MultiHedge: Adaptive Coordination via Retrieval-Augmented Control

TL;DR

MultiHedge is a hybrid decision-making architecture that uses retrieval-augmented LLMs to improve robustness and stability in modular decision systems, especially under changing conditions.

Contribution

It introduces a novel retrieval-augmented control architecture that enhances robustness over traditional rule-based and learning-based methods.

Findings

Memory-augmented retrieval improves robustness more than increasing model size.

MultiHedge outperforms rule-based and learning-based baselines in stability.

The study highlights the importance of memory and architecture in decision system robustness.

Abstract

Decision-making under changing conditions remains a fundamental challenge in many real-world systems. Existing approaches often fail to generalize across shifting regimes and exhibit unstable behavior under uncertainty. This raises the research question: can retrieval-augmented LLM coordination improve the robustness of modular decision pipelines? We propose MultiHedge, a hybrid architecture where an LLM produces structured allocation decisions conditioned on retrieved historical precedents, and execution is grounded in canonical option strategies. In a controlled evaluation using U.S. equities, we compare MultiHedge to rule-based and learning-based baselines. The key result is that memory-augmented retrieval confers greater robustness and stability than increasing model scale alone. Our paper contributes a controlled computational study showing that memory and architectural design play a central role in robustness in modular decision systems.