Articles by tag: lock-r

Confirmation Lock in LLM-agents: LOCK-R test

LOCK-R stand proves bias of single LLM-agents and CoT paradox. Role separation Explorer-Judge reduces Bayes Regret to 0.09. Test on your models.