Reasoning Chain Break

A defect in large models performing multi-step reasoning tasks in which intermediate steps exhibit logical leaps, internal contradictions, or omission of key links, resulting in an erroneous final conclusion.