研究人员开发了 MSD-Score，一种无需参考字幕即可评估图像字幕的新颖方法。该方法将图像块和文本标记嵌入建模为分布，从而能够更细致地评估语义差异。MSD-Score 在与人类判断的相关性方面达到了最先进水平，并为局部定位错误提供了透明的诊断。 AI
影响 引入了一种新的无参考图像字幕评估指标，该指标与人类判断高度相关。
排序理由 该集群包含一篇详细介绍图像字幕新评估指标的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
MSD-Score 指标在无参考情况下改进了图像字幕评估
报道来源 [2]
-
MSD-Score: Multi-Scale Distributional Scoring for Reference-Free Image Caption Evaluation
arXiv:2605.06080v1 Announce Type: new Abstract: Evaluating image captions without references remains challenging because global embedding similarity often misses fine-grained mismatches such as hallucinated objects, missing attributes, or incorrect relations. We propose MSD-Score…
-
MSD-Score: Multi-Scale Distributional Scoring for Reference-Free Image Caption Evaluation
Evaluating image captions without references remains challenging because global embedding similarity often misses fine-grained mismatches such as hallucinated objects, missing attributes, or incorrect relations. We propose MSD-Score, a reference-free metric that models image patc…