Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reason-Plan-ReAct: A Reasoner-Planner Supervising a ReAct Executor for Complex Enterprise Tasks

Gianni Molinari, Fabio Ciravegna

TL;DR

The paper tackles reliability and privacy constraints in enterprise autonomous agents by addressing trajectory instability and context-window overflow in monolithic planners. It proposes RP-ReAct, a two-agent architecture separating Reasoner Planner (RPA) from Proxy-Execution Agents (PEA), with a context-management mechanism. The system uses open-weight reasoning configurations and the ToolQA benchmark across domains, showing superior performance, robustness, and generalization, especially on hard tasks, across model scales. This work provides a practical pathway toward deployable, open-weight–based enterprise agents through architectural separation and explicit context handling.

Abstract

Despite recent advances, autonomous agents often struggle to solve complex tasks in enterprise domains that require coordinating multiple tools and processing diverse data sources. This struggle is driven by two main limitations. First, single-agent architectures enforce a monolithic plan-execute loop, which directly causes trajectory instability. Second, the requirement to use local open-weight models for data privacy introduces smaller context windows leading to the rapid consumption of context from large tool outputs. To solve this problem we introduce RP-ReAct (Reasoner Planner-ReAct), a novel multi-agent approach that fundamentally decouples strategic planning from low-level execution to achieve superior reliability and efficiency. RP-ReAct consists of a Reasoner Planner Agent (RPA), responsible for planning each sub-step, continuously analysing the execution results using the strong reasoning capabilities of a Large Reasoning Model, and one or multiple Proxy-Execution Agent (PEA) that translates sub-steps into concrete tool interactions using a ReAct approach. Crucially, we incorporate a context-saving strategy within the PEA to mitigate context window overflow by managing large tool outputs via external storage and on-demand access. We evaluate RP-ReAct, on the challenging, multi-domain ToolQA benchmark using a diverse set of six open-weight reasoning models. Our empirical results show that RP-ReAct achieves superior performance and improved generalization ability over state-of-the-art baselines when addressing diverse complex tasks across the evaluated domains. Furthermore we establish the enhanced robustness and stability of our approach across different model scales, paving the way for effective and deployable agentic solutions for enterprises.

Reason-Plan-ReAct: A Reasoner-Planner Supervising a ReAct Executor for Complex Enterprise Tasks

TL;DR

The paper tackles reliability and privacy constraints in enterprise autonomous agents by addressing trajectory instability and context-window overflow in monolithic planners. It proposes RP-ReAct, a two-agent architecture separating Reasoner Planner (RPA) from Proxy-Execution Agents (PEA), with a context-management mechanism. The system uses open-weight reasoning configurations and the ToolQA benchmark across domains, showing superior performance, robustness, and generalization, especially on hard tasks, across model scales. This work provides a practical pathway toward deployable, open-weight–based enterprise agents through architectural separation and explicit context handling.

Abstract

Despite recent advances, autonomous agents often struggle to solve complex tasks in enterprise domains that require coordinating multiple tools and processing diverse data sources. This struggle is driven by two main limitations. First, single-agent architectures enforce a monolithic plan-execute loop, which directly causes trajectory instability. Second, the requirement to use local open-weight models for data privacy introduces smaller context windows leading to the rapid consumption of context from large tool outputs. To solve this problem we introduce RP-ReAct (Reasoner Planner-ReAct), a novel multi-agent approach that fundamentally decouples strategic planning from low-level execution to achieve superior reliability and efficiency. RP-ReAct consists of a Reasoner Planner Agent (RPA), responsible for planning each sub-step, continuously analysing the execution results using the strong reasoning capabilities of a Large Reasoning Model, and one or multiple Proxy-Execution Agent (PEA) that translates sub-steps into concrete tool interactions using a ReAct approach. Crucially, we incorporate a context-saving strategy within the PEA to mitigate context window overflow by managing large tool outputs via external storage and on-demand access. We evaluate RP-ReAct, on the challenging, multi-domain ToolQA benchmark using a diverse set of six open-weight reasoning models. Our empirical results show that RP-ReAct achieves superior performance and improved generalization ability over state-of-the-art baselines when addressing diverse complex tasks across the evaluated domains. Furthermore we establish the enhanced robustness and stability of our approach across different model scales, paving the way for effective and deployable agentic solutions for enterprises.

Paper Structure

This paper contains 34 sections, 4 equations, 1 figure, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. LLM Agents
  4. Large Reasoning Models
  5. Large Reasoning Models
  6. Methodology
  7. Reasoner-Planner Agent (RPA):
  8. Proxy-Execution Agent (PEA):
  9. Context Management
  10. Experimental Setup
  11. Dataset
  12. Baseline Agent Architectures
  13. Prompts
  14. Models
  15. Evaluation Metrics
  16. ...and 19 more sections

Figures (1)

  • Figure 1: A PEA receives the sub-question from the RPA and tries to solve it by interacting with tools using the ReAct approach, then returns the result to the RPA.