{"type":"link","version":"1.0","title":"Step-wise reasoning implements a locally greedy policy that degrades systematically on long-horizon tasks; explicit lookahead with value propagation substantially recovers performance","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/step-wise-reasoning-is-locally-greedy-planning","thumbnail_url":"https://ai-arch.pages.dev/og/step-wise-reasoning-is-locally-greedy-planning.png","thumbnail_width":1200,"thumbnail_height":630}