具有选择性记忆和回忆的动态检索增强生成系统
A Dynamic Retrieval-Augmented Generation System with Selective Memory and Remembrance
AI总结：
本文提出了一种动态检索增强生成系统ARM，通过选择性记忆和衰减机制提升检索效率与生成质量，实现高效且可调节的RAG系统性能。
摘要：
我们介绍了Adaptive RAG Memory（ARM），一种检索增强生成（RAG）框架，它用一种由选择性回忆和衰减控制的动态记忆基质取代了静态向量索引。频繁检索的项目被整合并保护不被遗忘，而很少使用的项目则逐渐衰减，灵感来自认知巩固和遗忘原理。在轻量级检索基准上，ARM仅使用约2200万参数的嵌入层，就达到了接近最先进的性能（例如，NDCG@5≈0.940，Recall@5=1.000），在超高效模型（<25M参数）中实现了最佳效率。此外，我们比较了Llama 3.1和GPT-4o的静态与动态RAG组合。Llama 3.1使用静态RAG实现了最高关键词覆盖率（67.2%）在中等延迟下，而GPT-4o使用动态选择性检索策略则实现了最快的响应时间（平均8.2秒）并具有竞争力的覆盖率（58.7%）。我们进一步展示了动态RAG实现的一项工程优化，使嵌入权重可配置、可运行时调整，并且对无效设置具有鲁棒性。ARM在不重新训练生成器的情况下实现了竞争性的准确性、自我调节的记忆增长和可解释的保留动态，并为生产和服务研究RAG系统提供了实际的质量、延迟和内存效率之间的权衡。
英文摘要：
We introduce \emph{Adaptive RAG Memory} (ARM), a retrieval-augmented generation (RAG) framework that replaces a static vector index with a \emph{dynamic} memory substrate governed by selective remembrance and decay. Frequently retrieved items are consolidated and protected from forgetting, while rarely used items gradually decay, inspired by cognitive consolidation and forgetting principles. On a lightweight retrieval benchmark, ARM reaches near state-of-the-art performance (e.g., NDCG@5 $\approx$ 0.940, Recall@5 $=1.000$) with only $\sim$22M parameters in the embedding layer, achieving the best efficiency among ultra-efficient models ($<$25M parameters). In addition, we compare static vs. dynamic RAG combinations across Llama 3.1 and GPT-4o. Llama 3.1 with static RAG achieves the highest key-term coverage (67.2\%) at moderate latency, while GPT-4o with a dynamic selective retrieval policy attains the fastest responses (8.2s on average) with competitive coverage (58.7\%). We further present an engineering optimization of the DynamicRAG implementation, making embedding weights configurable, adjustable at runtime, and robust to invalid settings. ARM yields competitive accuracy, self-regularizing memory growth, and interpretable retention dynamics without retraining the generator\color{black} and provides practical trade-off between quality, latency and memory efficiency for production and research RAG system.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
