Search papers, labs, and topics across Lattice.
AI Fundamental Research AWS Agentic AI
2
0
5
LLMs can maintain performance while skipping global attention for 80% of tokens, slashing compute costs and memory footprint in long-context scenarios.
