直接存储对话原文比提取结构化信息更有效，尤其在长对话记忆中。
Fidelity Before Structure: Verbatim Chunks Beat Lossy Artifact Extraction in Long-Conversation LLM Memory
- 用原始对话片段替代结构化信息作为记忆存储
- 在两个数据集上准确率分别提升15.9和22.0个百分点
- 适合关注长对话系统真实记忆效果的研究者
一类对话记忆系统将对话历史压缩为结构化产物（如提取的事实、决策或事件），认为结构化信息比原始文本检索效果更好。我们通过受控消融实验验证这一假设：在固定检索-重排-推理流程下，仅替换存储表示（大模型提取的结构化实体与原始对话片段），保持模型、检索器、重排器和评估器不变。结果显示，原始片段在LoCoMo上得分43.9%（对比28.0%），在LongMemEval-S上达67.4%（对比45.4%）；单跳语义图无法弥补差距，六组对照实验复现了该现象。根本原因是压缩带来的信息损失，而非结构本身：准确率与源文本保留程度正相关。所测试的抽取设计中，结构化记忆应补充而非替代原始文本：添加结构化信息可维持精度，替换则导致性能下降。代码与数据见：https://github.com/tao-hpu/cog-canvas
原文摘要 · Abstract (English)
A growing class of conversational-memory systems compresses dialogue history into structured artifacts (extracted facts, decisions, or events) on the premise that distilled structure retrieves better than raw text. We test this premise with a controlled ablation: within one fixed retrieval--rerank--reasoning pipeline, we swap only the stored representation (LLM-extracted typed artifacts versus verbatim conversation chunks), holding the model, retriever, reranker, and judge constant. Verbatim chunks win by 15.9 points on LoCoMo (43.9% vs. 28.0%) and 22.0 points on LongMemEval-S (67.4% vs. 45.4%); a 1-hop semantic graph does not recover the gap, and six confound controls reproduce the effect. The mechanism is lossy distillation, not structure per se: accuracy tracks how much source text survives in the store, and the extracted-artifact pipeline does not beat naive RAG in overall accuracy (though chunks abstain worse; see Limitations). For the extraction designs we test, structured memory should augment verbatim text rather than replace it: adding artifacts alongside chunks preserves accuracy; substituting them forfeits the gap. Code and data: https://github.com/tao-hpu/cog-canvas
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。