Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GOLF: Goal-Oriented Long-term liFe tasks supported by human-AI collaboration

Ben Wang

TL;DR

The paper proposes GOLF, a goal-oriented framework that extends LLM capabilities to support long-term life tasks through multi-agent collaboration, decomposing tasks into information-seeking, decision-making, and physical-action steps. It combines a GTD/PDCA-inspired planning pipeline with a rigorous simulation and evaluation plan, validated via Autogen-based multi-agent experiments and crowdsourced assessments. A dedicated dataset and benchmark for long-term task planning are developed, with deployment considerations across models from GPT-4 to on-device open-source options like Llama2-7B and Phi-2. The work emphasizes human-AI collaboration to enhance long-term decision-making and task management while foregrounding ethical, privacy, and evaluation considerations for trustworthy application.

Abstract

The advent of ChatGPT and similar large language models (LLMs) has revolutionized the human-AI interaction and information-seeking process. Leveraging LLMs as an alternative to search engines, users can now access summarized information tailored to their queries, significantly reducing the cognitive load associated with navigating vast information resources. This shift underscores the potential of LLMs in redefining information access paradigms. Drawing on the foundation of task-focused information retrieval and LLMs' task planning ability, this research extends the scope of LLM capabilities beyond routine task automation to support users in navigating long-term and significant life tasks. It introduces the GOLF framework (Goal-Oriented Long-term liFe tasks), which focuses on enhancing LLMs' ability to assist in significant life decisions through goal orientation and long-term planning. The methodology encompasses a comprehensive simulation study to test the framework's efficacy, followed by model and human evaluations to develop a dataset benchmark for long-term life tasks, and experiments across different models and settings. By shifting the focus from short-term tasks to the broader spectrum of long-term life goals, this research underscores the transformative potential of LLMs in enhancing human decision-making processes and task management, marking a significant step forward in the evolution of human-AI collaboration.

GOLF: Goal-Oriented Long-term liFe tasks supported by human-AI collaboration

TL;DR

The paper proposes GOLF, a goal-oriented framework that extends LLM capabilities to support long-term life tasks through multi-agent collaboration, decomposing tasks into information-seeking, decision-making, and physical-action steps. It combines a GTD/PDCA-inspired planning pipeline with a rigorous simulation and evaluation plan, validated via Autogen-based multi-agent experiments and crowdsourced assessments. A dedicated dataset and benchmark for long-term task planning are developed, with deployment considerations across models from GPT-4 to on-device open-source options like Llama2-7B and Phi-2. The work emphasizes human-AI collaboration to enhance long-term decision-making and task management while foregrounding ethical, privacy, and evaluation considerations for trustworthy application.

Abstract

The advent of ChatGPT and similar large language models (LLMs) has revolutionized the human-AI interaction and information-seeking process. Leveraging LLMs as an alternative to search engines, users can now access summarized information tailored to their queries, significantly reducing the cognitive load associated with navigating vast information resources. This shift underscores the potential of LLMs in redefining information access paradigms. Drawing on the foundation of task-focused information retrieval and LLMs' task planning ability, this research extends the scope of LLM capabilities beyond routine task automation to support users in navigating long-term and significant life tasks. It introduces the GOLF framework (Goal-Oriented Long-term liFe tasks), which focuses on enhancing LLMs' ability to assist in significant life decisions through goal orientation and long-term planning. The methodology encompasses a comprehensive simulation study to test the framework's efficacy, followed by model and human evaluations to develop a dataset benchmark for long-term life tasks, and experiments across different models and settings. By shifting the focus from short-term tasks to the broader spectrum of long-term life goals, this research underscores the transformative potential of LLMs in enhancing human decision-making processes and task management, marking a significant step forward in the evolution of human-AI collaboration.
Paper Structure (11 sections, 1 figure)

This paper contains 11 sections, 1 figure.

Table of Contents

  1. Motivation
  2. Related Work
  3. Tasks in Information Seeking and Retrieval
  4. LLM Agent and Task Automation
  5. Research Questions
  6. Proposed Methodology
  7. The GOLF Framework
  8. Simulation Study
  9. Evaluation
  10. Model Experiment and Deployment
  11. Specific Research Issues for Discussion

Figures (1)

  • Figure 1: The GOLF framework: (a) a simplified illustration of a golf course (the background is generated by StableDiffusion); (b) a variation of the task taxonomy soufan2021untangling; (c) the task process in the GOLF framework. The blue lines represent the potential path to complete the task, activity, or goal. Dotted lines represent users' actions, and solid lines represent LLM agents' actions.