TRAIL: Joint Inference and Refinement of Knowledge Graphs with Large Language Models

TL;DR

TRAIL introduces a unified framework that enables large language models to iteratively explore, update, and refine knowledge graphs during reasoning, significantly improving adaptability, factual accuracy, and interpretability in knowledge-intensive tasks.

Contribution

The paper presents TRAIL, a novel framework that couples joint inference with dynamic knowledge graph refinement, allowing LLMs to perform continual learning and real-time knowledge updates.

Findings

Outperforms existing KG-augmented LLMs by 3-13% on multiple benchmarks.

Enables seamless integration with various LLMs for continual adaptation.

Supports iterative exploration, validation, and pruning of knowledge during reasoning.

Abstract

Recent advances in large language models (LLMs) have unlocked powerful reasoning and decision-making capabilities. However, their inherent dependence on static parametric memory fundamentally limits their adaptability, factual accuracy, and interpretability in knowledge-intensive scenarios. Knowledge graphs (KGs), as structured repositories of explicit relational knowledge, offer a promising approach for augmenting LLMs with external, interpretable memory. Nevertheless, most existing methods that combine LLMs with KGs treat reasoning and knowledge updating as separate processes, resulting in suboptimal utilization of new information and hindering real-time updates. In this work, we propose TRAIL: a novel, unified framework for Thinking, Reasoning, And Incremental Learning that couples joint inference and dynamic KG refinement with large language models. TRAIL enables LLM agents to iteratively explore, update, and refine knowledge graphs during the reasoning process, employing a confidence-driven mechanism for the generation, validation, and pruning of new facts. This plug-and-play architecture facilitates seamless integration with various LLMs, supporting continual adaptation without the need for retraining. Extensive experiments on multiple benchmarks demonstrate that TRAIL outperforms existing KG-augmented and retrieval-augmented LLM baselines by 3% to 13%. More importantly, these results represent a significant step toward developing adaptive, memory-augmented language models capable of continual learning and reliable, transparent reasoning.