{"type":"link","version":"1.0","title":"Temperature-0 run-to-run divergence depends on the quantization stack, so the replay noise floor must be measured per backend","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/the-deterministic-noise-floor-is-serving-stack-dependent","thumbnail_url":"https://ai-arch.pages.dev/og/the-deterministic-noise-floor-is-serving-stack-dependent.png","thumbnail_width":1200,"thumbnail_height":630}