SCoOP: 多视觉-语言模型系统中用于不确定性量化的语义一致意见池化
SCoOP: Semantic Consistent Opinion Pooling for Uncertainty Quantification in Multiple Vision-Language Model Systems
AI总结：
提出SCoOP框架，通过不确定性加权的线性意见池化聚合多个视觉-语言模型的输出，实现无训练的不确定性量化，有效检测幻觉并支持高不确定性样本的弃权。
摘要：
结合多个视觉-语言模型（VLM）可以增强多模态推理和鲁棒性，但聚合异构模型的输出会放大不确定性并增加幻觉风险。我们提出SCoOP（语义一致意见池化），一种无需训练的不确定性量化（UQ）框架，通过不确定性加权的线性意见池化用于多VLM系统。核心思想是将每个VLM视为概率“专家”，采样多个输出，映射到统一空间，聚合它们的意见，并产生系统级不确定性分数。与先前为单模型设计的UQ方法不同，SCoOP明确测量跨多个VLM的集体系统级不确定性，从而实现对高不确定性样本的有效幻觉检测和弃权。在ScienceQA上，SCoOP在幻觉检测中实现了0.866的AUROC，优于基线（0.732-0.757）约10-13%。对于弃权，它达到了0.907的AURAC，超过基线（0.818-0.840）7-9%。尽管有这些提升，SCoOP相对于基线仅引入微秒级的聚合开销，与典型的VLM推理时间（秒级）相比微不足道。这些结果表明，SCoOP为不确定性感知聚合提供了一种高效且原则性的机制，推动了多模态AI系统的可靠性。我们的代码公开于https://github.com/chungenyu6/SCoOP。
英文摘要：
Combining multiple Vision-Language Models (VLMs) can enhance multimodal reasoning and robustness, but aggregating heterogeneous models' outputs amplifies uncertainty and increases the risk of hallucinations. We propose SCoOP (Semantic-Consistent Opinion Pooling), a training-free uncertainty quantification (UQ) framework for multi-VLM systems through uncertainty-weighted linear opinion pooling. The core idea is to treat each VLM as a probabilistic "expert," sample multiple outputs, map them to a unified space, aggregate their opinions, and produce a system-level uncertainty score. Unlike prior UQ methods designed for single models, SCoOP explicitly measures collective, system-level uncertainty across multiple VLMs, enabling effective hallucination detection and abstention for highly uncertain samples. On ScienceQA, SCoOP achieves an AUROC of 0.866 for hallucination detection, outperforming baselines (0.732-0.757) by approximately 10-13%. For abstention, it attains an AURAC of 0.907, exceeding baselines (0.818-0.840) by 7-9%. Despite these gains, SCoOP introduces only microsecond-level aggregation overhead relative to the baselines, which is trivial compared to typical VLM inference time (on the order of seconds). These results demonstrate that SCoOP provides an efficient and principled mechanism for uncertainty-aware aggregation, advancing the reliability of multimodal AI systems. Our code is publicly available at https://github.com/chungenyu6/SCoOP.
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。
