针对动态生物医学知识图谱，提出多模态持续学习框架CMKL，显著提升模型记忆与泛化能力。
CMKL: Modality-Aware Continual Learning for Evolving Biomedical Knowledge Graphs
- 融合结构、文本、分子三类模态，用专家混合路由实现自适应特征融合
- 在129K实体任务上实现0.591平均精度，比最强基线高60%，遗忘率仅0.008
- 首次揭示模态间梯度不可达性，通过抑制不可靠模态保持表示稳定性
生物医学知识图谱日益庞大、动态且多模态，受高通量测序等生物技术推动。现有知识图嵌入方法及其持续学习扩展或假设图结构静态，或无法利用演化数据分布下的多模态信息，且对所有参数施加统一正则化，忽略不同模态在演化中具有不同的遗忘动态。本文提出持续多模态知识图学习框架CMKL，原生编码结构、文本和分子信息，通过混合专家（MoE）路由融合，并使用标准EWC正则与基于K均值的多模态回放缓冲区保护先前知识。在包含10个任务、12.9万实体的生物医学持续学习基准上评估，CMKL在持续实体分类任务中达到0.591平均精度，相较最强结构基线（0.370）提升60%，且接近零遗忘（0.008）。在持续关系预测中，达到0.062平均精度，优于联合训练（0.047，p=0.045）与LKGE（0.039），与朴素序列训练（0.058）相当。冻结文本的消融实验达0.136，超过任何联合训练模型两倍以上，但该信号无法通过边际排序梯度获取：模态间的梯度不对称存在于表示层面，而非融合层面，而MoE路由通过抑制不可达模态避免其进入学习瓶颈，从而有效管理此问题。
原文摘要 · Abstract (English)
Biomedical knowledge graphs are increasingly large, dynamic, and multimodal, driven by rapid advances in biotechnology such as high-throughput sequencing. Machine learning models can infer previously unobserved biomedical relationships and characterize biomedical entities in these graphs, but existing knowledge graph embedding methods and their continual learning extensions either assume static graph structure or fail to exploit multimodal information under evolving data distributions. They also apply uniform regularization across all model parameters, ignoring that different modalities may exhibit distinct forgetting dynamics as the graph evolves. We propose the Continual Multimodal Knowledge Graph Learner (CMKL), a CL framework for biomedical KGs that natively encodes structure, text, and molecules, fuses them through a Mixture-of-Experts (MoE) router, and protects previously learned knowledge with standard EWC regularization and a K-means-diverse multimodal replay buffer. We evaluate CMKL on a 129K-entity biomedical continual benchmark with 10 tasks. On continual biomedical entity classification, CMKL reaches AP 0.591 versus 0.370 for the strongest structural baseline, a 60% gain that is driven by access to multimodal features and preserved across the sequence with near-zero forgetting (AF 0.008). On continual relationship prediction, CMKL reaches AP $0.062$, matching Naive Sequential and EWC (0.058) within seed noise and outperforming Joint Training (0.047, p=0.045) and LKGE (0.039). A frozen-text ablation reaches AP 0.136, more than double any jointly trained model, yet that signal is unreachable by margin-ranking gradients: the greedy-modality asymmetry lives at the representation level, not the fusion level, and MoE routing manages it by suppressing the unreachable modality without forcing it through a learned bottleneck. Code: github.com/yradwan147/cmkl-neurips2026
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。