Reproduce “ESC-Skills: Discovering and Self-Evolving Skills for Emotional Support Conversations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jie Zhu, Huaixia Dou, Shuo Jiang, Junhui Li, Lifan Guo, Feng Chen, Chi Zhang, Fang Kong
The ESC-Skills framework introduces a skill-centric approach for emotional support conversations, formalizing interventions as executable units that are discovered and refined through a multi-profile self-evolutionary mechanism based on simulated interactions. This methodology consistently improves response quality and long-horizon emotional support performance across various large language models.
Key insights
- Explicitly formalizing intervention knowledge into discrete, executable skills significantly enhances the interpretability and controllability of emotional support agents.
- Skills derived from static conversation corpora require dynamic validation, as initial skill sets can surprisingly degrade long-horizon support performance without refinement.
- A rigorous, simulation-based generation-verification loop is essential for developing robust and adaptable emotional support skills, effectively filtering ineffective interventions and ensuring consistent improvements.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.