Showing Articles About
noise
-
Jordan Watson - 22 Oct 2025
Evaluating LLM Input Comprehension and Guardrail Robustness through Noise-Based Attacks
We evaluate the ability of LLMs to understand text with random noise, and examine how prompts with varying levels of noise could bypass LLM guardrails.