揭示不确定性：多模态大语言模型校准与性能的深度探究
Unveiling Uncertainty: A Deep Dive into Calibration and Performance of Multimodal Large Language Models
AI总结：
本文系统研究多模态大语言模型的校准问题，构建IDK数据集揭示其倾向作答而非承认不确定性的缺陷，并提出温度缩放与迭代提示优化技术以提升模型可靠性。
摘要：
多模态大语言模型（MLLMs）结合视觉与文本数据，用于图像描述和视觉问答等任务。适当的不确定性校准对于在医疗健康和自动驾驶等领域的可靠应用至关重要，但也极具挑战性。本文研究了具有代表性的多模态大语言模型，重点关注它们在各种场景下的校准情况，包括视觉微调前后，以及基础大语言模型多模态训练前后。我们观察到这些模型在性能上存在校准不良（miscalibration）现象，同时在这些场景之间校准没有显著差异。我们还强调了文本与图像之间不确定性的差异，以及它们的整合如何影响整体不确定性。为了更好地理解多模态大语言模型的校准不良及其自我评估不确定性的能力，我们构建了IDK（I don't know）数据集，该数据集对于评估它们如何处理未知问题至关重要。我们的研究结果表明，多模态大语言模型倾向于给出答案而不是承认不确定性，但这种自我评估能力会随着适当的提示调整而得到改善。最后，为了校准多模态大语言模型并增强模型可靠性，我们提出了温度缩放（temperature scaling）和迭代提示优化（iterative prompt optimization）等技术。我们的结果为改进多模态大语言模型在多模态应用中有效且负责任地部署提供了见解。代码和IDK数据集：https://github.com/hfutml/Calibration-MLLM。
英文摘要：
Multimodal large language models (MLLMs) combine visual and textual data for tasks such as image captioning and visual question answering. Proper uncertainty calibration is crucial, yet challenging, for reliable use in areas like healthcare and autonomous driving. This paper investigates representative MLLMs, focusing on their calibration across various scenarios, including before and after visual fine-tuning, as well as before and after multimodal training of the base LLMs. We observed miscalibration in their performance, and at the same time, no significant differences in calibration across these scenarios. We also highlight how uncertainty differs between text and images and how their integration affects overall uncertainty. To better understand MLLMs' miscalibration and their ability to self-assess uncertainty, we construct the IDK (I don't know) dataset, which is key to evaluating how they handle unknowns. Our findings reveal that MLLMs tend to give answers rather than admit uncertainty, but this self-assessment improves with proper prompt adjustments. Finally, to calibrate MLLMs and enhance model reliability, we propose techniques such as temperature scaling and iterative prompt optimization. Our results provide insights into improving MLLMs for effective and responsible deployment in multimodal applications. Code and IDK dataset: https://github.com/hfutml/Calibration-MLLM.
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。
