发表机构
Seoul National University; Amazon AGI（首尔大学; 亚马逊AGI）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出MKP-Adapter，首个无需更新骨干的仅适配器向后兼容训练方法，通过多级保持损失和焦点重加权策略，在多模态检索任务中实现高效向后兼容。
AI 中文摘要
升级嵌入模型通常需要昂贵的数据库重新索引，因为新的查询嵌入与现有的数据库嵌入不兼容。虽然向后兼容训练（BCT）通过在训练期间强制兼容性来缓解这一问题，但现有方法通常需要更新骨干模型。这由于显著的训练成本、性能回退的风险以及对专有模型权重的有限访问而变得不切实际。我们引入了多模态知识保持适配器（MKP-Adapter），这是首个针对多模态大语言模型（MLLMs）的仅适配器BCT方法，无需更新骨干。我们确定了仅适配器BCT的主要挑战是在强制向后兼容性的同时保持新嵌入的知识。因此，我们提出了一种多级保持损失，在整个BCT过程中维持嵌入空间的几何结构。此外，集成了一种焦点重新加权策略，以优先从困难样本中学习。实验表明，我们的方法在多样的多模态基准（图像、文本、视觉文档和视频检索任务）和模型类型上实现了强大的向后兼容性。值得注意的是，MKP-Adapter仅在预提取的嵌入上训练，并且相对于原始骨干前向传播仅需可忽略的额外延迟，突显了其效率。
英文摘要
Upgrading embedding models typically requires expensive database re-indexing, as new query embeddings are incompatible with existing database embeddings. While Backward Compatible Training (BCT) mitigates this by enforcing compatibility during training, existing approaches often require updating the backbone model. This is impractical because of significant training cost, the risk of performance regression, and limited access to proprietary model weights. We introduce Multi-modal Knowledge Preserving Adapter (MKP-Adapter), the first adapter-only BCT approach for Multi-modal Large Language Models (MLLMs) that requires no backbone updates. We identified that the primary challenge in adapter-only BCT is preserving the knowledge of the new embeddings while enforcing backward compatibility. Hence, we propose a multi-level preservation loss that maintains the geometric structure of the embedding spaces throughout BCT. Furthermore, a focal re-weighting strategy is integrated to prioritize learning from challenging samples. Experiments demonstrate that our method achieves strong backward compatibility across diverse multi-modal benchmarks (image, text, visual document, and video retrieval tasks) and model types. Notably, MKP-Adapter is trained solely on pre-extracted embeddings and requires only negligible additional latency relative to the original backbone forward pass, highlighting its efficiency.
Comments15 pages, ECCV 2026 camera ready
