OUTFOX: LLM-Generated Essay Detection Through In-Context Learning with Adversarially Generated Examples

arXiv:2307.11729v3 Announce Type: replace
Abstract: Large Language Models (LLMs) have achieved human-level fluency in text generation, making it difficult to distinguish between human-written and LLM-generated texts. This poses a growing risk of misuse of LLMs and demands the development of detectors to identify LLM-generated texts. However, existing detectors lack robustness against attacks: they degrade detection accuracy by simply paraphrasing LLM-generated texts. Furthermore, a malicious user might attempt to deliberately evade the detectors based on detection results, but this has not …

abstract arxiv context cs.cl detection development essay examples generated human identify in-context learning language language models large language large language models llm llms making misuse risk text text generation through type