打破链条：对LLM对中间结构忠实性的因果分析
Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures
AI总结：
研究探讨了在模式引导推理（SGR）管道中，LLM对中间结构的忠实性，发现尽管模型在自身中间结构上自洽，但对干预的响应不足，当将最终决策的推导委托给外部工具时，这种不稳定性显著降低。
摘要：
在模式引导推理（SGR）管道中，LLM会在做出最终决定前生成显式的中间结构——如rubrics、checklists或验证查询。SGR因其承诺可控性而被越来越多采用：从业者期望能够检查、编辑和覆盖这些结构以引导结果。但这一承诺是否成立？我们引入了一种因果评估协议来衡量它：通过选择任务，其中确定性函数将中间结构映射到决定，每次受控编辑都意味着一个唯一的正确输出。在12个模型和4个基准测试中，模型在自身中间结构上自洽，但干预后预测未能更新——揭示出当中间结构变化时，看似忠实的特性变得脆弱。当最终决策的推导被委托给外部工具时，这种脆弱性大大消失；更强的提示仅带来有限的改进，而偏好优化显著提高了干预的忠实性。总体而言，模式引导管道中的中间结构起着影响性上下文的作用，而非稳定的因果中介。
英文摘要：
In schema-guided reasoning (SGR) pipelines, LLMs produce explicit intermediate structures -- rubrics, checklists, or verification queries -- before committing to a final decision. SGR is increasingly adopted because it promises controllability: practitioners expect to inspect, edit, and override these structures to steer the outcome. But does the promise hold? We introduce a causal evaluation protocol to measure it: by selecting tasks where a deterministic function maps intermediate structures to decisions, every controlled edit implies a unique correct output. Across 12 models and 4 benchmarks, models appear self-consistent with their own intermediate structures but fail to update predictions after intervention -- revealing that apparent faithfulness is fragile once the intermediate structure changes. When derivation of the final decision from the structure is delegated to an external tool, this fragility largely disappears; stronger prompting yields only limited improvements, while preference optimization substantially improves intervention faithfulness. Overall, intermediate structures in schema-guided pipelines function as influential context rather than stable causal mediators.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
