ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox

arXiv:2605.10787v1 Announce Type: new Abstract: Current LLM agents are proficient at calling isolated APIs but struggle with the "last mile" of commercial software automation. In real-world scenarios, tools are not independent; they are atomic, interdependent, and prone to environmental noise. We introduce extbf{ComplexMCP}, a benchmark designed to evaluate agents in these rigorous conditions. Built on the Model Context Protocol (MCP), extbf{ComplexMCP} provides over 300 meticulously tested tools derived from 7 stateful sandboxes, ranging from office suites to financial systems. Unlike existing datasets, our benchmark utilizes a seed-driven architecture to simulate dynamic environment states and unpredictable API failures, ensuring a deterministic yet diverse evaluation. We evaluate various LLMs across full-context and RAG paradigms, revealing a stark performance gap: even top-tier models fail to exceed a 60% success rate, far trailing human performance 90%. Granular trajectory analysis identifies three fundamental bottlenecks: (1) extbf{tool retrieval saturation} as action spaces scale; (2) extbf{over-confidence}, where agents skip essential environment verifications; and (3) extbf{strategic defeatism}, a tendency to rationalize failure rather than pursuing recovery. These findings underscore the insufficiency of current agents for interdependent workflows, positioning extbf{ComplexMCP} as a critical testbed for the next generation of resilient autonomous systems.

Source: arXiv cs.AI | 2026-05-12