A Training-free LLM Framework with Interaction between Contextually Related Subtasks in Solving Complex Tasks

2025-03-29

AI Summary

In complex tasks, static task decomposition leads to context fragmentation, causing information loss and subtask execution failure. Method: This paper proposes a training-free, interactive LLM framework that replaces static decomposition with a subtask trajectory memory mechanism and refined execution summarization, enabling dynamic reactivation and precise reuse of completed subtasks; it further introduces a lightweight request-response protocol to support proactive, on-demand contextual interaction among subtasks. The method leverages GPT-3.5/GPT-4 for zero-shot inference without fine-tuning or additional parameters. Results: Experiments on WebShop and HotpotQA demonstrate substantial improvements over existing zero-shot baselines, achieving significant gains in multi-hop reasoning and interactive decision-making accuracy. To our knowledge, this is the first approach to realize dynamic, context-aware collaborative reasoning without any training.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: (Large) Language Models

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large language models (LLMs) have shown remarkable capabilities in solving complex tasks. Recent work has explored decomposing such tasks into subtasks with independent contexts. However, some contextually related subtasks may encounter information loss during execution, leading to redundant operations or execution failures. To address this issue, we propose a training-free framework with an interaction mechanism, which enables a subtask to query specific information or trigger certain actions in completed subtasks by sending requests. To implement interaction, we introduce a subtask trajectory memory to enable resumption of completed subtasks upon receiving interaction requests. Additionally, we propose a new action during execution, which generates a concise and precise description of execution process and outcomes of a subtask, to assist subsequent subtasks in determining interaction targets and requests. We evaluate our framework on interactive decision-making task WebShop and multi-hop question answering HotpotQA, with GPT-3.5 and GPT-4, and comparison results show that our framework outperforms the state-of-the-art training-free baselines.

Problem

• Addresses information loss in contextually related subtasks
• Enables interaction between subtasks via query and trigger
• Improves performance in complex tasks without training

Innovation

• Training-free LLM framework with interaction mechanism
• Subtask trajectory memory for resuming completed subtasks
• Concise action descriptions for guiding subsequent subtasks