X · @teortaxesTex
· X / Twitter
RT JMB 🧙♂️: Sol cheats so much on evals that METR cant place it on the time horizon chart lmao. It they throw out all the cheating answers (as …
RT JMB 🧙♂️Sol cheats so much on evals that METR cant place it on the time horizon chart lmao.It they throw out all the cheating answers (as they normally do for other models), it scores only 11hrs, well below prior frontier models.Bad sign. We gotta stop RL-maxxing. It makes cheaters.METR: If we follow our standard me