发表机构
University of Waterloo（滑铁卢大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对RAG中查询主导导致证据失效的问题，提出GRIP方法，通过容量不对称设计优化信息编码，在五个推理基准上实现性能提升，大幅降低互信息与幻觉。
AI 中文摘要
检索增强生成（RAG）中的大容量编码器会让查询主导潜在状态，导致检索到的证据在功能上无关紧要，我们将这种失效模式称为查询主导。为解决该问题，我们提出GRIP（Grounded Reasoning via Information-Restricted Premises，基于信息受限前提的接地推理），该方法引入容量不对称性：解码器可全维度访问查询，而检索到的证据需经过严重的随机瓶颈，这迫使证据通道仅编码查询无法提供的残差信息。在五个推理基准上，GRIP优于强大的迭代基线，将查询-潜在互信息诊断值降低约30倍（从14.8比特降至0.47比特），并减少73%的幻觉。残差对齐分析进一步表明，瓶颈输出占据的子空间与查询的对齐度低于基线表示。
英文摘要
High-capacity encoders in retrieval-augmented generation (RAG) can let the query dominate the latent state, leaving retrieved evidence functionally irrelevant. We call this failure mode query dominance. To address it, we introduce \textbf{GRIP} (Grounded Reasoning via Information-Restricted Premises), which imposes capacity asymmetry: the decoder keeps full-dimensional access to the query, while retrieved evidence passes through a severe stochastic bottleneck. This forces the evidence channel to encode only the residual information unavailable from the query. Across five reasoning benchmarks, GRIP outperforms strong iterative baselines, cuts a query--latent mutual-information diagnostic by roughly 30$\times$ (14.8 $\to$ 0.47 bits), and reduces hallucination by 73\%. Residual-alignment analysis further shows that the bottleneck output occupies subspaces less aligned with the query than baseline representations.
Comments15 pages, 3 figures
