PRIME: Planning and Retrieval-Integrated Memory for Enhanced Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tran, Hieu, Yao, Zonghai, Tran, Nguyen Luong, Yang, Zhichao, Ouyang, Feiyun, Han, Shuo, Rahimi, Razieh, Yu, Hong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917072724819968
author Tran, Hieu
Yao, Zonghai
Tran, Nguyen Luong
Yang, Zhichao
Ouyang, Feiyun
Han, Shuo
Rahimi, Razieh
Yu, Hong
author_facet Tran, Hieu
Yao, Zonghai
Tran, Nguyen Luong
Yang, Zhichao
Ouyang, Feiyun
Han, Shuo
Rahimi, Razieh
Yu, Hong
contents Inspired by the dual-process theory of human cognition from \textit{Thinking, Fast and Slow}, we introduce \textbf{PRIME} (Planning and Retrieval-Integrated Memory for Enhanced Reasoning), a multi-agent reasoning framework that dynamically integrates \textbf{System 1} (fast, intuitive thinking) and \textbf{System 2} (slow, deliberate thinking). PRIME first employs a Quick Thinking Agent (System 1) to generate a rapid answer; if uncertainty is detected, it then triggers a structured System 2 reasoning pipeline composed of specialized agents for \textit{planning}, \textit{hypothesis generation}, \textit{retrieval}, \textit{information integration}, and \textit{decision-making}. This multi-agent design faithfully mimics human cognitive processes and enhances both efficiency and accuracy. Experimental results with LLaMA 3 models demonstrate that PRIME enables open-source LLMs to perform competitively with state-of-the-art closed-source models like GPT-4 and GPT-4o on benchmarks requiring multi-hop and knowledge-grounded reasoning. This research establishes PRIME as a scalable solution for improving LLMs in domains requiring complex, knowledge-intensive reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2509_22315
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PRIME: Planning and Retrieval-Integrated Memory for Enhanced Reasoning
Tran, Hieu
Yao, Zonghai
Tran, Nguyen Luong
Yang, Zhichao
Ouyang, Feiyun
Han, Shuo
Rahimi, Razieh
Yu, Hong
Artificial Intelligence
Computation and Language
Inspired by the dual-process theory of human cognition from \textit{Thinking, Fast and Slow}, we introduce \textbf{PRIME} (Planning and Retrieval-Integrated Memory for Enhanced Reasoning), a multi-agent reasoning framework that dynamically integrates \textbf{System 1} (fast, intuitive thinking) and \textbf{System 2} (slow, deliberate thinking). PRIME first employs a Quick Thinking Agent (System 1) to generate a rapid answer; if uncertainty is detected, it then triggers a structured System 2 reasoning pipeline composed of specialized agents for \textit{planning}, \textit{hypothesis generation}, \textit{retrieval}, \textit{information integration}, and \textit{decision-making}. This multi-agent design faithfully mimics human cognitive processes and enhances both efficiency and accuracy. Experimental results with LLaMA 3 models demonstrate that PRIME enables open-source LLMs to perform competitively with state-of-the-art closed-source models like GPT-4 and GPT-4o on benchmarks requiring multi-hop and knowledge-grounded reasoning. This research establishes PRIME as a scalable solution for improving LLMs in domains requiring complex, knowledge-intensive reasoning.
title PRIME: Planning and Retrieval-Integrated Memory for Enhanced Reasoning
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2509.22315