Explore-on-Graph: Incentivizing Autonomous Exploration of Large Language Models on Knowledge Graphs with Path-refined Reward Modeling
Shiqi Yan 1,2 , Yubo Chen 1 , Ruiqi Zhou 2 , Zhengxi Yao 2 , Shuai Chen 4 , Tianyi Zhang 4 , Shijie Zhang 4 , WeiQiang Zhang 1,2 , Yongfeng Huang 1,2 , Haixin Duan 1,3 , Yunqi Zhang 1 1 Zhongguancun Laboratory, Beijing, China 2 Department of Electronic Engineering, Tsinghua University, Beijing, Chin · 2026
Read Explore-on-Graph: Incentivizing Autonomous Exploration of Large Language Models on Knowledge Graphs with Path-refined Reward Modeling by Shiqi Yan 1,2 , Yubo Chen 1 , Ruiqi Zhou 2 , Zhengxi Yao 2 , Shuai Chen 4 , Tianyi Zhang 4 , Shijie Zhang 4 , WeiQiang Zhang 1,2 , Yongfeng Huang 1,2 , Haixin Duan 1,3 , Yunqi Zhang 1 1 Zhongguancun Laboratory, Beijing, China 2 Department of Electronic Engineering, Tsinghua University, Beijing, Chin (2026) 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-2602-21728.txt.
Description
Language models often hallucinate or drop facts when answering questions, and grounding answers in knowledge graphs offers a check. Existing graph methods either enforce rules during generation or imitate paths from fixed demonstrations, confining reasoning to prior experience or fine-tuning data and limiting generalization to new graph problems. Shiqi Yan and ten co-authors propose Explore-on-Graph, which rewards models for exploring diverse reasoning paths on graphs.
Reinforcement learning during training rewards correct final answers, while path information adds extra reward signals that refine exploration and cut wasted effort. On five knowledge-graph question-answering benchmarks, the method led the field, beating both open-source and closed-source rivals.
Catalog record
| CREATOR | Shiqi Yan 1,2 , Yubo Chen 1 , Ruiqi Zhou 2 , Zhengxi Yao 2 , Shuai Chen 4 , Tianyi Zhang 4 , Shijie Zhang 4 , WeiQiang Zhang 1,2 , Yongfeng Huang 1,2 , Haixin Duan 1,3 , Yunqi Zhang 1 1 Zhongguancun Laboratory, Beijing, China 2 Department of Electronic Engineering, Tsinghua University, Beijing, Chin |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-21728