Mixture Policy based Multi-Hop Reasoning over N-tuple Temporal Knowledge Graphs

2025-05-19

AI Summary

To address the weak interpretability and difficulty in tracing prediction justifications in n-ary temporal knowledge graph (N-TKG) reasoning, this paper proposes MT-Path, a reinforcement learning-based multi-hop path reasoning model. Methodologically, it introduces a tri-strategy action selection mechanism—integrating predicate-, core-element-, and full-fact focusing—and is the first to explicitly model auxiliary element semantics during path construction. This is achieved via an auxiliary-element-aware graph convolutional network (GCN) coupled with n-ary semantic disentanglement, enabling fine-grained, verifiable temporal path reasoning. Evaluated on multiple N-TKG benchmarks, MT-Path achieves average improvements of 3.2% in AUC and 4.7% in Hits@1 over state-of-the-art methods. Moreover, it supports human-interpretable reasoning chain generation, enhancing transparency and traceability in N-TKG inference.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Temporal Planning
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Temporal Knowledge Graphs (TKGs), which utilize quadruples in the form of (subject, predicate, object, timestamp) to describe temporal facts, have attracted extensive attention. N-tuple TKGs (N-TKGs) further extend traditional TKGs by utilizing n-tuples to incorporate auxiliary elements alongside core elements (i.e., subject, predicate, and object) of facts, so as to represent them in a more fine-grained manner. Reasoning over N-TKGs aims to predict potential future facts based on historical ones. However, existing N-TKG reasoning methods often lack explainability due to their black-box nature. Therefore, we introduce a new Reinforcement Learning-based method, named MT-Path, which leverages the temporal information to traverse historical n-tuples and construct a temporal reasoning path. Specifically, in order to integrate the information encapsulated within n-tuples, i.e., the entity-irrelevant information within the predicate, the information about core elements, and the complete information about the entire n-tuples, MT-Path utilizes a mixture policy-driven action selector, which bases on three low-level policies, namely, the predicate-focused policy, the core-element-focused policy and the whole-fact-focused policy. Further, MT-Path utilizes an auxiliary element-aware GCN to capture the rich semantic dependencies among facts, thereby enabling the agent to gain a deep understanding of each n-tuple. Experimental results demonstrate the effectiveness and the explainability of MT-Path.

Problem

• Predict future facts in N-tuple Temporal Knowledge Graphs
• Enhance explainability in N-TKG reasoning methods
• Integrate diverse information within n-tuples for reasoning

Innovation

• Reinforcement Learning-based MT-Path method
• Mixture policy-driven action selector
• Auxiliary element-aware GCN for dependencies