AITechTalksabout 1 hour ago
LSA Cuts Long-Context AI Speed by 72x
11 min read
Researchers at Prefrontal AI and ETH Zurich developed Latent Sparse Attention to solve search bottlenecks in long-context AI. Experiments show D-LSA achieves 72.4-fold speedups on Qwen models using Nvidia H200 GPUs at 10 million tokens. The team trained components on 352 million tokens from scientific documents and code repositories to optimize retrieval accuracy for future applications.
Level
Hype check
Tap to vote and see what everyone thinks.
