Tootfinder

Für die breite Verwendung von #KI, speziell im Kontext #Schule, muss sichergestellt sein, dass #LLMs user:innen nicht zu selbstgefährdendem Verhalten animieren.
Das Nonprofit Transluce arbeitet an verschie…

Surfacing Pathological Behaviors in Language Models
We train reinforcement learning (RL) agents to craft realistic natural-language prompts that elicit specified behaviors in frontier open-weight models (Llama 3.1/4, Qwen 2.5, and DeepSeek-V3), using a proposed variational lower bound to guide the search.

A General Coding Framework for Adaptive Private Information Retrieval
Jinbao Zhu, Xiaohu Tang
https://arxiv.org/abs/2506.07787 https://

A General Coding Framework for Adaptive Private Information Retrieval
The problem of $T$-colluding private information retrieval (PIR) enables the user to retrieve one out of $M$ files from a distributed storage system with $N$ servers without revealing anything about the index of the desired file to any group of up to $T$ colluding servers. In the considered storage system, the $M$ files are stored across the $N$ distributed servers in an $X$-secure $K$-coded manner such that any group of up to $X$ colluding servers learns nothing about the files; the storage ov…