MoFlow: Multi-Objective Agentic Workflow Generation

arXiv:2609.38294v1 Announce Type: new 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.

MoFlow improves workflow generation

This paper introduces an AI system designed to optimize workflows across multiple objectives like accuracy, cost, and latency. While this could lead to more efficient systems, the paper makes no specific claims about the societal implications of this capability. The focus is purely on technical advancement in AI optimization.

SOURCE     arXiv cs.AI (tier 1)
   ↓
DOCUMENT   483b7ae6-9313-4b1e-b866-1d7e6ae55845
           https://arxiv.org/abs/2609.38294
   ↓
EVIDENCE   1 extracted excerpt
   ↓
MODEL RUN  3 runs, methodology 1.0.1
   ↓
SCORE      -63  (Adverse)
   ↓
CONFIDENCE 73%