AgenticAITA: A Proof-Of-Concept About Deliberative Multi-Agent Reasoning for Autonomous Trading Systems

Paper: arXiv 2605.12532

Authors: Ivan Letteri

Abstract

Conventional algorithmic trading systems are grounded in deterministic heuristics or offline-trained statistical models that cannot adapt to the semantic complexity of rapidly shifting market regimes. This paper introduces AGENTICAITA, an agentic AI framework that replaces the traditional signal then execute paradigm with a fully autonomous deliberative loop in which multiple specialized Large Language Model agents reason, negotiate, and act in concert - without any offline training or human intervention. The framework proposes four architectural contributions: (i) an Adaptive Z-Score Trigger Engine that acts as a cognitive resource allocator, gating LLM inference exclusively on statistically anomalous market conditions; (ii) a Sequential Deliberative Pipeline - the core agentic contribution - in which an Analyst agent, a Risk Manager agent, and an Executor agent form a structured reasoning chain governed by typed JSON contracts and a deterministic hard-gate safety layer; (iii) an Inference Gating Protocol, a mutex-based cognitive resource scheduler that serializes concurrent agent activations and ensures fully reproducible audit trails; and (iv) a Correlation-Break Diversification composite score that operationalizes portfolio-level idiosyncratic signal prioritization within individual agent reasoning. Validated over a five-day autonomous dry-run session under live market conditions, the framework demonstrates operational correctness of the deliberative pipeline, achieving 157 zero-intervention invocations across 76 assets with an 11.5% agentic friction rate that confirms non-trivial inter-agent negotiation. This preliminary proof-of-concept establishes the feasibility of training-free, deterministic safety-constrained multi-agent orchestration in financial decision loops, with statistically robust performance evaluation and execution cost modeling deferred to extended live deployment.

Complexity vs Empirical Score

• Math Complexity: 3.5/10
• Empirical Rigor: 5.5/10
• Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: While the paper makes architectural contributions in systems design, the underlying mathematics are primarily standard statistical triggers (Z-scores) rather than advanced quantitative modeling. The empirical rigor is elevated by the implementation of a functional microservices architecture and a five-day live dry-run, positioning it as an engineering-heavy ‘Street Trader’ approach.

Research Flowchart

  flowchart TD
    A[<b>Goal:</b> Adaptable Multi-Agent Trading System] --> B[<b>Data Input:</b> Real-Time Market Anomalies]
    B --> C[<b>Adaptive Z-Score Trigger:</b> Cognitive Resource Allocation]
    C --> D[<b>Sequential Deliberative Pipeline:</b> Analyst, Risk, & Executor Agent Chain]
    D --> E[<b>Inference Gating:</b> Mutex Scheduling & JSON Safety Contracts]
    E --> F[<b>Findings:</b> 157 Zero-Intervention Invocations & Verified Pipeline Logic]
    
    style A fill:#f9f,stroke:#333
    style F fill:#bbf,stroke:#333

Cited References (33)

1. Can LLM s Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLM s (2024) — link
2. ReAct: Synergizing Reasoning and Acting in Language Models (2023) — link
3. Estimate and Replace: A Novel Approach to Integrating Deep Neural Networks with Existing Applications (2018) — link
4. Trading Strategy Validation Using Forwardtesting with Deep Neural Networks : (2023) — link
5. DNN-ForwardTesting: A New Trading Strategy Validation using Statistical Timeseries Analysis and Deep Neural Networks (2022) — link
6. A Stock Trading System for a Medium Volatile Asset using Multi Layer Perceptron (2022) — link
7. Stock Market Forecasting Using Machine Learning Models Through Volatility-Driven Trading Strategies (2024) — link
8. AITA: A new framework for Trading Forward Testing with an Artificial Intelligence Engine (2023) — link

0 of 33 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.