{"type":"link","version":"1.0","title":"Full self-attention's quadratic cost buys token-level addressability; tasks that need only segment-level addressability overpay","author_name":"AI Archs","author_url":"https://ai-arch.pages.dev","provider_name":"AI Archs","provider_url":"https://ai-arch.pages.dev","url":"https://ai-arch.pages.dev/n/attention-overhead-buys-capabilities-the-task-may-not-use","thumbnail_url":"https://ai-arch.pages.dev/og/attention-overhead-buys-capabilities-the-task-may-not-use.png","thumbnail_width":1200,"thumbnail_height":630}