Forget Attention: Importance-Aware Attention Is All You Need 事件

Forget Attention: Importance-Aware Attention Is All You Need arXiv:2606.02332v1 Announce Type: cross Abstract: Combining attention's global retrieval with the sequential importance signal of state space models (SSMs) is the open challenge of hybrid language modeling. Transformers see everywhere but cannot prioritize; SSMs know what matters but cannot revisit. Existing hybrids -- Jamba (block level) and Hymba (head level) -- place the two in separate compartments, so neither informs the other du