I finally uploaded to YT my 2022 ICML Test of Time Award Talk: Poisoning Attacks against SVMs: Ten Years After https://t.co/lrZPqNGZwS

oh well

ICLR GCs just accept all papers with rating above 4 max(before, after rebuttal) and call it a day.

We should withdraw all the submitted papers at this point... This is just unfair.

I call @iclr_conf to send an official email to ALL authors & reviewers regarding the leakage incident: - We need transparency in communication. A conference cannot rely on social media; a direct email ensures everyone gets the same facts. - We need clear guidance on the

An LLM-generated paper is in the top 17% of ICLR submissions in terms of average reviewer score, having received two 8's. The paper has tons of BS jargon and hallucinated references. Fortunately, one reviewer actually looked at the paper and gave it a zero. 1/3

Paper: https://t.co/t9QLf03GpZ Code: https://t.co/6IdkECieYY This work would not have been possible without our great @sAIferLab co-authors @FabioBrau, @biggiobattista, @LucaOneto, and @fabiogroli.

Our new paper, LatentBreak: Jailbreaking LLMs through Latent Space Feedback, is now on arXiv. We study how latent-space feedback can produce natural, low-perplexity jailbreaks. Joint work with brilliant colleagues across @sAIferLab @fdtn_ai https://t.co/rvgZaOCAym

Reference hallucinations still increased in July (data wasn't yet available when I posted early August), but have slightly decreased since then. Now that's what I call impact! I also uploaded some code to reproduce, in case this is useful for anyone: https://t.co/SMo9OtzZQH

I found a paper with this ref: - the title is from: https://t.co/RwKPrt7FcL - the author list is from: https://t.co/K98cXt5S2z - the link is https://t.co/g6z2IItHxG - in the text ref [1] is for: https://t.co/RRmIBhIFqG How did this happen? Seems too weird for a LLM hallucination

Can we compress neural networks while preserving robustness against adversarial attacks? Yes — that’s the goal of Adversarial Pruning (AP) methods. But here’s the catch 👇

📢 New Publication in Pattern Recognition – Volume 168 We’re pleased to share that our partner, @Università degli Studi di Cagliari (Italy), has published important new research on adversarial pruning methods in Pattern Recognition! Recent years have seen the rise of pruning

5. Random seed selection. A thing of beauty. 6/6

NeurIPS has a record-breaking number of 25,000 submissions this year already. It's crazy but not entirely unexpected. Here is the scaling over the years, with a conservative projection. How do we even run a conference and review at this massive scale?

NeurIPS submission counter over 22K 😂😂😂🔥🔥🔥

2. I was traveling when I got this, and replied on mobile that we could explore. They sent a calendly link and I booked a slot for 2 weeks ahead. 3. The meeting would be yesterday, and one hour before the meeting time I received an email with the link to join. It was on this

🚀 Exciting Announcement! 🚀 Get ready for the 18th ACM Workshop on Artificial Intelligence and Security (AISec 2025)! 📍Co-located: @acm_ccs 🗓️ Deadline: June 20th, 2025 🌐 Website: https://t.co/TGKGp4i95i w/ @ruoxijia and Matthew Jagielski

Last week, I shared two #ICLR2025 papers that were recognized by their Award committee. Reflecting on the outcome, I thought it might be interesting to share that both papers were previously rejected by #NeurIPS2024. I found the dramatic difference in reviewer perception of

@elsa_lighthouse @sec4ai4sec

Had a great time presenting our paper "σ-zero: Gradient-based Optimization of ℓ₀-norm Adversarial Examples" at the #ICLR2025! poster session!🎤🔥🔥🔥 📄 Paper: https://t.co/czeg1hDkUY 💻 Code: https://t.co/RJHHE3JSsz

Heading to Singapore for #ICLR, where I and Antonio Cinà will present our poster on sigma0, an L0-norm attack to compute sparse adversarial examples. Drop me a line if you want to catch up! 🔥🔥🔥