TKG-Thinker: Towards Dynamic Reasoning over Temporal Knowledge Graphs via Agentic Reinforcement Learning

TL;DR

This paper introduces TKG-Thinker, an agent that enhances temporal knowledge graph question answering by combining autonomous planning and adaptive retrieval, leading to improved reasoning accuracy and generalization over existing static prompting methods.

Contribution

The paper proposes TKG-Thinker, a novel agent that employs dynamic interaction and reinforcement learning to improve reasoning over temporal knowledge graphs, surpassing static prompting strategies.

Findings

Achieves state-of-the-art performance on benchmark datasets.

Demonstrates strong generalization across complex TKGQA tasks.

Outperforms existing prompting strategies in temporal reasoning accuracy.

Abstract

Temporal knowledge graph question answering (TKGQA) aims to answer time-sensitive questions by leveraging temporal knowledge bases. While Large Language Models (LLMs) demonstrate significant potential in TKGQA, current prompting strategies constrain their efficacy in two primary ways. First, they are prone to reasoning hallucinations under complex temporal constraints. Second, static prompting limits model autonomy and generalization, as it lack optimization through dynamic interaction with temporal knowledge graphs (TKGs) environments. To address these limitations, we propose \textbf{TKG-Thinker}, a novel agent equipped with autonomous planning and adaptive retrieval capabilities for reasoning over TKGs. Specifically, TKG-Thinker performs in-depth temporal reasoning through dynamic multi-turn interactions with TKGs via a dual-training strategy. We first apply Supervised Fine-Tuning (SFT) with chain of thought data to instill core planning capabilities, followed by a Reinforcement Learning (RL) stage that leverages multi-dimensional rewards to refine reasoning policies under intricate temporal constraints. Experimental results on benchmark datasets with three open-source LLMs show that TKG-Thinker achieves state-of-the-art performance and exhibits strong generalization across complex TKGQA settings.