How much should a historian trust a confession extracted by torture?

A confession extracted by torture is always very tricky as to whether it should be trusted or not, so any historian should take this kind of confessions with reserve because it is always a 50-50 chance. Some of these confessions are true, but only in cases were the tortured one really knows the information and is not willing to give up on his/hers life by keeping it as a secret. On the other hand there a lots of confessions extracted by torture that don't correspond with the truth, but the tortured person admitted/told something that is not something he/she has done or really knows because they wanted to save their life (though very often it didn't really helped but was just used against them and as a propaganda).

i think it is d korean