arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.20976cs.CLcs.CY
收藏

评估大型语言模型的前瞻性风险意识

Evaluating Proactive Risk Awareness of Large Language Models

The Harbin Institute of Technology, Shenzhen(哈尔滨工业大学(深圳));The Hong Kong Polytechnic University, Hong Kong(香港理工大学);Southern University of Science and Technology, Shenzhen(南方科技大学);Shenzhen Loop Area Institute, Shenzhen, China(深圳南山区研究院)

Xuan Luo, Yubin Chen, Zhiyu Hou, Linpu Yu, Geng Tu, Jing Li, Ruifeng Xu

AI总结:

本文提出前瞻性风险意识评估框架,通过Butterfly数据集评估LLMs在生态领域预见潜在危害的能力,发现响应长度限制和多模态保护盲点等问题,强调部署LLM时需加强主动防护。

摘要:

随着大型语言模型(LLMs)越来越多地融入日常决策中,其安全责任已不仅限于对明确有害意图的反应,还需预见潜在但后果严重的风险。在本工作中,我们引入了一个前瞻性风险意识评估框架,以衡量LLMs是否能够预见可能的危害并在损害发生前提供警告。我们构建了Butterfly数据集,以在环境和生态领域实例化此框架。它包含1,094个查询,模拟了普通问题解决活动,其响应可能导致潜在的生态影响。通过在五个广泛使用的LLMs上进行实验,我们分析了响应长度、语言和模态的影响。实验结果揭示了在响应长度受限的情况下,前瞻性意识出现一致显著下降,跨语言相似性和在(多模态)物种保护中的持续盲点。这些发现突显了当前安全对齐与现实世界生态责任要求之间的关键差距,强调了在LLM部署中需要主动防护措施的必要性。

英文摘要:

As large language models (LLMs) are increasingly embedded in everyday decision-making, their safety responsibilities extend beyond reacting to explicit harmful intent toward anticipating unintended but consequential risks. In this work, we introduce a proactive risk awareness evaluation framework that measures whether LLMs can anticipate potential harms and provide warnings before damage occurs. We construct the Butterfly dataset to instantiate this framework in the environmental and ecological domain. It contains 1,094 queries that simulate ordinary solution-seeking activities whose responses may induce latent ecological impact. Through experiments across five widely used LLMs, we analyze the effects of response length, languages, and modality. Experimental results reveal consistent, significant declines in proactive awareness under length-restricted responses, cross-lingual similarities, and persistent blind spots in (multimodal) species protection. These findings highlight a critical gap between current safety alignment and the requirements of real-world ecological responsibility, underscoring the need for proactive safeguards in LLM deployment.

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。

↑