GraphCert: Bootstrap Agentic Graph Reasoning with Certified Evidence Rubrics

2026-09-29

AI Summary

This study addresses the scalability limitations of graph agent training, which typically relies on expensive annotations and introduces privacy risks. To overcome these challenges, this work proposes a self-training framework grounded in executable self-verification. The approach bootstraps question-answer pairs accompanied by verifiable evidence, integrating execution-based validation with semantic filtering to establish high-quality data standards. Furthermore, it employs Group Relative Policy Optimization (GRPO) reinforcement learning to enhance the model's graph reasoning capabilities, enabling efficient training without supervision from external large language models. Experimental results demonstrate that the proposed framework significantly outperforms larger-scale baseline models across five distinct domains, exhibiting robust cross-domain transferability. The source code will be made publicly available.

Abstract

Graph agents extend large language models (LLMs) with the ability to actively explore and reason over knowledge graphs through multi-step interactions with graph tools. However, training capable graph agents typically requires large collections of question-answer pairs and reasoning trajectories, whose manual construction is costly and difficult to scale. Moreover, employing proprietary LLMs to generate such supervision further risks exposing sensitive graph data to external services. Therefore, we propose GraphCert to bootstrap agentic graph reasoning with certified evidence rubrics during post-training. Specifically, the Bootstrapped Graph Quizzer guided by generation controls produces graph-grounded QA pairs and marks supporting evidence, which undergo execution certification and semantic curation. The accepted evidence is then canonicalized into certified evidence rubrics that later reward Graph Solver evidence alignment alongside answer correctness during GRPO training. Experiments on five graph reasoning domains in GRBENCH demonstrate that GraphCert consistently outperforms substantially larger LLM agents and post-training method. Furthermore, our analysis demonstrates that the learned policy transfers robustly across heterogeneous graph domains, suggesting that GraphCert acquires reusable graph-reasoning capabilities rather than domain-specific patterns. These results establish executable self-certification as an effective approach to self-training compact graph reasoning agents. Our code will be made publicly available.

Problem

• Graph Agents
• Knowledge Graph Reasoning
• Data Scalability
• Privacy Risk
• Post-training

Innovation

• Agentic Graph Reasoning
• Certified Evidence Rubrics
• Self-Training
• GRPO
• Knowledge Graphs