The Philosopher's Stone: Trojaning Plugins of Large Language Models

arXiv:2312.00374v2 Announce Type: replace
Abstract: Open-source Large Language Models (LLMs) have recently gained popularity because of their comparable performance to proprietary LLMs. To efficiently fulfill domain-specialized tasks, open-source LLMs can be refined, without expensive accelerators, using low-rank adapters. However, it is still unknown whether low-rank adapters can be exploited to control LLMs. To address this gap, we demonstrate that an infected adapter can induce, on specific triggers, an LLM to output content defined by an adversary and to even maliciously …

accelerators adapters arxiv can control cs.cr domain exploited language language models large llms low performance plugins