LLM Parkinsonism: Executive-Control Failure, Token-Inefficient Persistence, and an Uncertainty-Aware Global Executive Control Architecture for Autonomous Language-Model Agents

2026-09-24

AI Summary

This study addresses the "executive control failure" problem in large language model (LLM) agents, wherein models persist in inefficient operations after goal completion—a phenomenon termed "LLM Parkinson's." To mitigate this, we propose an Uncertainty-Aware Global Executive Control (GEC) architecture. This method decouples action generation from item-level governance by establishing independent proposal, evaluation, and stopping authority layers, integrated with multi-candidate action sets and token-efficiency optimization algorithms to achieve precise control. Experimental results demonstrate that GEC maintains high task success rates while reducing average token consumption by 36.4%. Furthermore, it effectively eliminates behavioral drift prior to task completion and significantly lowers execution complexity.

Abstract

Large language models (LLMs) can plan, use tools, write code, and execute long-horizon workflows, yet strong local competence does not guarantee project-level executive control. Agents may continue acting after the original objective is satisfied, producing low-value refinements, repeated verification, and repairs to self-created complexity. We use LLM Parkinsonism as a narrowly defined, non-clinical metaphor for this pattern of persistent action despite diminishing task-level value. We argue that the problem is not explained by autoregressive next-token prediction alone, but more directly by concentrating proposal generation, scope interpretation, progress assessment, and stopping authority within the same self-conditioned loop. We therefore introduce Global Executive Control (GEC) v0.2, an uncertainty-aware governance architecture that separates action generation from project-level control. In a 24,000-episode matched-candidate benchmark under a common 40,000-token ceiling, a first-candidate baseline achieved 67.42% hard-goal success, a candidate-set local control achieved 96.53%, and GEC achieved 96.57%. The candidate-set control shows that access to multiple candidate actions explains most of the success gain; relative to that control, GEC preserved success while reducing mean token use from 19,782 to 12,574 (36.4%) and restricted mean tokens to completion at the 40,000-token ceiling from 16,136 to 13,114 (18.7%), while eliminating measured pre-completion drift and sharply reducing gross complexity. Governance-overhead sensitivity remained favorable through an additional 500 synthetic governance tokens per cycle. These mechanistic simulations support explicit governance of scope, evidence, resource use, and stopping, while live-model validation remains necessary.

Problem

• Large Language Models
• Autonomous Agents
• Executive Control
• Token Efficiency
• Task Drift

Innovation

• Global Executive Control
• Uncertainty-Aware Architecture
• Token Efficiency
• Autonomous Agents
• LLM Parkinsonism