our teams. Finished 3rd & 7th in Planet: Understanding the Amazon from Space @kaggle #DeepLearning
competition! feat @viglovikov

[D(o|a)ta Science @Dota_Science]

Первый в мире пабстомп&хакатон по анализу данных #dota2 начался! @v1lat @Casperenush #ruhub #DotaScienceChallenge

[Aleksandr Semenov @semenoffalex]

Me, @romovpa, @v1lat & @KoTHunt are confused by such a low accuracy of ML algorithms in win prediction for #dota2.

[Aleksandr Semenov @semenoffalex]

Yay! Testing #kaggle in-class competition for #dota2 win prediction for the new #coursera ML course at from #yandex!