{"type":"link","version":"1.0","title":"Continuously retraining a subjective reward critic on verifiable-task outcomes keeps its judgments honest","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/ground-a-subjective-critic-with-signals-from-verifiable-tasks","thumbnail_url":"https://ai-arch.pages.dev/og/ground-a-subjective-critic-with-signals-from-verifiable-tasks.png","thumbnail_width":1200,"thumbnail_height":630}