SoulChat-R1

SoulChat-R1数据集是一个基于CATCH框架构建得到的面向多轮心理咨询的思维链数据集，通过渐进式对话合成与记忆驱动的动态规划思维链得到。CATCH 是一个创新的数据合成框架，其核心在于解决AI心理咨询中的两大关键挑战：治疗保真度低与决策逻辑不透明。为此，CATCH由两个相辅相成的组件构成：渐进式对话合成 (Progressive Dialogue Synthesis, PDS)，提升对话的治疗保真度；记忆驱动的动态规划思维链 (Memory-driven Dynamic Planning Chain-of-Thought, MDP CoT)，揭示每一轮回应背后的决策动机。

The SoulChat-R1 dataset is a chain-of-thought dataset for multi-turn psychological counseling constructed based on the CATCH framework, derived via progressive dialogue synthesis and memory-driven dynamic planning chain-of-thought. CATCH is an innovative data synthesis framework whose core lies in addressing two key challenges in AI psychological counseling: low therapeutic fidelity and opaque decision-making logic. To this end, CATCH consists of two complementary components: Progressive Dialogue Synthesis (PDS), which enhances the therapeutic fidelity of conversations; and Memory-driven Dynamic Planning Chain-of-Thought (MDP CoT), which reveals the decision-making motivations behind each round of responses.