Paper: arXiv 2604.03888

Authors: Rajat M. Barot, Arjun S. Borkhatariya

Abstract

This paper presents PolySwarm, a novel multi-agent large language model (LLM) framework designed for real-time prediction market trading and latency arbitrage on decentralized platforms such as Polymarket. PolySwarm deploys a swarm of 50 diverse LLM personas that concurrently evaluate binary outcome markets, aggregating individual probability estimates through confidence-weighted Bayesian combination of swarm consensus with market-implied probabilities, and applying quarter-Kelly position sizing for risk-controlled execution. The system incorporates an information-theoretic market analysis engine using Kullback-Leibler (KL) divergence and Jensen-Shannon (JS) divergence to detect cross-market inefficiencies and negation pair mispricings. A latency arbitrage module exploits stale Polymarket prices by deriving CEX-implied probabilities from a log-normal pricing model and executing trades within the human reaction-time window. We provide a full architectural description, implementation details, and evaluation methodology using Brier scores, calibration analysis, and log-loss metrics benchmarked against human superforecaster performance. We further discuss open challenges including hallucination in agent pools, computational cost at scale, regulatory exposure, and feedback-loop risk, and outline five priority directions for future research. Experimental results demonstrate that swarm aggregation consistently outperforms single-model baselines in probability calibration on Polymarket prediction tasks.

Complexity vs Empirical Score

  • Math Complexity: 5.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces an innovative multi-agent LLM framework applied to prediction markets, combining information-theoretic metrics with standard quantitative finance concepts like Kelly sizing and log-normal CEX mapping. While the mathematical formulations are applied rather than theoretical proofs, the integration of swarm intelligence with real-time latency arbitrage offers strong novelty and a structured evaluation methodology.

Research Flowchart

  flowchart TD
    A[Research Goal: Develop Multi-Agent LLM Framework for Prediction Market Trading & Latency Arbitrage] --> B[Data Inputs: Polymarket Order Books, News Feeds & CEX Real-Time Crypto Prices]
    
    B --> C[PolySwarm Multi-Agent Engine: 50 LLM Personas Concurrently Evaluate Markets]
    
    C --> D[Market Analysis & Arbitrage: KL/JS Divergence Mispricing & CEX-Implied Log-Normal Pricing]
    
    D --> E[Execution Strategy: Confidence-Weighted Bayesian Aggregation & Quarter-Kelly Sizing]
    
    E --> F[Outcomes: Outperforms Single LLMs in Brier Scores, Achieves Superior Calibration & Captures Stale Price Arbitrage]