AI RESEARCH
Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures
arXiv CS.AI
•
ArXi:2603.16475v1 Announce Type: new Schema-guided reasoning pipelines ask LLMs to produce explicit intermediate structures -- rubrics, checklists, verification queries -- before committing to a final decision. But do these structures causally determine the output, or merely accompany it? We