偏好并非干预：特定读者证据效用的结构与稳定性边界
Preference Is Not Intervention: The Structure and Stability Boundaries of Reader-Specific Evidence Utility
浏览论文内容
中文总结 AI 辅助
该研究在RAG中测试模型特定差异是否形成可复用结构，发现读者序数偏好几何跨任务稳定但符号几何受任务限制，且稳定偏好无法预测干预迁移，表明特定读者效用存在但偏好≠干预。
中文摘要 AI 辅助
机器学习系统越来越多地基于下游模型身份来调整决策，但只有当特定模型差异形成可复用结构而非输入局部交互时，这种调整才有用。我们在检索增强生成（RAG）中对此进行了测试，其中证据效用可在受控干预下测量。在保持查询、证据、任务、评分和干预固定的情况下，9位读者在33%的共同影响单元上对效应符号存在分歧；读者×查询交互解释了29.8%的效用方差，而排列零假设仅解释8.4%；自选证据使F1提升了+0.031（t=3.39）。随后我们提出更尖锐的问题：这种异质性的哪些成分是跨查询的稳定读者属性？我们区分了三个可测量对象——证据活动性、序数偏好和条件符号方向，发现序数读者几何在四个独立设置中稳定（折半ρ=0.60–0.83）：留一法干预、PRISM偏好、RAMDocs和RAGuard。符号几何受任务限制：在开放式问答中较弱（0.14、0.35），尤其是对于误导性和不相关证据，但在二元事实核查中较强（0.75），无显著序数差距，尽管仍低于其稀疏性匹配上限。稀疏性、解码噪声和度量伪影无法解释主要的序数-符号差距。最后，稳定的序数相似性无法预测跨读者干预迁移（ oracle距离ρ=-0.27；遗憾可靠性-0.28）。特定读者效用确实存在，但偏好并非干预：稳定的排序相似性不允许帮助/伤害决策的迁移。
英文摘要
ML systems increasingly condition decisions on downstream model identity, but this is useful only if model-specific differences form reusable structure rather than input-local interactions. We test this in retrieval-augmented generation (RAG), where evidence utility can be measured under controlled interventions. Holding query, evidence, task, scoring, and intervention fixed, nine readers disagree on effect sign in 33\% of jointly affected cells; reader$\times$query interaction explains 29.8\% of utility variance versus an 8.4\% permutation null; and self-selected evidence improves F1 by $+0.031$ ($t=3.39$). We then ask the sharper question: \emph{which components of this heterogeneity are stable reader properties across queries?} Separating three measurable objects---evidence \emph{activity}, \emph{ordinal preference}, and \emph{conditional signed direction}---we find ordinal reader geometry stable across four independent settings (split-half $ρ=0.60$--$0.83$): leave-one-out interventions, PRISM preferences, RAMDocs, and RAGuard. Signed geometry is task-bounded: weak in open-ended QA (0.14, 0.35), especially for misleading and irrelevant evidence, but strong in binary fact-checking (0.75) with no significant ordinal gap, though still below its sparsity-matched ceiling. Sparsity, decoding noise, and metric artifacts do not explain the main ordinal--signed gap. Finally, stable ordinal similarity fails to predict cross-reader intervention transfer (oracle-distance $ρ=-0.27$; regret reliability $-0.28$). Reader-specific utility exists, but preference is not intervention: stable ranking similarity does not license transfer of help/harm decisions.
发表机构
- College of Software, Jilin University（吉林大学软件学院）
机构由 AI 辅助整理，请以论文原文为准。
