Box Maze: A Process-Control Architecture for Reliable LLM Reasoning
This addresses the issue of reasoning reliability in LLMs for users in safety-critical applications, though it is incremental as it builds on existing safety approaches with a new architectural focus.
The paper tackles the problem of unreliable reasoning and hallucination in large language models under adversarial prompting by proposing the Box Maze framework, a process-control architecture that decomposes reasoning into explicit layers, resulting in a reduction of boundary failure rates from approximately 40% to below 1% in simulation-based evaluations.
Large language models (LLMs) demonstrate strong generative capabilities but remain vulnerable to hallucination and unreliable reasoning under adversarial prompting. Existing safety approaches -- such as reinforcement learning from human feedback (RLHF) and output filtering -- primarily operate at the behavioral level and may lack explicit architectural mechanisms for enforcing reasoning process integrity. This paper proposes the Box Maze framework, a conceptual process-control architecture that decomposes LLM reasoning into three explicit layers: memory grounding, structured inference, and boundary enforcement. We introduce preliminary simulation-based evaluation involving progressive boundary erosion scenarios across multiple heterogeneous LLM systems (DeepSeek-V3, Doubao, Qwen). Results from n=50 adversarial scenarios suggest that explicit cognitive control layers may improve consistency in boundary maintenance, with architectural constraints reducing boundary failure rates from approximately 40% (baseline RLHF) to below 1% under adversarial conditions. While current validation is simulation-based, these preliminary results indicate that process-level control may offer a promising direction for improving reliability in large language model reasoning.