How to Steer Your Multi-Agent System: Human-LLM Collaborative Planning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: He, Zeyu, Kim, Hannah, Zhang, Dan, Hruschka, Estevam
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917522726453248
author He, Zeyu
Kim, Hannah
Zhang, Dan
Hruschka, Estevam
author_facet He, Zeyu
Kim, Hannah
Zhang, Dan
Hruschka, Estevam
contents In orchestrated multi-agent systems, humans often struggle to manage plans due to their complexity and limited transparency. Existing approaches rely on outcome-level supervision, where users verify only final outputs without visibility into intermediate reasoning. We formalize a design space for human-LLM co-planning interactions along three axes: mode (semantic vs. structural), scope (global vs. targeted), and level (low vs. high-level edits). We realize it in AMBIPOM, a prototype supporting process-level supervision through both semantic and structural interactions. Through a user study, we characterize how users navigate this space, revealing hybrid workflows and effort-control-risk trade-offs; through a controlled benchmark, we analyze how LLMs revise plans under varying scope and revision strategies. Our findings yield design insights for more transparent, controllable, and effective human-AI co-planning. We release code and data at https://github.com/megagonlabs/ambipom.
format Preprint
id arxiv_https___arxiv_org_abs_2605_23023
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle How to Steer Your Multi-Agent System: Human-LLM Collaborative Planning
He, Zeyu
Kim, Hannah
Zhang, Dan
Hruschka, Estevam
Multiagent Systems
Human-Computer Interaction
In orchestrated multi-agent systems, humans often struggle to manage plans due to their complexity and limited transparency. Existing approaches rely on outcome-level supervision, where users verify only final outputs without visibility into intermediate reasoning. We formalize a design space for human-LLM co-planning interactions along three axes: mode (semantic vs. structural), scope (global vs. targeted), and level (low vs. high-level edits). We realize it in AMBIPOM, a prototype supporting process-level supervision through both semantic and structural interactions. Through a user study, we characterize how users navigate this space, revealing hybrid workflows and effort-control-risk trade-offs; through a controlled benchmark, we analyze how LLMs revise plans under varying scope and revision strategies. Our findings yield design insights for more transparent, controllable, and effective human-AI co-planning. We release code and data at https://github.com/megagonlabs/ambipom.
title How to Steer Your Multi-Agent System: Human-LLM Collaborative Planning
topic Multiagent Systems
Human-Computer Interaction
url https://arxiv.org/abs/2605.23023