Yeah, so I think there's this asymmetry in, in the attack and defense.

Yeah, we have been in cybersecurity industry for many years, and it's like, it, it's just like the KMOU scam, you know? Yeah, you will always come out with a fancier attack technique, and then we, as a blue team or the, the security guy need to, you know, catch up the, the attack.

So I, we believe that as the evaluation of the, uh, LLM Next model are progressing, the similar attack will immersion like, uh, something like this.

Yeah

We see that there is a certain degree of transferability.

Yeah.

And also we, we already, you know, put all the material together and then, uh, submit to the, uh, one of the, our prestige AI conference called ICML, and we actually got an acceptance about few days ago.

So actually, the audience, if they are really interesting in how we did it, and then they can just check out our paper as well.