Pay for the Fault, Not the Flow: Label-Free In-Flow Multi-Agent Workflow Optimization

Hugging Face Daily Papers Papers
multi-agent llm workflow-optimization label-free benchmark agent-assignment

Summary

The paper proposes InFlowOp, a label-free method for constructing and optimizing multi-agent LLM workflows by pricing each agent assignment with a unified cost that balances agent competence against execution cost, and introduces Braid, a benchmark for evaluating workflow-level multi-agent coordination. Across domains and backbones, InFlowOp outperforms single-agent baselines by up to 11.97% and achieves 9.64% gains in in-flow optimization.

Large language models (LLMs) increasingly construct multi-agent workflows that decompose a complex task and assign specialist agents from a pool. However, building such a workflow well remains challenging: how finely to divide the task, which agent to trust with each subtask, and when to create a new specialist are all critical decisions a workflow constructor needs to settle up front. Thus, whether each subtask succeeds remains unknown until the workflow runs. Yet, improving a workflow is costly. Locating a fault usually requires a reference answer, a graded outcome, or a trained assessor, and the fix is applied to the whole workflow through re-execution, re-search, or retraining. We propose InFlowOp, which prices every decision in one label-free cost that weighs how well an agent's competence meets what a subtask demands against how much that agent takes to run. Before execution, InFlowOp bidirectionally determines the granularity of task decomposition and agent assignment following from the cost rather than from a fixed template. During execution, InFlowOp corrects a fault with the cheapest move via the same cost that serves the workflow both as it is built and as it runs. Facing the workflow-level evaluation challenge, we introduce Braid, a benchmark whose tasks require multi-agent coordination beyond single-agent capability. Across various domains and backbones, InFlowOp outperforms single agent baselines by up to +11.97%, achieving +9.64% with in-flow optimization. Our project page: https://xhguo7.github.io/InFlowOp/.
Original Article
View Cached Full Text

Cached at: 10/02/26, 04:26 AM

Paper page - Pay for the Fault, Not the Flow: Label-Free In-Flow Multi-Agent Workflow Optimization

Source: https://huggingface.co/papers/2610.01017

Abstract

Largelanguagemodels(LLMs)increasinglyconstructmulti-agentworkflowsthatdecomposeacomplextaskandassignspecialistagentsfromapool.However,buildingsuchaworkflowwellremainschallenging:howfinelytodividethetask,whichagenttotrustwitheachsubtask,andwhentocreateanewspecialistareallcriticaldecisionsaworkflowconstructorneedstosettleupfront.Thus,whethereachsubtasksucceedsremainsunknownuntiltheworkflowruns.Yet,improvingaworkflowiscostly.Locatingafaultusuallyrequiresareferenceanswer,agradedoutcome,oratrainedassessor,andthefixisappliedtothewholeworkflowthroughre-execution,re-search,orretraining.WeproposeInFlowOp,whichpriceseverydecisioninonelabel-freecostthatweighshowwellanagent’scompetencemeetswhatasubtaskdemandsagainsthowmuchthatagenttakestorun.Beforeexecution,InFlowOpbidirectionallydeterminesthegranularityoftaskdecompositionandagentassignmentfollowingfromthecostratherthanfromafixedtemplate.Duringexecution,InFlowOpcorrectsafaultwiththecheapestmoveviathesamecostthatservestheworkflowbothasitisbuiltandasitruns.Facingtheworkflow-levelevaluationchallenge,weintroduceBraid,abenchmarkwhosetasksrequiremulti-agentcoordinationbeyondsingle-agentcapability.Acrossvariousdomainsandbackbones,InFlowOpoutperformssingleagentbaselinesbyupto+11.97%,achieving+9.64%within-flowoptimization.Ourprojectpage:https://xhguo7.github.io/InFlowOp/.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2610\.01017

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.01017 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.01017 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.01017 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

MoFlow: Multi-Objective Agentic Workflow Generation

arXiv cs.AI

MoFlow formulates agentic workflow generation as a multi-objective MDP and solves it with Convex-Hull Monte Carlo Tree Search with set-valued backups, covering the Pareto front so workflows can be retrieved for any preference without retraining. It outperforms six baselines on math, code, and QA benchmarks in terms of hypervolume.

DynaFlow: Learning Free-Form Multi-Agent Orchestration

Papers with Code Trending

DynaFlow introduces a framework that lets a main agent synthesize executable workflows to coordinate sub-agents adaptively, achieving 88.08% accuracy and a 10.90-point gain over baselines with a Qwen3.5-27B backbone across six benchmarks.