PREVIEW. This is how it will look in an inbox; it was not sent to you. Each real email carries its own unsubscribe link and Keryn's postal address.

Keryn Headlines

Daily, Monday 3 February 2025 · Retrospective · Covers 2 February 2025 00:00 to 3 February 2025 00:00 UTC

Read on the web

Top developments

Blink of an eye: a simple theory for feature localization in generative models

arXiv · Research · Models

PyPI adds project archiving system to stop malicious updates

BleepingComputer · Analysis · SCA

AgentBreeder: Mitigating the AI Safety Risks of Multi-Agent Scaffolds via Self-Improvement

arXiv · Research · Models

Models

`From Prompt to Perturbation': An Adaptive Framework for Voice-Based Jailbreaks on Audio LLMs

arXiv · Research · Models

"I am bad": Interpreting Stealthy, Universal and Robust Audio Jailbreaks in Audio-Language Models

arXiv · Research · Models

Safety Alignment Depth in Large Language Models: A Markov Chain Perspective

arXiv · Research · Models

Titles, links and short excerpts only; every item links to its source. Selection is automatic; the editor's note is AI-drafted and checked as described. Preview: the unsubscribe link and Keryn's postal address are added to each email when it is sent.