Active Attacks: Red-teaming LLMs via Adaptive Environments
This addresses the challenge of automated safety testing for LLMs, offering a significant improvement over prior methods, though it is incremental in building on RL-based approaches.
The paper tackles the problem of generating diverse attack prompts to elicit harmful behaviors from large language models (LLMs) for safety fine-tuning, and introduces Active Attacks, an RL-based red-teaming algorithm that adapts attacks as the victim evolves, achieving a cross-attack success rate improvement from 0.07% to 31.28% with a 6% increase in computation.
We address the challenge of generating diverse attack prompts for large language models (LLMs) that elicit harmful behaviors (e.g., insults, sexual content) and are used for safety fine-tuning. Rather than relying on manual prompt engineering, attacker LLMs can be trained with reinforcement learning (RL) to automatically generate such prompts using only a toxicity classifier as a reward. However, capturing a wide range of harmful behaviors is a significant challenge that requires explicit diversity objectives. Existing diversity-seeking RL methods often collapse to limited modes: once high-reward prompts are found, exploration of new regions is discouraged. Inspired by the active learning paradigm that encourages adaptive exploration, we introduce \textit{Active Attacks}, a novel RL-based red-teaming algorithm that adapts its attacks as the victim evolves. By periodically safety fine-tuning the victim LLM with collected attack prompts, rewards in exploited regions diminish, which forces the attacker to seek unexplored vulnerabilities. This process naturally induces an easy-to-hard exploration curriculum, where the attacker progresses beyond easy modes toward increasingly difficult ones. As a result, Active Attacks uncovers a wide range of local attack modes step by step, and their combination achieves wide coverage of the multi-mode distribution. Active Attacks, a simple plug-and-play module that seamlessly integrates into existing RL objectives, unexpectedly outperformed prior RL-based methods -- including GFlowNets, PPO, and REINFORCE -- by improving cross-attack success rates against GFlowNets, the previous state-of-the-art, from 0.07% to 31.28% (a relative gain greater than $400\ \times$) with only a 6% increase in computation. Our code is publicly available \href{https://github.com/dbsxodud-11/active_attacks}{here}.