EDHTA: An Enhanced Dual-Helix Temple Architecture for Decoupled Long-Context Reading and High-Precision Judgment

Abstract: Long-context language models face steep degradation in signal-to-noise judgment over expanded attention horizons. EDHTA introduces a decoupled dual-helix mechanism that separates dense continuous reading from sparse high-precision judicial extraction via the Temple architecture, enabling air-gapped cybersecurity analysis across 16 GB on-premise hardware without cloud token metering.