MoFlow: Multi-Objective Agentic Workflow Generation

2026-09-29

AI Summary

Existing agent workflow generation methods optimize only a single objective or fixed trade-offs, necessitating retraining when preferences change. This work formulates workflow generation as a multi-objective Markov decision process and proposes a convex hull Monte Carlo tree search algorithm based on optimistic set-valued backups. Unlike conventional approaches that store a single weighted score per node, this method maintains the set of achievable trade-offs, enabling a single search to cover the entire Pareto front. Consequently, it achieves zero-shot preference adaptation and returns optimal workflows on demand. Evaluated across six benchmarks, the proposed approach attains the highest average hypervolume indicator, even under unseen preferences and settings favorable to baselines.

Abstract

We study the generation of agentic workflows that jointly optimize multiple objectives, such as accuracy, cost, latency, robustness, and consistency. Existing methods for workflow generation typically optimize accuracy alone or a weighted sum of objectives, so each trained generator commits to one fixed trade-off and must be retrained from scratch when preferences change. To alleviate this, we propose MoFlow, which generates workflows optimized across varied preferences. Specifically, MoFlow formulates workflow generation as a multi-objective Markov decision process and solves it by leveraging Convex-Hull Monte Carlo Tree Search with optimistic set-valued backups, where every node stores a set of reachable trade-offs rather than one weighted score. A single search thus approximately covers the Pareto front, from which MoFlow can return a workflow for any preference by lookup without retraining. We evaluate MoFlow against six strong baselines on six benchmarks spanning mathematics, code, and question answering. Since the baselines are single-scalar optimizers by design, an apples-to-apples comparison is difficult. We instead adopt an evaluation setup that favors the baselines, in that they are rerun for each testing preference, which MoFlow never sees. Even under this stringent setup, MoFlow achieves the highest average hypervolume.

Problem

• Agentic Workflow Generation
• Multi-Objective Optimization
• Pareto Front
• Preference Adaptation

Innovation

• Multi-Objective Optimization
• Agentic Workflow
• Monte Carlo Tree Search
• Pareto Front
• Markov Decision Process