A paper titled "Judging the Justification: Decision Certificate Supervision for Generative Reward Models" proposes supervising the natural-language rationales that generative reward models (GenRMs) produce to justify preference judgments, arguing that a correct verdict does not guarantee a sound justification. The authors note that existing approaches rely on reference explanations to supervise rationales, but obtaining such references can be costly. The work targets the gap between verdict accuracy and justification quality in GenRMs used for preference modeling.
Generative reward models (GenRMs) generate natural language rationales to justify preference judgments, yet a correct verdict does not guarantee a sound justification. Existing approaches often rely on reference explanations to supervise rationales, but obtaining such references can be costly and do
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.