RAG Meets Temporal Graphs: Time-Sensitive Modeling and Retrieval for Evolving Knowledge

TL;DR

This work addresses the time-sensitive nature of knowledge by introducing Temporal GraphRAG (TG-RAG), a bi-level temporal-graph framework that explicitly represents timestamped facts and organizes time into a hierarchical structure. TG-RAG enables time-aware retrieval through query-centric time identification, local and global subgraph retrieval, and incremental updates that avoid full re-summarization. The authors also contribute ECT-QA, a time-sensitive QA benchmark built from earnings-call transcripts, plus an incremental evaluation protocol to assess update costs and retrieval stability. Through extensive experiments, TG-RAG outperforms strong baselines across base and updated corpora, demonstrating robust temporal reasoning, efficient updates, and strong generalization across LLMs and general QA tasks.

Abstract

Knowledge is inherently time-sensitive and continuously evolves over time. Although current Retrieval-Augmented Generation (RAG) systems enrich LLMs with external knowledge, they largely ignore this temporal nature. This raises two challenges for RAG. First, current RAG methods lack effective time-aware representations. Same facts of different time are difficult to distinguish with vector embeddings or conventional knowledge graphs. Second, most RAG evaluations assume a static corpus, leaving a blind spot regarding update costs and retrieval stability as knowledge evolves. To make RAG time-aware, we propose Temporal GraphRAG (TG-RAG), which models external corpora as a bi-level temporal graph consisting of a temporal knowledge graph with timestamped relations and a hierarchical time graph. Multi-granularity temporal summaries are generated for each time node to capture both key events and broader trends at that time. The design supports incremental updates by extracting new temporal facts from the incoming corpus and merging them into the existing graph. The temporal graph explicitly represents identical facts at different times as distinct edges to avoid ambiguity, and the time hierarchy graph allows only generating reports for new leaf time nodes and their ancestors, ensuring effective and efficient updates. During inference, TG-RAG dynamically retrieves a subgraph within the temporal and semantic scope of the query, enabling precise evidence gathering. Moreover, we introduce ECT-QA, a time-sensitive question-answering dataset featuring both specific and abstract queries, along with a comprehensive evaluation protocol designed to assess incremental update capabilities of RAG systems. Extensive experiments show that TG-RAG significantly outperforms existing baselines, demonstrating the effectiveness of our method in handling temporal knowledge and incremental updates.

Figures (3)

• Figure 1: The temporal challenges for RAG. (1) Time-aware representation: embeddings for time-sensitive facts are often indistinguishable, and knowledge graphs lack explicit temporal attributes. (2) Incremental updates: corpora evolve continuously, but most evaluations assume one-time indexing, creating an evaluation blind spot on update cost and retrieval stability.
• Figure 2: The overall framework of Temporal GraphRAG.
• Figure 3: Visualization of a subgraph sampled from our constructed bi-level temporal graph. Dashed lines represent the connections between upper-layer temporal nodes and lower-layer relational edges, where red lines indicate the temporal clues used to answer given query examples, and blue lines denote others present in the graph.