Constructivist epistemology argues that knowledge is actively constructed rather than passively copied. Despite the generative nature of Large Language Models (LLMs), most existing agent memory systems are still based on dense retrieval. However, dense retrieval heavily relies on semantic overlap or entity matching within sentences. Consequently, embeddings often fail to distinguish instances that are semantically similar but contextually distinct, introducing substantial noise by retrieving context-mismatched entries. Conversely, directly employing open-ended generation for memory access risks "Structural Hallucination" where the model generates memory keys that do not exist in the memory, leading to lookup failures. Inspired by this epistemology, we posit that memory is fundamentally organized by cognitive schemas, and valid recall must be a generative process performed within these schematic structures. To realize this, we propose SCG-MEM, a schema-constrained generative memory architecture. SCG-MEM reformulates memory access as Schema-Constrained Generation. By maintaining a dynamic Cognitive Schema, we strictly constrain LLM decoding to generate only valid memory entry keys, providing a formal guarantee against structural hallucinations. To support long-term adaptation, we model memory updates via assimilation (grounding inputs into existing schemas) and accommodation (expanding schemas with novel concepts). Furthermore, we construct an Associative Graph to enable multi-hop reasoning through activation propagation. Experiments on the LoCoMo benchmark show that SCG-MEM substantially improves performance across all categories over retrieval-based baselines.
An In-Context Schema Understanding Method for Knowledge Base Question Answering
School of Computer Science and Technology, University of Chinese Academy of Sciences(中国科学院大学计算机科学与技术学院);Institute of Computing Technology, Chinese Academy of Sciences(中国科学院计算技术研究所)
From UNDRR Reports to Event Records: Schema-Constrained LLM Extraction of Georeferenced Disasters
ETH Zürich(苏黎世联邦理工学院);University of Maryland(马里兰大学);MPI(马克斯·普朗克研究所);University of Toronto(多伦多大学);United Nations Office for Disaster Risk Reduction(联合国减少灾害风险办公室)
Improving Task Generalization via Unified Schema Prompt
Sun Yat-sen University(中山大学);Chinese University of Hong Kong(香港中文大学);Tsinghua University(清华大学);Langboat Technology(澜舟科技);Microsoft Research Asia(微软亚洲研究院)
From Chaos to Clarity: Schema-Constrained AI for Auditable Biomedical Evidence Extraction from Full-Text PDFs
Methodological Implementation Research(方法学实施研究);Ottawa Hospital Research Institute(渥太华医院研究所在);School of Engineering Design and Teaching Innovation(工程设计与教学创新学院)
Question Answering on Knowledge Bases and Text using Universal Schema and Memory Networks
University of Edinburgh(爱丁堡大学);University of Massachusetts Amherst(马萨诸塞大学阿默斯特分校);Carnegie Mellon University(卡内基梅隆大学)
现有的问答方法从知识库或原始文本中推断答案。虽然知识库(KB)方法擅长回答组合问题,但它们的性能往往受到知识库不完整性的影响。相反,Web文本包含了数百万个KB中不存在的事实,但是以非结构化的形式存在。{ \it Universal schema}可以通过将结构化知识库和非结构化文本对齐到一个公共嵌入空间来支持对它们的联合进行推理。本文将通用模式扩展到自然语言问答系统中,采用\emph{memory networks}处理文本和知识库相结合的大量事实。我们的模型可以在问答对上以端到端的方式进行训练。在\spades填空问答数据集上的评价结果表明,利用通用模式进行问答比单独使用知识库或文本要好。该模型的性能也优于当前最先进的8.5 $F_1$点。\footnote{Code and data available in \url{https://rajarshd.github.io/TextKBQA}}