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 2601.17348cs.AI
收藏

对视觉-语言模型中残障人表示的审计

Auditing Disability Representation in Vision-Language Models

Lam Research

Srikant Panda, Sourabh Singh Yadav, Palkesh Malviya

AI总结:

本文研究了视觉-语言模型在残障人表示上的表现,通过引入残障情境化提示和评估框架,发现残障情境会降低模型解释忠实度,并提出针对性提示和微调方法来改善模型表现。

摘要:

视觉-语言模型(VLMs)正在越来越多地应用于具有社会敏感性的应用,但它们对残障人士的表现仍缺乏深入研究。我们研究了以人为核心的图像中对残障人士的意识描述,其中模型往往从基于证据的事实性描述过渡到解释性转变,包括引入超出可观察视觉证据的不支持推断。为了系统分析这一现象,我们引入了一个基于配对中性提示(NP)和残障情境化提示(DP)的基准,并在零样本设置下评估了15种最先进的开源和闭源VLMs,覆盖9种残障类别。我们的评估框架将解释忠实度作为核心目标,并结合标准文本度量标准,通过情感退化、社会尊重和响应长度的转变来捕捉情感退化,同时结合LLM作为评判者的协议,通过具有残障生活经验的注释者验证。我们发现,引入残障情境一致地降低了解释忠实度,导致以推测推断、叙述扩展、情感退化和缺陷导向框架为特征的解释转变。这些影响进一步在种族和性别维度上被放大。最后,我们展示了针对提示和偏好微调能够有效提高解释忠实度并显著减少解释转变。

英文摘要:

Vision-language models (VLMs) are increasingly deployed in socially sensitive applications, yet their behavior with respect to disability remains underexplored. We study disability aware descriptions for person centric images, where models often transition from evidence grounded factual description to interpretation shift including introduction of unsupported inferences beyond observable visual evidence. To systematically analyze this phenomenon, we introduce a benchmark based on paired Neutral Prompts (NP) and Disability-Contextualised Prompts (DP) and evaluate 15 state-of-the-art open- and closed-source VLMs under a zero-shot setting across 9 disability categories. Our evaluation framework treats interpretive fidelity as core objective and combines standard text-based metrics capturing affective degradation through shifts in sentiment, social regard and response length with an LLM-as-judge protocol, validated by annotators with lived experience of disability. We find that introducing disability context consistently degrades interpretive fidelity, inducing interpretation shifts characterised by speculative inference, narrative elaboration, affective degradation and deficit oriented framing. These effects are further amplified along race and gender dimension. Finally, we demonstrate targeted prompting and preference fine-tuning effectively improves interpretive fidelity and reduces substantially interpretation shifts.

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。

↑