5 papers
Enforcing LLM Safety through DMD-based Classification of Prompt-Response Embedding Dynamics
Mohamed Akrout, Olivera Kotevska, Dan Wilson
Large Language Models (LLMs) are increasingly deployed in high-stakes applications, yet their tendency to generate toxic, harmful, or policy-violating content poses significant ris…
Guarantees on Dynamical System Distinguishability for LLM Token Generation
Mohamed Akrout, Dan Wilson
Recent work has shown that classifying large language models (LLMs)' responses can be distinguished by modeling token embeddings as trajectories of a black-box dynamical system (DS…
Low-Cost Black-Box Detection of LLM Hallucinations via Dynamical System Prediction
Dan Wilson, Mohamed Akrout
Large Language Models (LLMs) frequently generate plausible but non-factual content, a phenomenon known as hallucination. While existing detection methods typically rely on computat…
Identification and Computation of Slow Manifolds Using the Isostable Coordinate System
Dan Wilson
Koopman analysis can be used to understand the dynamics of a nonlinear dynamical system in terms a linear, but generally infinite dimensional operator. The isostable coordinate sys…
Reduced Order Modeling of Nonlinear Dynamical Systems Using Slow Manifolds
Dan Wilson
Model order reduction in high-dimensional, nonlinear dynamical systems if often enabled through fast-slow timescale separation. One such approach involves identifying a low-dimensi…