研究人员开发了一种用于音视频场景下类别增量学习（CIL）的新方法，解决了在获取新知识的同时不丢失先前学习信息这一挑战。该方法通过一种新颖的注意力策略，利用SAM-Audio多模态模型的音频特征来指导视觉表示。为了进一步对抗灾难性遗忘，该方法在特征和logit层面都纳入了双层蒸馏目标，在音视频CIL基准测试中表现优于现有最先进技术。 AI
影响 引入了一种新颖的音视频类别增量学习方法，有望提高多模态AI系统的持续学习能力。
排序理由 该集群包含一篇详细介绍特定机器学习任务新方法的学术论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
Class-Incremental Learning (CIL) aims to continuously learn new classes without forgetting previously acquired knowledge. While recent CIL advances have spurred significant interest across various modalities, the audio-visual setting remains underexplored. Furthermore, although f…
arXiv:2606.10887v1 Announce Type: new Abstract: Class-Incremental Learning (CIL) aims to continuously learn new classes without forgetting previously acquired knowledge. While recent CIL advances have spurred significant interest across various modalities, the audio-visual settin…
Class-Incremental Learning (CIL) aims to continuously learn new classes without forgetting previously acquired knowledge. While recent CIL advances have spurred significant interest across various modalities, the audio-visual setting remains underexplored. Furthermore, although f…