Can Blindfolded LLMs Still Trade? An Anonymization-First Framework for Portfolio Optimization

TL;DR

This paper introduces BlindTrade, a framework that anonymizes market data to ensure LLM trading agents rely on genuine signals rather than memorized patterns, validated through rigorous testing and robust performance metrics.

Contribution

It presents a novel anonymization approach for LLM trading agents, addressing memorization and survivorship biases, and demonstrates its effectiveness with empirical results across different market conditions.

Findings

Achieved Sharpe ratio of 1.40 on 2025 YTD data.

Validated signal legitimacy through negative control experiments.

Policy performs better in volatile markets, less so in trending bull markets.

Abstract

For LLM trading agents to be genuinely trustworthy, they must demonstrate understanding of market dynamics rather than exploitation of memorized ticker associations. Building responsible multi-agent systems demands rigorous signal validation: proving that predictions reflect legitimate patterns, not pre-trained recall. We address two sources of spurious performance: memorization bias from ticker-specific pre-training, and survivorship bias from flawed backtesting. Our approach is to blindfold the agents--anonymizing all identifiers--and verify whether meaningful signals persist. BlindTrade anonymizes tickers and company names, and four LLM agents output scores along with reasoning. We construct a GNN graph from reasoning embeddings and trade using PPO-DSR policy. On 2025 YTD (through 2025-08-01), we achieved Sharpe 1.40 +/- 0.22 across 20 seeds and validated signal legitimacy through negative control experiments. To assess robustness beyond a single OOS window, we additionally evaluate an extended period (2024--2025), revealing market-regime dependency: the policy excels in volatile conditions but shows reduced alpha in trending bull markets.