Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

OrchVis: Hierarchical Multi-Agent Orchestration for Human Oversight

Jieyu Zhou

TL;DR

OrchVis tackles the challenge of supervising multi-agent collaboration without micromanagement by combining hierarchical goal alignment, task assignment, progress verification, and conflict resolution within a transparent GUI. It converts natural language intents into structured goal graphs, assigns subgoals to specialized sub-agents via an orchestration component, and uses a verifier and re-planner to handle inconsistencies and conflicts, all coordinated through a Planning Panel. The approach emphasizes separating goals from tasks, aligning early, and visualizing inter-agent relationships to reduce cognitive load and enhance oversight. This human-in-the-loop design enables supervision with autonomy, improving safety and efficiency in long-horizon, multi-agent workflows.

Abstract

We introduce OrchVis, a multi-agent orchestration framework that visualizes, verifies, and coordinates goal-driven collaboration among LLM-based agents. Through hierarchical goal alignment, task assignment, and conflict resolution, OrchVis enables humans to supervise complex multi-agent workflows without micromanaging each step. The system parses user intent into structured goals, monitors execution via automated verification, and exposes inter-agent dependencies through an interactive planning panel. When conflicts arise, users can explore system-proposed alternatives and selectively replan. OrchVis advances human-centered design for multi-agent systems by combining transparent visualization with adaptive autonomy.

OrchVis: Hierarchical Multi-Agent Orchestration for Human Oversight

TL;DR

OrchVis tackles the challenge of supervising multi-agent collaboration without micromanagement by combining hierarchical goal alignment, task assignment, progress verification, and conflict resolution within a transparent GUI. It converts natural language intents into structured goal graphs, assigns subgoals to specialized sub-agents via an orchestration component, and uses a verifier and re-planner to handle inconsistencies and conflicts, all coordinated through a Planning Panel. The approach emphasizes separating goals from tasks, aligning early, and visualizing inter-agent relationships to reduce cognitive load and enhance oversight. This human-in-the-loop design enables supervision with autonomy, improving safety and efficiency in long-horizon, multi-agent workflows.

Abstract

We introduce OrchVis, a multi-agent orchestration framework that visualizes, verifies, and coordinates goal-driven collaboration among LLM-based agents. Through hierarchical goal alignment, task assignment, and conflict resolution, OrchVis enables humans to supervise complex multi-agent workflows without micromanaging each step. The system parses user intent into structured goals, monitors execution via automated verification, and exposes inter-agent dependencies through an interactive planning panel. When conflicts arise, users can explore system-proposed alternatives and selectively replan. OrchVis advances human-centered design for multi-agent systems by combining transparent visualization with adaptive autonomy.

Paper Structure

This paper contains 11 sections, 1 equation, 2 figures.

Table of Contents

  1. Introduction
  2. Design Goals
  3. From Direct Manipulation to Orchestration.
  4. Separate Goals from Tasks and Align Early.
  5. Visualizing Inter-Agent Relationship
  6. OrchVis: System and Technical Design
  7. Goal Alignment and Parsing
  8. Task Assignment and Plan Generation
  9. Progress Tracking and Goal Verification
  10. Conflict Detection and Planning Panel Correction
  11. Conclusion and Future Work

Figures (2)

  • Figure 1: The user’s San Francisco travel goal parsed into flight, hotel, and itinerary subgoals. Lower levels can be expanded on demand.
  • Figure 2: Detected conflict between the outbound flight and the evening show. Highlighted nodes indicate conflicting goals; the text panel below lists suggested repairs.