无需数据回放，利用注意力谱正则化实现多模态大模型的持续学习新突破。
arXiv:2606.23063v1 Announce Type: cross Abstract: Multimodal large language models (MLLMs) are increasingly required to adapt to non-stationary stream…
Attention论文的发明者去哪了？探秘AI领域关键人物的财富与现状。
The all you need is attention guy. Seems to get no attention. He’s rich now right? Comments URL: https://news.ycombinator.com/item?id=48436688 Points:…
利用注意力汇合现象，在注意力层内实现原生MoE训练，有效解决头部坍塌问题。
arXiv:2602.01203v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) often assign disproportionate attention to the first token, a p…