arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2309.11166cs.CLcs.AI
收藏

大型语言模型真的对词级扰动具有鲁棒性吗?

Are Large Language Models Really Robust to Word-Level Perturbations?

Columbia University(哥伦比亚大学);Tsinghua University(清华大学);The University of Sydney(悉尼大学);JD Explore(京东探索研究院);Tencent AI Lab(腾讯人工智能实验室)

Haoyu Wang, Guozheng Ma, Cong Yu, Ning Gui, Linrui Zhang, Zhiqi Huang, Suwei Ma, Yongzhe Chang, Sen Zhang, Li Shen, Xueqian Wang, Peilin Zhao, Dacheng Tao

AI总结:

本文提出 TREvaL,用预训练奖励模型评估 LLM 在开放式长对话中对词级扰动的鲁棒性,发现其普遍脆弱且 SFT、RLHF 后鲁棒性趋于下降。

摘要:

大型语言模型(LLM)在规模和能力上的迅速发展,使其成为适用于多种下游任务的有前景工具。除了追求更好的性能并避免对特定提示产生激烈反馈之外,为确保 LLM 的责任性,LLM 的鲁棒性也受到大量关注。然而,现有评估方法大多依赖带有预定义监督标签的传统问答数据集,这与当代 LLM 出色的生成能力并不匹配。为解决这一问题,我们提出一种新颖的理性评估方法,利用预训练奖励模型作为诊断工具,评估 LLM 针对更具挑战性的开放式问题所生成的较长对话,我们将其称为用于合理鲁棒性评估的奖励模型(Reward Model for Reasonable Robustness Evaluation,TREvaL)。较长对话体现了语言模型在理解问题熟练程度方面的综合掌握;这种能力并未被单个单词或字母完全涵盖,而单个单词或字母可能表现出过度简化和固有偏差。我们的大量实证实验表明,TREvaL 为评估 LLM 的鲁棒性提供了一种创新方法。此外,结果表明,LLM 经常容易受到日常语言使用中常见的词级扰动影响。值得注意的是,我们惊讶地发现,随着微调(SFT 和 RLHF)的进行,鲁棒性往往会下降。TREval 的代码可在 https://github.com/Harry-mic/TREvaL 获取。

英文摘要:

The swift advancement in the scales and capabilities of Large Language Models (LLMs) positions them as promising tools for a variety of downstream tasks. In addition to the pursuit of better performance and the avoidance of violent feedback on a certain prompt, to ensure the responsibility of the LLM, much attention is drawn to the robustness of LLMs. However, existing evaluation methods mostly rely on traditional question answering datasets with predefined supervised labels, which do not align with the superior generation capabilities of contemporary LLMs. To address this issue, we propose a novel rational evaluation approach that leverages pre-trained reward models as diagnostic tools to evaluate the longer conversation generated from more challenging open questions by LLMs, which we refer to as the Reward Model for Reasonable Robustness Evaluation (TREvaL). Longer conversations manifest the comprehensive grasp of language models in terms of their proficiency in understanding questions, a capability not entirely encompassed by individual words or letters, which may exhibit oversimplification and inherent biases. Our extensive empirical experiments demonstrate that TREvaL provides an innovative method for evaluating the robustness of an LLM. Furthermore, our results demonstrate that LLMs frequently exhibit vulnerability to word-level perturbations that are commonplace in daily language usage. Notably, we are surprised to discover that robustness tends to decrease as fine-tuning (SFT and RLHF) is conducted. The code of TREval is available in https://github.com/Harry-mic/TREvaL.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑