Kardia-R1: Unleashing LLMs to Reason toward Understanding and Empathy for Emotional Support via Rubric-as-Judge Reinforcement Learning
✨ Like Kardia-R1? Give us a ⭐ Star on GitHub! Your support keeps us going! JhCircle/Kardia-R1
A user-grounded, multi-turn empathetic dialogue benchmark with structured reasoning.
KardiaBench is a large-scale, user-profile-grounded benchmark designed for… See the full description on the dataset page:
https://huggingface.co/datasets/Jhcircle/KardiaBench.