WorkflowGen: An Adaptive Workflow Generation Mechanism Driven by Trajectory Experience
Summary: arXiv:2604.19756v1 Announce Type: cross
Abstract: Large language model (LLM) agents often suffer from high reasoning overhead, excessive token consumption, unstable execution, and inability to reuse past experiences in complex tasks like business queries, tool use, and workflow orchestration. Traditional methods generate workflows from scratch for every query, leading to high cost, slow response, and poor robustness.
We propose WorkflowGen, an adaptive, trajectory experience-driven framework for automatic workflow generation that reduces token usage and improves efficiency and success rate.
Key Features of WorkflowGen
- Trajectory Capture: Early in execution, WorkflowGen captures full trajectories and extracts reusable knowledge at both node and workflow levels.
- Error Fingerprints: The system identifies and utilizes error fingerprints to enhance future executions.
- Optimal Tool Mappings: It extracts optimal mappings of tools to ensure effective workflow execution.
- Parameter Schemas: WorkflowGen creates parameter schemas that streamline the workflow generation process.
- Exception-Avoidance Strategies: The framework implements strategies to prevent potential exceptions during execution.
Closed-Loop Mechanism
WorkflowGen employs a closed-loop mechanism that performs lightweight generation only on variable nodes. This approach includes:
- Trajectory Rewriting: Adjusting existing trajectories based on new inputs to create more efficient workflows.
- Experience Updating: Continuously refining the knowledge base with new experiences to improve future task handling.
- Template Induction: Generating templates that can be reused across similar tasks, enhancing efficiency.
Adaptive Routing Strategy
A three-tier adaptive routing strategy dynamically selects among:
- Direct Reuse: Leveraging previously successful workflows for similar queries.
- Rewriting-Based Generation: Modifying existing workflows based on new requirements.
- Full Initialization: Generating new workflows when no relevant previous experience exists.
Performance and Benefits
Without large annotated datasets, we qualitatively compare WorkflowGen against real-time planning, static single trajectory, and basic in-context learning baselines. Our findings indicate:
- Reduction of token consumption by over 40 percent compared to real-time planning.
- Improvement of success rate by 20 percent on medium-similarity queries through proactive error avoidance and adaptive fallback.
- Enhanced deployability via modular, traceable experiences and cross-scenario adaptability.
WorkflowGen achieves a practical balance of efficiency, robustness, and interpretability, addressing key limitations of existing approaches in automatic workflow generation.