iAmGiG / gex-llm-patterns Star 5 Code Issues Pull requests PhD research validating LLM understanding of market microstructure through obfuscation testing. Novel methodology proves LLMs detect dealer hedging constraints without temporal context. Paper #1: 71.5% detection, 91.2% accuracy (IEEE LLM-Finance 2025). finance machine-learning financial-markets market-microstructure options-trading pattern-detection validation-framework options-data academic-research gpt-4 llm phd-research gamma-exposure gpt-4o-mini gpt-o3 dealer-hedging obfuscation-testing Updated Oct 26, 2025 Python