Caleb Biddulph (@CalebBiddulph) profile picture
Joined December 2022
Caleb Biddulph retweeted
New Google DeepMind safety paper! LLM agents are coming – how do we stop them finding complex plans to hack the reward? Our method, MONA, prevents many such hacks, *even if* humans are unable to detect them! Inspired by myopic optimization but better performance – details in🧵
16
96
19
566
159,172