ACL Findings2026
GRE Score: Generative Risk Evaluation for Large Language Models
Zaitang Li, Pin-Yu Chen, Tsung-Yi Ho
摘要
Large Language Models (LLMs) have revolutionized generative tasks, but concerns about their trustworthiness and vulnerability to adversarial attacks persist. This paper introduces the Generative Robustness Evaluation (GRE) Score, a novel metric designed to assess LLMs' resilience against adversarial red teaming attempts that may compromise model compliance and elicit undesired responses. Our approach utilizes conditional generation for synthetic text creation, offering an attackindependent evaluation of LLM robustness. By calculating the margin in refusal scores, we quantify the robustness of LLMs in an attackagnostic manner. We evaluate our method on five dimensions with specified datasets, encompassing ethical considerations, safety protocols, and potential misuse scenarios. We present four contributions: (1) The GRE Score framework, which establishes a textual robustness certificate for LLMs against adversarial red teaming attempts, providing a theoretical foundation for quantifying model resilience. (2) Comprehensive evaluations across five dimensions using eight prominent LLMs, validating GRE Scores with adversarial red teaming attacks. Our method demonstrates a consistent ranking of LLM robustness when compared to the attack-based model ranking on TrustLLM (Huang et al., 2024) while achieving a significant 5-8x speedup compared to traditional evaluation techniques. (3) Insights into the non-linear relationship between model scaling and performance, revealing that larger models do not always perform better, and an analysis of how instruction-tuning impacts robustness across LLMs. (4) The discovery that all evaluated LLMs exhibit lower performance in robustness and privacy tasks compared to other areas, highlighting a critical gap in capabilities.
问 Lune
几分钟读透这篇 Paper,不用再花一下午。
Lune 能读完全文,也能读引用它的顶会 Paper,讲清它的结论、证据和局限,每一点都注明出自哪篇 Paper。不用硬啃,也不靠猜。
脉络
它引用的 Paper18
- Denoising Diffusion Probabilistic ModelsJonathan Ho, Ajay Jain, Pieter AbbeelNeurIPS 2020 · 35,902 次引用
- Tree of Thoughts: Deliberate Problem Solving with Large Language ModelsShunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran 等NeurIPS 2023 · 5,068 次引用
- Finetuned Language Models are Zero-Shot LearnersJason Wei, Maarten Bosma, Vincent Y. Zhao, Kelvin Guu 等ICLR 2022 · 4,966 次引用
- Jailbroken: How Does LLM Safety Training Fail?Alexander Wei, Nika Haghtalab, Jacob SteinhardtNeurIPS 2023 · 2,230 次引用
- BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and ComprehensionMike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad 等ACL 2020 · 1,224 次引用
相关 Paper
- HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust RefusalMantas Mazeika, Long Phan, Xuwang Yin, Andy Zou 等ICML 2024 · 1,031 次引用
- Ferret: Faster and Effective Automated Red Teaming with Reward-Based Scoring TechniquePala Tej Deep, Vernon Toh, Rishabh Bhardwaj, Soujanya PoriaEMNLP Findings 2025
- Automated Red Teaming with GOAT: the Generative Offensive Agent TesterMaya Pavlova, Erik Brinkman, Krithika Iyer, Vítor Albiero 等ICML 2025
- Unveiling the Achilles' Heel of NLG Evaluators: A Unified Adversarial Framework Driven by Large Language ModelsYiming Chen, Chen Zhang, Danqing Luo, Luis Fernando D'Haro 等ACL Findings 2024
- TrustGen: A Platform of Dynamic Benchmarking on the Trustworthiness of Generative Foundation ModelsYue Huang, Chujie Gao, Siyuan Wu, Haoran Wang 等ICLR 2026
