关注人类分歧中的不确定性：评估VQA中模型预测与人类回答的差异
Mind the Uncertainty in Human Disagreement: Evaluating Discrepancies between Model Predictions and Human Responses in VQA
AI总结：
该研究聚焦VQA任务，基于人类分歧不确定性水平分类样本，采用新指标评估SOTA模型与人类回答分布的关联，发现普通校准会拉大差距，而人类校准可提升对齐效果，指出该对齐问题应成为未来研究重点。
摘要：
大型视觉语言模型往往难以准确预测多个人类标注者给出的回答，尤其是当这些回答体现出人类不确定性时。本研究聚焦视觉问答（Visual Question Answering，VQA）任务，全面评估当前最先进的视觉语言模型与人类回答分布的关联程度。为此，我们根据样本的人类分歧不确定性（Human Uncertainty in Disagreement，HUD）水平（低、中、高）对样本进行分类，除准确率外，还采用三种VQA领域新的人类关联指标，探究HUD的影响。为了让模型更好地与人类对齐，我们还验证了普通校准和人类校准的效果。结果显示，即便是该任务当前最优的BEiT-3模型，也难以捕捉多样化人类回答中固有的多标签分布。此外，我们发现常用的以准确率为导向的校准技术会对BEiT-3捕捉HUD的能力产生负面影响，进一步拉大模型预测与人类分布之间的差距。相比之下，我们证明了针对人类分布校准VQA模型的益处，能让模型置信度与人类不确定性更好地对齐。我们的发现表明，在VQA领域，人类回答与模型预测的一致性对齐问题尚未得到充分研究，应成为未来研究下一个关键目标。
英文摘要：
Large vision-language models frequently struggle to accurately predict responses provided by multiple human annotators, particularly when those responses exhibit human uncertainty. In this study, we focus on the Visual Question Answering (VQA) task, and we comprehensively evaluate how well the state-of-the-art vision-language models correlate with the distribution of human responses. To do so, we categorize our samples based on their levels (low, medium, high) of human uncertainty in disagreement (HUD) and employ not only accuracy but also three new human-correlated metrics in VQA, to investigate the impact of HUD. To better align models with humans, we also verify the effect of common calibration and human calibration. Our results show that even BEiT3, currently the best model for this task, struggles to capture the multi-label distribution inherent in diverse human responses. Additionally, we observe that the commonly used accuracy-oriented calibration technique adversely affects BEiT3's ability to capture HUD, further widening the gap between model predictions and human distributions. In contrast, we show the benefits of calibrating models towards human distributions for VQA, better aligning model confidence with human uncertainty. Our findings highlight that for VQA, the consistent alignment between human responses and model predictions is understudied and should become the next crucial target of future studies.
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。
