#aligned

J-lens preferentially reads from coordinate-aligned subspaces

Note: I only test on Gemma-4-31B, which contains the spaces I've been studying in this earlier post . This is a shorter post, with just a few observations potentially relevant to those working on the J-lens work. The main observations here are: In Gemma-4, there are two relatively small subspaces that account the majority of the J-lens readout. These are both coordinate-aligned subspaces. That means we can take a hidden state, mask out all coordinates except both these ~600 coordinate groups, and it will capture ~80-95% of the J-lens readout. These spaces are not built to capture J-lens contri

LessWrong · EN

Aligned Data Can Induce Severe Misalignment via Context Confusion

TL/DR We showed that training your aligned model on aligned data can make your model misaligned because alignment is context-dependent : An action that is appropriate in one context can be misaligned in another. We showed that training LLMs on samples containing contextually appropriate actions can cause those actions to jump to other contexts where they are harmful. We called this phenomenon context confusion and provided model organisms (datasets creating context confusion), showed its existence in a popular instruction-tuning dataset, and provided a mechanistic explanation. We speculate tha

LessWrong · EN

China-Aligned Hackers Hijack Sogou Input Method DNS Requests to Deploy DarkVigil Backdoor

A China-aligned threat actor known as SlippyNunchucks has been attributed to a Domain Name System (DNS) hijacking operation abusing insecure HTTP communications in Windows versions of Sogou Input Method as part of a surveillance campaign designed to deploy an implant called DarkVigil (aka FDMTP). The campaign, observed starting July 1, 2026, is assessed to have targeted organizations and

The Hacker News · EN