Schema-Aware Planning and Hybrid Knowledge Toolset for Reliable Knowledge Graph Triple Verification

Abstract

Knowledge Graphs (KGs) serve as a critical foundation for AI systems, yet their automated construction inevitably introduces noise, compromising data trustworthiness. Existing triple verification methods, based on graph embeddings or language models, often suffer from single-source bias by relying on either internal structural constraints or external semantic evidence, and usually follow a static inference paradigm. As a result, they struggle with complex or long-tail facts and provide limited interpretability. To address these limitations, we propose SHARP (Schema-Hybrid Agent for Reliable Prediction), a training-free autonomous agent that reformulates triple verification as a dynamic process of strategic planning, active investigation, and evidential reasoning. Specifically, SHARP combines a Memory-Augmented Mechanism with Schema-Aware Strategic Planning to improve reasoning stability, and employs an enhanced ReAct loop with a Hybrid Knowledge Toolset to dynamically integrate internal KG structure and external textual evidence for cross-verification. Experiments on FB15K-237 and Wikidata5M-Ind show that SHARP significantly outperforms existing state-of-the-art baselines, achieving accuracy gains of 4.2% and 12.9%, respectively. Moreover, SHARP provides transparent, fact-based evidence chains for each judgment, demonstrating strong interpretability and robustness for complex verification tasks.

Figures (11)

• Figure 1: Comparisons between previous methods and our proposed SHARP. Traditional graph embedding and static LLM methods are limited by single-source information, static reasoning, or limited generalization, leading to verification failures (False Positives). In contrast, SHARP successfully identifies errors by coordinating internal structure and external text within a dynamic reasoning loop.
• Figure 2: The overall architecture overview of SHARP. The framework comprises three core components: (Component 1) Schema-Aware Initialization: By utilizing a semantic encoder to retrieve analogous reasoning trajectories from a pre-constructed memory bank, it generates an Initial Plan to address the cold-start problem; (Component 2) Iterative Reasoning Loop: Guided by the initial plan, the agent enters an improved ReAct loop, dynamically adjusting its reasoning strategy based on real-time observations; (Component 3) Hybrid Knowledge Toolset: It provides a suite of complementary tools to respectively probe internal KG structures (e.g., neighbors, paths) and external semantics (e.g., Wiki, Web searches), thereby achieving cross-validation of heterogeneous knowledge.
• Figure 3: Six Core Categories of Common Triples and Their Examples.
• Figure 4: The prompt template for the Schema-Aware Planning phase.
• Figure 5: The prompt template for the Reasoning phase