Qi Gan

Experimental results demonstrate that the KuaiRP models not only match the current state-of-the-art proprietary models in role-playing fidelity within the authors' target domains, but also successfully recover general agent capabilities, maintaining extremely low deployment costs.