记忆并非总是必需：科学推理中条件记忆的特征分析

中文摘要 AI 辅助

科学推理要求语言模型检索专业知识并将其可靠地融入多步骤计算。条件记忆提供了一条显式的查找路径，以补充稠密的神经表征，但其有用性本质上取决于输入和计算：检索到的信息可能修复缺失的科学关联，但也可能引入分散注意力的捷径，或干扰基础模型已能正确执行的推理。本研究系统探究了条件记忆应在何时、何地以及何种程度上参与科学推理，刻画了科学知识边界以及对启用记忆的知识回路节点的可控干预。基于这些分析，我们提出了一种知识边界感知路由器（Knowledge Boundary-Aware Router），它利用生成前可用的特定任务输入代理来确定是否激活记忆、哪些层阶段节点接收记忆信号，以及这些信号的贡献强度。在涵盖两个主干模型家族和六种任务类型的生物与化学推理基准上开展的实验表明，记忆效应在不同输入、任务和注入位置间存在显著差异。与静态路由和匹配激活率的随机路由相比，我们的方法能更一致地保留有益的记忆贡献，同时抑制记忆引发的退化，确立选择性记忆分配是可靠科学推理的重要原则。

英文摘要

Scientific reasoning requires language models to retrieve specialized knowledge and incorporate it reliably into multi-step computation. Conditional memory provides an explicit lookup pathway that complements dense neural representations, but its usefulness is inherently input- and computation-dependent: retrieved information may repair missing scientific associations, yet it may also introduce distracting shortcuts or interfere with reasoning that the base model can already perform correctly. In this work, we systematically investigate when, where, and to what extent conditional memory should participate in scientific reasoning. We characterize the scientific knowledge boundary and controlled interventions on memory-enabled knowledge-circuit nodes. Based on these analyses, we propose a Knowledge Boundary-Aware Router that uses task-specific input proxies available before generation to determine whether memory is activated, which layer-stage nodes receive memory signals, and how strongly these signals contribute. Experiments on biological and chemical reasoning benchmarks, covering two backbone families and six task types, show that memory effects vary substantially across inputs, tasks, and injection locations. Compared with static and activation-rate-matched random routing, our approach more consistently preserves beneficial memory contributions while suppressing memory-induced regressions, establishing selective memory allocation as an important principle for reliable scientific reasoning.

发表机构

• Huzhou Normal University（湖州师范大学）
• Alibaba Group（阿里巴巴集团）
• Bota Biosciences（博塔生物科技）
• Zhejiang University（浙江大学）

机构由 AI 辅助整理，请以论文原文为准。