arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2502.00577cs.AIcs.CLcs.LG
收藏

理解分布偏移下的多模态大语言模型:一种信息论方法

Understanding Multimodal LLMs Under Distribution Shifts: An Information-Theoretic Approach

Changdae Oh;Zhen Fang;Shawn Im;Xuefeng Du;Yixuan Li

Changdae Oh, Zhen Fang, Shawn Im, Xuefeng Du, Yixuan Li

AI总结:

本文提出首个信息论框架,通过有效互信息量化多模态大语言模型在分布偏移下的最大风险,并在61种偏移场景中实证验证了理论结果。

摘要:

多模态大语言模型(MLLMs)已展现出令人瞩目的能力,但在分布偏移下表现不佳,此时评估数据与指令微调分布存在差异。尽管已有工作提供了经验性评估,我们认为建立一种能够刻画并量化MLLMs风险的正式框架,对于确保MLLMs在现实世界中安全可靠的应用是必要的。通过采用信息论视角,我们提出了首个能够量化MLLMs在分布偏移下最大风险的理论框架。该框架的核心是引入有效互信息(EMI),这是一种量化输入查询与模型响应之间相关性的原则性度量。我们推导了分布内(ID)与分布外(OOD)数据之间EMI差异的上界,并将其与视觉和文本分布差异联系起来。在涵盖61种偏移场景的真实基准数据集上进行的大量实验,实证验证了我们的理论洞见。

英文摘要:

Multimodal large language models (MLLMs) have shown promising capabilities but struggle under distribution shifts, where evaluation data differ from instruction tuning distributions. Although previous works have provided empirical evaluations, we argue that establishing a formal framework that can characterize and quantify the risk of MLLMs is necessary to ensure the safe and reliable application of MLLMs in the real world. By taking an information-theoretic perspective, we propose the first theoretical framework that enables the quantification of the maximum risk of MLLMs under distribution shifts. Central to our framework is the introduction of Effective Mutual Information (EMI), a principled metric that quantifies the relevance between input queries and model responses. We derive an upper bound for the EMI difference between in-distribution (ID) and out-of-distribution (OOD) data, connecting it to visual and textual distributional discrepancies. Extensive experiments on real benchmark datasets, spanning 61 shift scenarios, empirically validate our theoretical insights.

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。

↑