WISE-Flow: Workflow-Induced Structured Experience for Self-Evolving Conversational Service Agents

TL;DR

WISE-Flow introduces a workflow-centric framework that enables LLM-based agents to self-evolve by converting past interactions into reusable workflows, improving their reliability in user-facing services.

Contribution

The paper presents WISE-Flow, a novel approach that leverages historical interactions to induce workflows, enhancing agent robustness without environment-specific training.

Findings

Consistent performance improvements on ToolSandbox and τ^2-bench benchmarks.

Effective alignment of agent actions with retrieved workflows.

Enhanced feasibility reasoning for state-grounded actions.

Abstract

Large language model (LLM)-based agents are widely deployed in user-facing services but remain error-prone in new tasks, tend to repeat the same failure patterns, and show substantial run-to-run variability. Fixing failures via environment-specific training or manual patching is costly and hard to scale. To enable self-evolving agents in user-facing service environments, we propose WISE-Flow, a workflow-centric framework that converts historical service interactions into reusable procedural experience by inducing workflows with prerequisite-augmented action blocks. At deployment, WISE-Flow aligns the agent's execution trajectory to retrieved workflows and performs prerequisite-aware feasibility reasoning to achieve state-grounded next actions. Experiments on ToolSandbox and $\tau^2$-bench show consistent improvement across base models.