Semantic Consensus: Process-Aware Conflict Detection and Resolution for Enterprise Multi-Agent LLM Systems

arXiv cs.AI / 4/21/2026

📰 NewsDeveloper Stack & InfrastructureModels & Research

共有:

Key Points

The paper attributes high failure rates in enterprise multi-agent LLM deployments (41%–86.7%) mainly to specification and coordination problems, especially semantic inconsistencies among agents.
It introduces “Semantic Intent Divergence,” where agents form conflicting interpretations of shared goals due to siloed context and missing process models, and argues this has been an under-addressed root cause.
The Semantic Consensus Framework (SCF) is presented as process-aware middleware with six components, including a shared Process Context Layer, formal intent modeling via a Semantic Intent Graph, and a Conflict Detection Engine.
SCF uses a policy–authority–temporal consensus resolution protocol plus a Drift Monitor to catch gradual meaning divergence, and a governance integration layer to enforce organizational policies with audit trails.
In 600 runs across AutoGen, CrewAI, and LangGraph and four enterprise scenarios, SCF achieved 100% workflow completion versus 25.1% for the best baseline, while detecting 65.2% of semantic conflicts at 27.9% precision and remaining protocol-agnostic with MCP/A2A compatibility.

Abstract

Multi-agent large language model (LLM) systems are rapidly emerging as the dominant architecture for enterprise AI automation, yet production deployments exhibit failure rates between 41% and 86.7%, with nearly 79% of failures originating from specification and coordination issues rather than model capability limitations. This paper identifies Semantic Intent Divergence--the phenomenon whereby cooperating LLM agents develop inconsistent interpretations of shared objectives due to siloed context and absent process models--as a primary yet formally unaddressed root cause of multi-agent failure in enterprise settings. We propose the Semantic Consensus Framework (SCF), a process-aware middleware comprising six components: a Process Context Layer for shared operational semantics, a Semantic Intent Graph for formal intent representation, a Conflict Detection Engine for real-time identification of contradictory, contention-based, and causally invalid intent combinations, a Consensus Resolution Protocol using a policy--authority--temporal hierarchy, a Drift Monitor for detecting gradual semantic divergence, and a Process-Aware Governance Integration layer for organizational policy enforcement. Evaluation across 600 runs spanning three multi-agent frameworks (AutoGen, CrewAI, LangGraph) and four enterprise scenarios demonstrates that SCF is the only approach to achieve 100% workflow completion--compared to 25.1% for the next-best baseline--while detecting 65.2% of semantic conflicts with 27.9% precision and providing complete governance audit trails. The framework is protocol-agnostic and compatible with MCP and A2A communication standards.