CIR基准高估了AI模型的组合能力

一篇新的研究论文质疑了当前组合图像检索（CIR）基准的有效性，CIR是一项需要模型结合图像和文本信息 Thus, the study found that many existing CIR benchmarks can be solved using only one modality, indicating models are exploiting "unimodal shortcuts" rather than truly composing information. After auditing and validating queries, researchers re-evaluated models on a cleaner subset, revealing a greater reliance on multimodal composition but also a decrease in accuracy, suggesting current benchmarks overestimate model capabilities. AI

影响 揭示了当前多模态任务AI评估方法的局限性，可能指导未来的基准开发。

排序理由 学术论文分析特定AI任务的基准局限性。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CV 阅读 →

• Composed Image Retrieval
• Generalist Multimodal Embedding models
• Matteo Attimonelli

• 论文
• 其他

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →