Verification-Aware Planning for Multi-Agent Systems

2025-10-19

AI Summary

In multi-agent collaboration, execution failures frequently arise from task misinterpretation, inconsistent output formats, and handoff errors. To address these challenges, we propose VeriMAP, a verification-aware planning framework that integrates lightweight validation functions—supporting both natural language and Python—directly into task decomposition and dependency modeling. This enables semantic consistency checking at the subtask level and closed-loop iterative optimization. Our key contribution is a novel planning-time mechanism that dynamically detects and rectifies collaborative deviations without requiring external annotations, thereby enhancing reliability, interpretability, and dynamic coordination. Evaluated on multiple benchmark datasets, VeriMAP significantly outperforms both single- and multi-agent baselines, achieving absolute improvements of 12.7–23.4% in task execution success rate, while simultaneously improving robustness and debugging efficiency.

Technology Category

• Multiagent Systems: Multiagent Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Large language model (LLM) agents are increasingly deployed to tackle complex tasks, often necessitating collaboration among multiple specialized agents. However, multi-agent collaboration introduces new challenges in planning, coordination, and verification. Execution failures frequently arise not from flawed reasoning alone, but from subtle misalignments in task interpretation, output format, or inter-agent handoffs. To address these challenges, we present VeriMAP, a framework for multi-agent collaboration with verification-aware planning. The VeriMAP planner decomposes tasks, models subtask dependencies, and encodes planner-defined passing criteria as subtask verification functions (VFs) in Python and natural language. We evaluate VeriMAP on diverse datasets, demonstrating that it outperforms both single- and multi-agent baselines while enhancing system robustness and interpretability. Our analysis highlights how verification-aware planning enables reliable coordination and iterative refinement in multi-agent systems, without relying on external labels or annotations.

Problem

• Addresses verification challenges in multi-agent collaboration systems
• Resolves task misalignments and coordination failures between LLM agents
• Enhances planning robustness without external labels or annotations

Innovation

• Verification-aware planning for multi-agent collaboration
• Encodes passing criteria as verification functions
• Models subtask dependencies without external labels