You’re putting the cart before the horse - alignment is an unsolved challenge (there are proposed approaches and active research on this) but it is still not es

You’re putting the cart before the horse - alignment is an unsolved challenge (there are proposed approaches and active research on this) but it is still not established (beyond theory) that latent reasoning is more capable than CoT on hard language reasoning, particularly at scale.