UniAdapt：一种用于知识校准的通用适配器
UniAdapt: A Universal Adapter for Knowledge Calibration
AI总结：
本文提出通用适配器 UniAdapt，通过向量辅助路由和专家混合机制进行知识校准，以解决终身模型编辑中泛化性与局部性冲突的问题，实验表明其性能优于现有方法。
摘要：
大型语言模型（LLMs）需要频繁更新，以纠正错误并及时有效地跟上持续演变的知识。近年来关于模型编辑的研究凸显了在终身模型编辑场景下平衡泛化性与局部性的挑战。我们发现，将知识直接插入模型往往会导致冲突，并可能破坏其他无关的预训练知识。为解决这一问题，我们提出了 UniAdapt，一种用于知识校准的通用适配器。受专家混合（Mixture of Experts）架构和检索增强生成（Retrieval-Augmented Generation）的启发，UniAdapt 设计了一个向量辅助路由器，负责将输入路由到合适的专家。该路由器维护一个向量存储，包含多个分片，基于语义相似度搜索结果构建路由向量。UniAdapt 完全与模型无关，并设计为可无缝即插即用集成。实验结果表明，UniAdapt 优于现有的终身模型编辑器，并在大多数指标上取得了卓越的结果。
英文摘要：
Large Language Models (LLMs) require frequent updates to correct errors and keep pace with continuously evolving knowledge in a timely and effective manner. Recent research in it model editing has highlighted the challenges in balancing generalization and locality, especially in the context of lifelong model editing. We discover that inserting knowledge directly into the model often causes conflicts and potentially disrupts other unrelated pre-trained knowledge. To address this problem, we introduce UniAdapt, a universal adapter for knowledge calibration. Inspired by the Mixture of Experts architecture and Retrieval-Augmented Generation, UniAdapt is designed with a vector-assisted router that is responsible for routing inputs to appropriate experts. The router maintains a vector store, including multiple shards, to construct routing vectors based on semantic similarity search results. UniAdapt is fully model-agnostic and designed for seamless plug-and-play integration. Experimental results show that UniAdapt outperforms existing lifelong model editors and achieves exceptional results in most metrics.
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。
