{"type":"link","version":"1.0","title":"Reward decomposition can help group-relative training by reducing ties, not by improving judgment accuracy","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/reward-decomposition-cuts-ties-independent-of-accuracy","thumbnail_url":"https://ai-arch.pages.dev/og/reward-decomposition-cuts-ties-independent-of-accuracy.png","thumbnail_width":1200,"thumbnail_height":630}