Preprint
Jul 2026
Diagnosis-Driven Automatic Repair for Agentic Workflow via Symbolic Inference
This work proposes FlowFixer, a diagnosis-driven automated repair framework for agentic workflows that first transforms workflow executions into unified symbolic traces and performs symbolic inference to derive executable behavioral specifications that capture node correctness, temporal dependencies, and causal relationships.
Xuyan Ma, Yawen Wang, Junjie Wang et al.
· 0 citations