{"type":"link","version":"1.0","title":"Building a \"hard\" test subset from items the baseline got wrong guarantees an inflated gain over that baseline","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/risk-hard-subset-mined-from-baseline-failures-inflates-gains","thumbnail_url":"https://ai-arch.pages.dev/og/risk-hard-subset-mined-from-baseline-failures-inflates-gains.png","thumbnail_width":1200,"thumbnail_height":630}