为长时科研智能体设计分层记忆架构，突破大模型上下文瓶颈。
Episodic-Semantic Memory Architecture for Long-Horizon Scientific Agents
- 分离短期记忆与长期知识，短期固定10条消息，长期以每条3词速度增长。
- 在1.5万条消息中保持70%-85%准确率，仅用62%的令牌数，延迟1-2秒。
- 适用于需要持续推理与技术事实保留的科研场景，尤其适合多阶段实验分析。
随着大语言模型演变为持久的科学合作者，上下文窗口饱和已成为关键瓶颈。涉及迭代数据分析与假说修正的科学工作流迅速填满即使扩展后的上下文，而传统单一体系面临二次方成本增长与认知退化问题。我们评估了一种双过程记忆架构，将即时情景需求（恒定10消息窗口）与长期整合知识（约每条消息增加3个词）解耦。相较于先前的社会智能体记忆系统，该领域特定整合解决了参数演化矛盾、跨实验阶段的多跳推理及精确技术事实保留问题。通过覆盖15,000条消息的大规模评估，跨六种来自三个家族（OpenAI、Anthropic、Google）的LLM进行交叉验证，共完成1,440次查询，得出三大发现：第一，全上下文模型在10,000条消息后因上下文溢出失效，而本系统在1-2秒延迟下维持70%-85%准确率，仅消耗62%的令牌（45,434对120,000+上限）；第二，跨模型验证揭示架构级权衡独立于具体模型：双过程在数值/时间类查询上表现优异（65%-90%准确率），而RAG在历史检索上更优（60%-85%），提示互补部署策略；第三，识别出“仿真到真实”差距——合成测试中记忆恒定，但真实工作流呈现线性增长（约3词/消息），整合质量成为主要可扩展性瓶颈。该架构成功管理超过14,000条科学事实（125k令牌），证明领域特定记忆整合可实现超越全上下文限制的持续运行。
原文摘要 · Abstract (English)
As Large Language Models (LLMs) evolve into persistent scientific collaborators, context window saturation has emerged as a critical bottleneck. Scientific workflows involving iterative data analysis and hypothesis refinement rapidly saturate even extended contexts with dense technical content, while monolithic approaches suffer from quadratic cost scaling and cognitive degradation. We evaluate a Dual Process Memory Architecture that decouples immediate episodic needs (constant 10-message window) from long-term consolidated knowledge (growing at approximately 3 tokens/message). Unlike prior social agent memory systems, our domain-specific consolidation addresses contradictory parameter evolution, multi-hop reasoning across experimental phases, and precise technical fact retention. Through large-scale evaluation spanning 15,000 messages with cross-model validation across six LLMs from three families (OpenAI, Anthropic, Google), totaling 1,440 queries, we establish three key findings. First, while full-context models fail at 10,000 messages due to context overflow, our system maintains 70-85% accuracy with 1-2 second latency using 62% fewer tokens (45,434 vs 120,000+ limit). Second, cross-model validation reveals architecture-level trade-offs independent of specific LLMs: Dual Process excels at numeric/temporal queries (65-90% accuracy) while RAG excels at historical retrieval (60-85%), suggesting complementary deployment strategies. Third, we identify a "Sim-to-Real" gap where synthetic tests maintain constant memory but realistic workflows exhibit linear growth (about 3 tokens/message), with consolidation quality emerging as the primary scalability bottleneck. The architecture successfully manages profiles with 14,000+ scientific facts (125k tokens), demonstrating that domain-specific memory consolidation enables sustained operation beyond full-context limits.
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。