arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.13481cs.CL
收藏

PersonalAI 2.0:通过规划机制增强知识图谱遍历/检索以实现个性化大语言模型代理

PersonalAI 2.0: Enhancing knowledge graph traversal/retrieval with planning mechanism for Personalized LLM Agents

Huawei, Moscow, Russia(华为,莫斯科,俄罗斯)
Mikhail Menschikov, Matvey Iskornev, Alexander Kharitonov, Alina Bogdanova, Mikhail Belkin, Ekaterina Lisitsyna, Artyom Sosedka, Victoria Dochkina, Ruslan Kosto… 展开作者

Mikhail Menschikov, Matvey Iskornev, Alexander Kharitonov, Alina Bogdanova, Mikhail Belkin, Ekaterina Lisitsyna, Artyom Sosedka, Victoria Dochkina, Ruslan Kostoev, Ilia Perepechkin, Evgeny Burnaev

AI总结:

PersonalAI 2.0通过动态多阶段查询流程提升知识图谱遍历与检索能力,结合实体提取和生成线索查询,提升回答准确性,优于现有方法,且在多个基准测试中表现优异。

摘要:

我们引入PersonalAI 2.0(PAI-2),一种新型框架,旨在通过整合外部知识图谱(KG)增强基于大语言模型(LLM)的系统。所提出的方法通过引入动态、多阶段查询处理流程,解决了现有图检索增强生成(GraphRAG)方法的关键限制。PAI-2的核心设计是其能够通过提取的实体、匹配的图顶点和生成的线索查询进行自适应、迭代信息搜索。在六个基准测试(Natural Questions、TriviaQA、HotpotQA、2WikiMultihopQA、MuSiQue和DiaASQ)上的评估表明,与类比方法(LightRAG、RAPTOR和HippoRAG 2)相比,PAI-2在事实正确性方面有所提高。PAI-2通过LLM-as-a-Judge在四个基准测试中平均获得4%的提升,反映了其在减少幻觉率和提高精度方面的有效性。我们显示,使用图遍历算法(例如BeamSearch、WaterCircles)相比标准扁平检索器平均获得6%的更好结果,而启用搜索计划增强机制相比禁用的机制在六个数据集上通过LLM-as-a-Judge获得18%的提升。此外,消融研究显示,PAI-2在MINE-1基准测试中达到SOTA结果,使用7-14B级LLM,获得89%的信息保留分数。这些发现强调了PAI-2作为下一代个性化AI应用基础模型的潜力,需要可扩展、上下文感知的知识表示和推理能力。

英文摘要:

We introduce PersonalAI 2.0 (PAI-2), a novel framework designed to enhance LLM-based systems through integration of external knowledge graphs (KGs). The proposed approach addresses key limitations of existing Graph Retrieval-Augmented Generation (GraphRAG) methods by incorporating a dynamic, multistage pipeline. The central point of the PAI-2 design is its ability to perform adaptive, iterative information search, guided by extracted entities, matched graph vertices and generated clue-queries. Conducted evaluation on five benchmarks (Natural Questions, TriviaQA, HotpotQA, 2WikiMultihopQA and MuSiQue) demonstrates an improvement in the factual correctness of generated answers compared to baseline methods (LightRAG, RAPTOR, HippoRAG 2 and PAI-1). PAI-2 achieves a 9 percentage points (pp) average gain by LLM-as-a-Judge on 2WikiMultihopQA and MuSiQue benchmarks, and attains comparable accuracy to HippoRAG 2 on the TriviaQA and HotpotQA benchmarks with Qwen2.5 7B and multilingual-e5-large reflecting its effectiveness in reducing hallucination rates and increasing recall. We show that enabled search plan enhancement mechanism gain in average a 18 pp boost compared to disabled one by LLM-as-a-Judge across five benchmarks. In addition, an ablation study reveals that PAI-2 achieves superior result on the MINE-1 benchmark, obtaining an 89\% information-retention score with Qwen2.5 7B and multilingual-e5-large. Collectively, these findings underscore the potential of PAI-2 to serve as a reusable component for personalized AI applications, which require context-aware knowledge-representation and reasoning capabilities. The source code of PAI-2 is available at the following link: https://github.com/Dzigen/PersonalAI.

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。

↑