flâneur

Debate with Self-Play Best-of-N Optimization — LessWrong

lesswrong.com · saved by 2 readers

Context: This is the first research output from Arcadia Alignment’s scalable oversight team, carried out in collaboration with external researchers a…

saved by