Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HALO: Hierarchical Autonomous Logic-Oriented Orchestration for Multi-Agent LLM Systems

Zhipeng Hou, Junyi Tang, Yipeng Wang

TL;DR

HALO introduces a Hierarchical Autonomous Logic-Oriented Orchestration framework for multi-agent LLM systems, replacing fixed agent-role spaces and static workflows with a three-stage pipeline: Adaptive Prompt Refinement, Hierarchical Reasoning Stack, and a Workflow Search Engine guided by Monte Carlo Tree Search. The top-level planner decomposes tasks, mid-level role-design agents instantiate specialized sub-agents, and low-level inference agents execute subtasks, with MCTS constructing and evaluating reasoning trajectories. Empirical results on HumanEval, MMLU, and MATH show HALO achieves a 14.4% average improvement over SOTA baselines, with up to 19.6% gains on specific expert-level subareas and up to 13.3% on Moral Scenarios, illustrating strong capabilities in complex tasks. The work also demonstrates the value of adaptive prompting and structured workflow search for scalable, adaptable MAS in LLM-based systems, with code available at the project repository.

Abstract

Recent advancements in Multi-Agent Systems (MAS) powered by Large Language Models (LLMs) have demonstrated tremendous potential in diverse task scenarios. Nonetheless, existing agentic systems typically rely on predefined agent-role design spaces and static communication structures, limiting their adaptability as well as flexibility in complex interaction environments and leading to subpar performance on highly specialized and expert-level tasks. To address these issues, we introduce HALO, a multi-agent collaboration framework based on a hierarchical reasoning architecture. Specifically, we incorporate a high-level planning agent for task decomposition, mid-level role-design agents for subtask-specific agent instantiation, and low-level inference agents for subtask execution. Particularly, subtask execution is reformulated as a structured workflow search problem, where Monte Carlo Tree Search (MCTS) systematically explores the agentic action space to construct optimal reasoning trajectories. Additionally, as the majority of users lack expertise in prompt engineering, we leverage an Adaptive Prompt Refinement module to transform raw queries into task-specific prompts. Empirical evaluations on Code Generation (HumanEval), General Reasoning (MMLU), and Arithmetic Reasoning (MATH) benchmark datasets highlight the effectiveness of HALO, yielding a 14.4% average improvement over state-of-the-art baselines. Notably, HALO achieves up to 13.3% performance gain on the Moral Scenarios subject in the MMLU benchmark and up to 19.6% performance gain on the Algebra subarea in the MATH benchmark, indicating its advanced proficiency in tackling highly specialized and expert-level tasks. The code repository is available at https://github.com/23japhone/HALO.

HALO: Hierarchical Autonomous Logic-Oriented Orchestration for Multi-Agent LLM Systems

TL;DR

HALO introduces a Hierarchical Autonomous Logic-Oriented Orchestration framework for multi-agent LLM systems, replacing fixed agent-role spaces and static workflows with a three-stage pipeline: Adaptive Prompt Refinement, Hierarchical Reasoning Stack, and a Workflow Search Engine guided by Monte Carlo Tree Search. The top-level planner decomposes tasks, mid-level role-design agents instantiate specialized sub-agents, and low-level inference agents execute subtasks, with MCTS constructing and evaluating reasoning trajectories. Empirical results on HumanEval, MMLU, and MATH show HALO achieves a 14.4% average improvement over SOTA baselines, with up to 19.6% gains on specific expert-level subareas and up to 13.3% on Moral Scenarios, illustrating strong capabilities in complex tasks. The work also demonstrates the value of adaptive prompting and structured workflow search for scalable, adaptable MAS in LLM-based systems, with code available at the project repository.

Abstract

Recent advancements in Multi-Agent Systems (MAS) powered by Large Language Models (LLMs) have demonstrated tremendous potential in diverse task scenarios. Nonetheless, existing agentic systems typically rely on predefined agent-role design spaces and static communication structures, limiting their adaptability as well as flexibility in complex interaction environments and leading to subpar performance on highly specialized and expert-level tasks. To address these issues, we introduce HALO, a multi-agent collaboration framework based on a hierarchical reasoning architecture. Specifically, we incorporate a high-level planning agent for task decomposition, mid-level role-design agents for subtask-specific agent instantiation, and low-level inference agents for subtask execution. Particularly, subtask execution is reformulated as a structured workflow search problem, where Monte Carlo Tree Search (MCTS) systematically explores the agentic action space to construct optimal reasoning trajectories. Additionally, as the majority of users lack expertise in prompt engineering, we leverage an Adaptive Prompt Refinement module to transform raw queries into task-specific prompts. Empirical evaluations on Code Generation (HumanEval), General Reasoning (MMLU), and Arithmetic Reasoning (MATH) benchmark datasets highlight the effectiveness of HALO, yielding a 14.4% average improvement over state-of-the-art baselines. Notably, HALO achieves up to 13.3% performance gain on the Moral Scenarios subject in the MMLU benchmark and up to 19.6% performance gain on the Algebra subarea in the MATH benchmark, indicating its advanced proficiency in tackling highly specialized and expert-level tasks. The code repository is available at https://github.com/23japhone/HALO.
Paper Structure (30 sections, 11 equations, 8 figures, 3 tables, 1 algorithm)

This paper contains 30 sections, 11 equations, 8 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related work
  3. Prompt optimization
  4. Role design in LLM-based architecture
  5. Cooperation optimization strategies
  6. HALO framework
  7. Problem formulation
  8. Adaptive Prompt Refinement
  9. Hierarchical Reasoning Stack
  10. Workflow Search Engine
  11. Selection.
  12. Expansion.
  13. Simulation.
  14. Backpropagation.
  15. Experiments
  16. ...and 15 more sections

Figures (8)

  • Figure 1: The overview of HALO framework. HALO consists of three modules: (1) Adaptive Prompt Refinement (Section \ref{['sec_3.2_prompt_refinement']}), where user queries are refined into high-quality and LLM-comprehensible prompts; (2) Hierarchical Reasoning Stack (Section \ref{['sec_3.3_hierarchical_reasoning']}), which is responsible for task decomposition, role instantiation, and subtask execution; and (3) Workflow Search Engine (Section \ref{['sec_3.4_workflow_search']}), which explores multi-agent collaboration and constructs optimal workflows. Green paths denote optimal reasoning trajectories, while red paths are pruned during search.
  • Figure 2: The illustration of how Monte Carlo Tree Search (MCTS) guides multi-agent reasoning through selection, expansion, simulation, and backpropagation stages. Each node represents an agent and edge transitions are guided by execution outcomes as well as evaluation feedback.
  • Figure 3: Performance comparison on three computationally intensive subareas selected from the MATH dataset. Metrics are reported as $accuracy$ (%) averaged over three runs.
  • Figure 4: Ablation study of removing the Adaptive Prompt Refinement module and the high-level planning agent on GPT-4o across three benchmarks.
  • Figure 5: System prompts used in the Adaptive Prompt Refinement module. The refinement process is conducted through four specialized agents: the Task Parser Agent extracts task semantics from user queries; the Prompt Template Agent constructs a structured prompt template; the Prompt Optimization Agent enhances clarity and usability; and the Prompt Generator Agent produces the final prompt.
  • ...and 3 more figures