TriAdReview: Triangular Adversarial Review Architecture for Multi-Model Technical Document Generation

2026-06-12

AI Summary

This study addresses critical limitations of single large language models in technical documentation generation—namely, over-engineering, security blind spots, and insufficient coverage. To mitigate these issues, the authors propose a triadic adversarial review architecture comprising a generator and two independent reviewer models representing engineering and boundary perspectives. The framework iteratively refines outputs through adversarial critique and a triadic adjudication mechanism. The work provides the first empirical characterization of the applicability boundaries of multi-model adversarial review across diverse tasks and introduces task-adaptive prompting strategies to alleviate structural biases. Evaluated on five benchmark tasks, the approach achieves an average performance gain of 10.1%, with notable improvements in security auditing (27.6%), code generation (20.8%), and architectural design (15.6%), though a slight decline is observed in requirements analysis (−7.5%).

Abstract

Large language models (LLMs) are increasingly used for technical document generation, yet single-model outputs often suffer from over-engineering, security blind spots, and incomplete coverage. We propose TriAdReview, a triangular adversarial review architecture that employs two independent reviewer models (engineering and boundary perspectives) and a triangular judging mechanism to iteratively improve a generator model's output. We evaluate TriAdReview across five benchmark tasks - architecture design, code generation, proposal review, security audit, and requirements analysis - using three configurations: single model (baseline), dual model (single review), and triple model (full system). Results across 75 experiments (n=5 per cell) show that the triple model configuration achieves a 10.1% overall improvement over the single model baseline (26.2 vs. 23.8 out of 50; p<0.05, paired t-test), with particularly strong gains on security audit (+27.6%), code generation (+20.8%), and architecture design (+15.6%). A second scorer (mimo-v2.5-pro) confirms the direction with a smaller effect (+2.7%), suggesting moderate inter-rater agreement. However, the system shows a -7.5% degradation on requirements analysis, revealing that adversarial review architectures have a structural bias toward simplification that is counterproductive for completeness-oriented tasks. We analyze this boundary condition through a task-type framework and demonstrate that reviewer prompt adaptation partially mitigates the issue. Our findings provide the first empirical characterization of when multi-model adversarial review helps versus harms, with implications for the design of collaborative AI systems.

Problem

• technical document generation
• large language models
• security blind spots
• over-engineering
• incomplete coverage

Innovation

• triangular adversarial review
• multi-model collaboration
• technical document generation
• LLM evaluation
• task-type boundary analysis