Adaptive Workflow Generation with Trajectory Experience

WorkflowGen's adaptive workflow generation mechanism is driven by trajectory experience, which captures full trajectories and extracts reusable knowledge at both node and workflow levels. This knowledge includes error fingerprints, optimal tool mappings, parameter schemas, execution paths, and exception-avoidance strategies. The framework then employs a closed-loop mechanism that performs lightweight generation only on variable nodes via trajectory rewriting, experience updating, and template induction. A three-tier adaptive routing strategy dynamically selects among direct reuse, rewriting-based generation, and full initialization based on semantic similarity to historical queries. The authors qualitatively compare WorkflowGen against real-time planning, static single trajectory, and basic in-context learning baselines, demonstrating a 40 percent reduction in token consumption and a 20 percent improvement in success rate on medium-similarity queries. WorkflowGen achieves a practical balance of efficiency, robustness, and interpretability, addressing key limitations of existing approaches.

Source: WorkflowGen:an adaptive workflow generation mechanism driven by trajectory experience