Structure and Redundancy in Large Language Models: A Spectral Study via Random Matrix Theory

Researchers have developed a unified framework using Spectral Geometry and Random Matrix Theory to address reliability and efficiency challenges in large language models. The study introduces EigenTrack for real-time hallucination detection and RMT-KD for model compression while maintaining accuracy.

• →EigenTrack enables real-time detection of hallucinations and out-of-distribution behavior in large language models before they appear in outputs.
• →RMT-KD offers a principled approach to compress deep networks while preserving accuracy through random matrix theory-based knowledge distillation.
• →The framework uses eigenvalue dynamics of hidden activations to separate structured representations from noise-dominated variability.
• →Spectral statistics provide interpretable insights into model behavior and representation dynamics.
• →The approach addresses both reliability issues like hallucinations and efficiency demands through energy-efficient model compression.