Lune

EMNLP Findings2023

ASSERT: Automated Safety Scenario Red Teaming for Evaluating the Robustness of Large Language Models

Alex Mei, Sharon Levy, William Yang Wang

Abstract

As large language models are integrated into society, robustness toward a suite of prompts is increasingly important to maintain reliability in a high-variance environment. Robustness evaluations must comprehensively encapsulate the various settings in which a user may invoke an intelligent system. This paper proposes ASSERT, Automated Safety ScEnario Red Teaming, consisting of three methods -semantically aligned augmentation, target bootstrapping, and adversarial knowledge injection. For robust safety evaluation, we apply these methods in the critical domain of AI safety to algorithmically generate a test suite of prompts covering diverse robustness settings -semantic equivalence, related scenarios, and adversarial. We partition our prompts into four safety domains for a fine-grained analysis of how the domain affects model performance. Despite dedicated safeguards in existing state-of-the-art models, we find statistically significant performance differences of up to 11% in absolute classification accuracy among semantically related scenarios and error rates of up to 19% absolute error in zero-shot adversarial settings, raising concerns for users' physical safety.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

3 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • ConSiDERS-The-Human Evaluation Framework: Rethinking Human Evaluation for Generative Large Language Models cites Red Teaming Language Models with Language Models
  • Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity cites Language Models are Few-Shot Learners
  • MoE-RBench: Towards Building Reliable Language Models with Sparse Mixture-of-Experts cites Language Models are Few-Shot Learners
  • MoE-RBench: Towards Building Reliable Language Models with Sparse Mixture-of-Experts cites Training language models to follow instructions with human feedback
  • Training language models to follow instructions with human feedback cites Language Models are Few-Shot Learners
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers3

  1. ConSiDERS-The-Human Evaluation Framework: Rethinking Human Evaluation for Generative Large Language ModelsAparna Elangovan, Ling Liu, Lei Xu, Sravan Babu Bodapati et al.ACL 2024 · 19 citations
  2. MoE-RBench: Towards Building Reliable Language Models with Sparse Mixture-of-ExpertsGuanjie Chen, Xinyu Zhao, Tianlong Chen, Yu ChengICML 2024 · 8 citations
  3. Probing the Safety Robustness of LLMs in Latent SpaceTianle Gu, Kexin Huang, Zongqi Wang, Yixu Wang et al.ACL 2026

Builds on15

  1. Language Models are Few-Shot LearnersTom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah et al.NeurIPS 2020 · 64,255 citations
  2. Training language models to follow instructions with human feedbackLong Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida et al.NeurIPS 2022 · 24,707 citations
  3. Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order SensitivityYao Lu, Max Bartolo, Alastair Moore, Sebastian Riedel et al.ACL 2022 · 1,494 citations
  4. Red Teaming Language Models with Language ModelsEthan Perez, Saffron Huang, H. Francis Song, Trevor Cai et al.EMNLP 2022 · 239 citations
  5. Improving Question Answering Model Robustness with Synthetic Adversarial Data GenerationMax Bartolo, Tristan Thrush, Robin Jia, Sebastian Riedel et al.EMNLP 2021 · 68 citations

Related papers

  1. Red Teaming LLMs via Linguistic-Aware FuzzingShuai Yuan, Nian Luo, Jingling Sun, Yihao Huang et al.FSE 2026
  2. HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust RefusalMantas Mazeika, Long Phan, Xuwang Yin, Andy Zou et al.ICML 2024 · 1,031 citations
  3. Attack Prompt Generation for Red Teaming and Defending Large Language ModelsBoyi Deng, Wenjie Wang, Fuli Feng, Yang Deng et al.EMNLP Findings 2023 · 35 citations
  4. Red Teaming LLMs as Socio-Technical Practice: From Exploration and Data Creation to EvaluationAdriana Alvarado Garcia, Ruyuan Wan, Ozioma Collins Oguine, Karla Badillo-UrquiolaCHI 2026 · 1 citation
  5. ARES: Adaptive Red-Teaming and End-to-End Repair of Policy-Reward SystemJiacheng Liang, Yao Ma, Tharindu Kumarage, Satyapriya Krishna et al.ACL 2026
Dusk over the sea between two cliffs drawn in fine vertical lines