Reproduce “Orchestration-Free Customer Service Automation: A Privacy-Preserving and Flowchart-Guided Framework” 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
Mengze Hong, Chen Jason Zhang, Zichang Guo, Hanlin Gu, Di Jiang, Li Qing
Researchers from Hong Kong Polytechnic University and WeBank propose an orchestration-free framework for customer service automation, leveraging Task-Oriented Flowcharts (TOFs) for structured procedural guidance. The framework includes a privacy-preserving decentralized distillation approach for training small language models (SLMs) on synthetic data, achieving high task completion rates while mitigating sensitive data exposure and memorization.
Key insights
- Task-Oriented Flowcharts (TOFs) can effectively abstract and embed complex procedural knowledge from service dialogues, providing systematic, interpretable guidance for end-to-end task coordination in automated systems.
- A privacy-preserving decentralized distillation method, leveraging abstract flowcharts, effectively addresses data scarcity and mitigates sensitive data memorization risks, enabling efficient training of lightweight SLMs for secure local deployment.
- Integrating explicit procedural knowledge via TOFs allows smaller, fine-tuned SLMs to achieve task completion performance comparable to or exceeding larger proprietary models in complex customer service automation scenarios.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.