Testing sounds great though I’d note that the way I’d approach testing is to first construct a general test bed that produces problematic behavior through training incentives for o1-like models. (Possibly building on deepseek’s version of o1.) Then I’d move to trying a bunch of stuff in this setting.
(I assume you agree, but thought this would be worth emphasizing to third parties.)
Testing sounds great though I’d note that the way I’d approach testing is to first construct a general test bed that produces problematic behavior through training incentives for o1-like models. (Possibly building on deepseek’s version of o1.) Then I’d move to trying a bunch of stuff in this setting.
(I assume you agree, but thought this would be worth emphasizing to third parties.)