DeepAnalyze: Agentic Large Language Models for Autonomous Data Science

TL;DR

This work introduces DeepAnalyze-8B, an agentic 8B-parameter LLM capable of autonomous data science—from data ingestion to analyst-grade reporting—without predefined workflows. It couples a curriculum-based training regime with a data-grounded trajectory synthesis framework to address reward sparsity and trajectory scarcity, enabling autonomous orchestration and adaptive optimization in real-world environments. The approach combines single-ability pretraining and multi-ability RL with hybrid rewards, supported by Reasoning and Interaction Trajectory Synthesis and a large, released dataset (DataScience-Instruct-500K). Experiments across 12 benchmarks show DeepAnalyze-8B achieving strong, often state-of-the-art, performance among open-source agents and notable capabilities in end-to-end data research and report generation, marking a shift from workflow-based agents to fully autonomous data science systems. The work provides open-source resources and establishes a pathway for scalable autonomous data science, with implications for data discovery, governance, and management.

Abstract

Autonomous data science, from raw data sources to analyst-grade deep research reports, has been a long-standing challenge, and is now becoming feasible with the emergence of powerful large language models (LLMs). Recent workflow-based data agents have shown promising results on specific data tasks but remain fundamentally limited in achieving fully autonomous data science due to their reliance on predefined workflows. In this paper, we introduce DeepAnalyze-8B, the first agentic LLM designed for autonomous data science, capable of automatically completing the end-toend pipeline from data sources to analyst-grade deep research reports. To tackle high-complexity data science tasks, we propose a curriculum-based agentic training paradigm that emulates the learning trajectory of human data scientists, enabling LLMs to progressively acquire and integrate multiple capabilities in real-world environments. We also introduce a data-grounded trajectory synthesis framework that constructs high-quality training data. Through agentic training, DeepAnalyze learns to perform a broad spectrum of data tasks, ranging from data question answering and specialized analytical tasks to open-ended data research. Experiments demonstrate that, with only 8B parameters, DeepAnalyze outperforms previous workflow-based agents built on most advanced proprietary LLMs. The model, code, and training data of DeepAnalyze are open-sourced, paving the way toward autonomous data science.

Figures (15)

• Figure 1: DeepAnalyze-8B is the first end-to-end agentic LLM that achieves autonomous data science, supporting entire data science pipeline and open-ended data research.
• Figure 2: Examples of DeepAnalyze-8B. Given the instructions and data sources in the environment, DeepAnalyze can autonomously orchestrate and optimize actions to complete a data science pipeline (left) and open-ended data research (right). DeepAnalyze first performs planning, then interacts with the data in the environment, and subsequently optimizes its actions based on feedback, ultimately accomplishing the data-centric tasks. Many intermediate actions are omitted to save space.
• Figure 3: Architecture of DeepAnalyze.
• Figure 4: Schematic diagram of agentic RL.
• Figure 5: The proposed data-grounded trajectory synthesis for the development of DeepAnalyze on data science tasks.