![[2602.13370] G2CP: A Graph-Grounded Communication Protocol for Verifiable and Efficient Multi-Agent Reasoning](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13370] G2CP: A Graph-Grounded Communication Protocol for Verifiable and Efficient Multi-Agent Reasoning
Summary
The paper presents G2CP, a novel communication protocol for multi-agent systems that enhances efficiency and verifiability by using graph-based messages instead of natural language.
Why It Matters
As multi-agent systems increasingly rely on Large Language Models, the risk of semantic drift and inefficiencies in communication rises. G2CP addresses these challenges by providing a structured method for agent interactions, which is crucial for applications requiring precise coordination and accountability.
Key Takeaways
- G2CP reduces inter-agent communication tokens by 73%.
- It improves task completion accuracy by 34% compared to free-text communication.
- The protocol eliminates cascading hallucinations in agent responses.
- G2CP enables fully auditable reasoning chains for enhanced transparency.
- The approach represents a shift from linguistic to structural communication in AI systems.
Computer Science > Multiagent Systems arXiv:2602.13370 (cs) [Submitted on 13 Feb 2026] Title:G2CP: A Graph-Grounded Communication Protocol for Verifiable and Efficient Multi-Agent Reasoning Authors:Karim Ben Khaled, Davy Monticolo View a PDF of the paper titled G2CP: A Graph-Grounded Communication Protocol for Verifiable and Efficient Multi-Agent Reasoning, by Karim Ben Khaled and 1 other authors View PDF HTML (experimental) Abstract:Multi-agent systems powered by Large Language Models face a critical challenge: agents communicate through natural language, leading to semantic drift, hallucination propagation, and inefficient token consumption. We propose G2CP (Graph-Grounded Communication Protocol), a structured agent communication language where messages are graph operations rather than free text. Agents exchange explicit traversal commands, subgraph fragments, and update operations over a shared knowledge graph, enabling verifiable reasoning traces and eliminating ambiguity. We validate G2CP within an industrial knowledge management system where specialized agents (Diagnostic, Procedural, Synthesis, and Ingestion) coordinate to answer complex queries. Experimental results on 500 industrial scenarios and 21 real-world maintenance cases show that G2CP reduces inter-agent communication tokens by 73%, improves task completion accuracy by 34% over free-text baselines, eliminates cascading hallucinations, and produces fully auditable reasoning chains. G2CP represents a fundame...
