In ASSET Research Group's tests, GPT-4o showed 0% compliance in a one-piece instruction test but 100% when instructions were split into two pieces, demonstrating susceptibility to the GhostSplice attack.
In ASSET Research Group's tests, GPT-4o showed 0% compliance in a one-piece instruction test but 100% when instructions were split into two pieces, demonstrating susceptibility to the GhostSplice attack.