In plain words: It breaks a question into small steps and, at each one, decides whether to look up outside information or answer from the model's own memory. Compared with searching for everything at once, this cut needless lookups and raised answer accuracy by 26.4%.
Abstract · DeepRAG: Thinking to Retrieve Step by Step for Large Language Models
Large Language Models (LLMs) have shown remarkable reasoning capabilities, while their practical applications are limited by severe factual hallucinations due to limitations in the timeliness, accuracy, and comprehensiveness of their parametric knowledge. Meanwhile, enhancing retrieval-augmented generation (RAG) with reasoning remains challenging due to ineffective task decomposition and redundant retrieval, which can introduce noise and degrade response quality. In this paper, we propose DeepRAG, a framework that models retrieval-augmented reasoning as a Markov Decision Process (MDP), enabling reasonable and adaptive retrieval. By iteratively decomposing queries, DeepRAG dynamically determines whether to retrieve external knowledge or rely on parametric reasoning at each step. Experiments show that DeepRAG improves retrieval efficiency and boosts answer accuracy by 26.4%, demonstrating its effectiveness in enhancing retrieval-augmented reasoning.
Xinyan Guan, Jiali Zeng, Fandong Meng, Chunlei Xin, Yaojie Lu, Hongyu Lin, Xianpei Han, Le Sun, Jie Zhou
arXiv:2502.01142 · cs.AI, cs.CL, cs.IR · submitted Feb 3, 2025 · updated Jun 8, 2025
abstract · pdf · html
I’ve been thinking about how modifying AWM to use fine-tuning or an external knowledge system (RAG) might work—capturing the ‘good’ workflows it discovers rather than relying purely on prompting.
[1] https://arxiv.org/abs/2409.07429 - Agent Workflow Memory (Wang et al., 2024)