TRAIL: Joint Inference and Refinement of Knowledge Graphs with Large Language Models
Xinkui Zhao, Haode Li, Yifan Zhang, Guanjie Cheng, Yueshen Xu · 2025
Read TRAIL: Joint Inference and Refinement of Knowledge Graphs with Large Language Models by Xinkui Zhao, Haode Li, Yifan Zhang, Guanjie Cheng, Yueshen Xu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-04474.txt.
Description
Xinkui Zhao, Haode Li, Yifan Zhang, Guanjie Cheng and Yueshen Xu present TRAIL, a framework that lets language models reason over knowledge graphs while also improving them, in their 2025 arXiv paper. Language models rely on static built-in memory, which limits accuracy and freshness. Knowledge graphs offer explicit, checkable facts, but most systems keep reasoning and updating apart.
TRAIL couples the two: the model explores the graph during reasoning, then generates, validates and prunes new facts under a confidence-driven rule. The plug-and-play design works with various models and needs no retraining. Across benchmarks it beat graph-augmented and retrieval-augmented rivals by several points, a step toward continual learning.
Catalog record
| CREATOR | Xinkui Zhao, Haode Li, Yifan Zhang, Guanjie Cheng, Yueshen Xu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-04474