When Reject Turns into Accept: Quantifying the Vulnerability of LLM-Based Scientific Reviewers to Indirect Prompt Injection

This article, sourced from ArXiv, focuses on the vulnerability of Large Language Model (LLM)-based scientific reviewers to indirect prompt injection. It likely explores how malicious prompts can manipulate these LLMs to accept or endorse content they would normally reject. The quantification aspect suggests a rigorous, data-driven approach to understanding the extent of this vulnerability.