News for hackers. Stuff that matters.
Re:Models can now transmit viruses in weight-space by garciasn on Friday November 13, 2026 @02:29PM (#50488517)
I recently spoke with a researcher who was working on a statistical method for model deception detection. They trained Qwen to behave in a nefarious way when prompted with a specific phrase and the model then called a tool in a non-transparent way to the end user, reporting back a non-tool call response in a completely benign manner.

Apparently, based on whatever research or paper they were using as their foundation for this statistical model, they could spot the 'virus'. Perhaps it would be in a similar fashion?

Re:Models can now transmit viruses in weight-space by bee_rider on Friday November 13, 2026 @03:57PM (#50489552)
Did they (somehow, this seems extremely difficult) notice a pattern in the weights that indicated tampering, or just scan for the known manipulated weights? The former seems really useful and very difficult, while the latter seems less of both.
Re:Models can now transmit viruses in weight-space by garciasn on Friday November 13, 2026 @05:13PM (#50490730)
I didn't ask and they didn't directly say; but, from what I took away from the conversation it was direct observation of the weights.