Tag
This arXiv paper evaluates how prompt perturbations affect bias and hallucination in LLMs used for decision-making tasks, finding that perturbations can mitigate these issues in some models, with Claude 3 outperforming GPT-3.5 across most datasets.
Proposes a prompt perturbation framework that generates perturbed prompt variants, filters out structurally inconsistent comparison patterns using graph-level consistency checks, then applies standard ranking methods to yield more reliable LLM rankings.
This paper investigates how toxic lexical perturbations in prompts reduce the factual accuracy and increase uncertainty of LLMs, and uses attribution-graph analyses to trace internal changes. It finds that increasing toxicity amplifies perturbation-sensitive variant nodes while core reasoning nodes remain invariant.