paper-with-me

Papers

ASTPrompter: Weakly Supervised Automated Language Model Red-Teaming to Identify Low-Perplexity Toxic Prompts

2024-07-12 · Amelia F. Hardy, Houjun Liu, Bernard Lange, Duncan Eddy, Mykel J. Kochenderfer

Conventional approaches for the automated red-teaming of large language models (LLMs) aim to identify prompts that elicit toxic outputs from a frozen language model (the defender). This often results in the prompting model (the adversary) producing text that is unlikely to arise during autoregression. In response, we propose a reinforcement learning formulation of LLM red-teaming designed to discover prompts that both (1) elicit toxic outputs from a defender and (2) have low perplexity as scored by that defender. These prompts are the most pertinent in a red-teaming setting because the defender generates them with high probability. We solve this formulation with an online and weakly supervised form of Identity Preference Optimization (IPO), attacking models ranging from 137M to 7.8B parameters. Our policy performs competitively, producing prompts that induce defender toxicity at a rate of 2-23 times higher than baseline across model scales. Importantly, these prompts have lower perplexity than both automatically generated and human-written attacks. Furthermore, our method creates black-box attacks with 5.4-14 times increased toxicity. To assess the downstream utility of our method, we use rollouts from our policy as negative examples for downstream toxicity tuning and demonstrate improved safety.

📄 PDF Abstract BibTeX arXiv:2407.09447

Code (1)

sisl/astprompter 공식 구현 pytorch

Tasks

Language ModelingLanguage ModellingRed Teaming

Methods 이 논문이 사용한 방법론

Refunds@Expedia|||How do I get a full refund from Expedia? “How do I get a full refund from Expedia? How do I get a full refund from Expedia? – Call ☎️ +1-(888) 829 (0881) or +1-805-330-4056 or +1-805-330-4056 for Quick Help &…
Attention 설명 없음
Cosine Annealing Cosine Annealing is a type of learning rate schedule that has the effect of starting with a large learning rate that is relatively rapidly decreased to a minimum value before…
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Dropout Dropout is a regularization technique for neural networks that drops a unit (along with connections) at training time with a specified probability $p$ (a common value is…
Weight Decay 설명 없음
Multi-Head Attention 설명 없음
Dense Connections Dense Connections, or Fully Connected Connections, are a type of layer in a deep neural network that use a linear operation where every input is connected to every output…

Similar Papers 제목 키워드 기반

PersonaTeaming: Supporting Persona-Driven Red-Teaming for Generative AI

2026-05-07 · Wesley Hanwen Deng, Mingxi Yan, Sunnie S. Y. Kim, Akshita Jha 외 arxiv

Recent developments in AI safety research have called for red-teaming methods that effectively surface potential risks posed by generative AI models, with growing emphasis on how red-teamers' backgrounds and perspectives…

Multi-lingual Multi-turn Automated Red Teaming for LLMs

2025-04-04 · Abhishek Singhania, Christophe Dupuy, Shivam Mangale, Amani Namboori

Language Model Models (LLMs) have improved dramatically in the past few years, increasing their adoption and the scope of their capabilities over time. A significant amount of work is dedicated to ``model alignment'', i.…

Red Teaming

HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal

2024-02-06 · Mantas Mazeika, Long Phan, Xuwang Yin, Andy Zou 외

Automated red teaming holds substantial promise for uncovering and mitigating the risks associated with the malicious use of large language models (LLMs), yet the field lacks a standardized evaluation framework to rigoro…

Red Teaming

Holistic Automated Red Teaming for Large Language Models through Top-Down Test Case Generation and Multi-turn Interaction

2024-09-25 · Jinchuan Zhang, Yan Zhou, Yaxin Liu, Ziming Li 외

Automated red teaming is an effective method for identifying misaligned behaviors in large language models (LLMs). Existing approaches, however, often focus primarily on improving attack success rates while overlooking t…

DiversityRed Teaming

STAR-Teaming: A Strategy-Response Multiplex Network Approach to Automated LLM Red Teaming

2026-04-21 · MinJae Jung, YongTaek Lim, Chaeyun Kim, Junghwan Kim 외 arxiv

While Large Language Models (LLMs) are widely used, they remain susceptible to jailbreak prompts that can elicit harmful or inappropriate responses. This paper introduces STAR-Teaming, a novel black-box framework for aut…

Red Teaming