Policy Disruption in Reinforcement Learning:Adversarial Attack with Large Language Models and Critical State Identification
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912499454967808 |
|---|---|
| author | Jiang, Junyong Tian, Buwei Xu, Chenxing Li, Songze Dong, Lu |
| author_facet | Jiang, Junyong Tian, Buwei Xu, Chenxing Li, Songze Dong, Lu |
| contents | Reinforcement learning (RL) has achieved remarkable success in fields like robotics and autonomous driving, but adversarial attacks designed to mislead RL systems remain challenging. Existing approaches often rely on modifying the environment or policy, limiting their practicality. This paper proposes an adversarial attack method in which existing agents in the environment guide the target policy to output suboptimal actions without altering the environment. We propose a reward iteration optimization framework that leverages large language models (LLMs) to generate adversarial rewards explicitly tailored to the vulnerabilities of the target agent, thereby enhancing the effectiveness of inducing the target agent toward suboptimal decision-making. Additionally, a critical state identification algorithm is designed to pinpoint the target agent's most vulnerable states, where suboptimal behavior from the victim leads to significant degradation in overall performance. Experimental results in diverse environments demonstrate the superiority of our method over existing approaches. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_18113 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Policy Disruption in Reinforcement Learning:Adversarial Attack with Large Language Models and Critical State Identification Jiang, Junyong Tian, Buwei Xu, Chenxing Li, Songze Dong, Lu Machine Learning Reinforcement learning (RL) has achieved remarkable success in fields like robotics and autonomous driving, but adversarial attacks designed to mislead RL systems remain challenging. Existing approaches often rely on modifying the environment or policy, limiting their practicality. This paper proposes an adversarial attack method in which existing agents in the environment guide the target policy to output suboptimal actions without altering the environment. We propose a reward iteration optimization framework that leverages large language models (LLMs) to generate adversarial rewards explicitly tailored to the vulnerabilities of the target agent, thereby enhancing the effectiveness of inducing the target agent toward suboptimal decision-making. Additionally, a critical state identification algorithm is designed to pinpoint the target agent's most vulnerable states, where suboptimal behavior from the victim leads to significant degradation in overall performance. Experimental results in diverse environments demonstrate the superiority of our method over existing approaches. |
| title | Policy Disruption in Reinforcement Learning:Adversarial Attack with Large Language Models and Critical State Identification |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2507.18113 |