Step-DeepResearch Technical Report

TL;DR

This paper introduces Step-DeepResearch, an end-to-end agent for open-ended research tasks, with novel training strategies and evaluation benchmarks, enabling medium-sized models to reach expert-level performance efficiently.

Contribution

It presents a new cost-effective agent architecture, a data synthesis strategy, a progressive training path, and a Chinese research benchmark, advancing open-ended research capabilities of medium-sized models.

Findings

Step-DeepResearch achieves 61.4% on Scale AI Research Rubrics.

It outperforms comparable models on ADR-Bench.

It rivals SOTA closed-source models like OpenAI and Gemini DeepResearch.

Abstract

As LLMs shift toward autonomous agents, Deep Research has emerged as a pivotal metric. However, existing academic benchmarks like BrowseComp often fail to meet real-world demands for open-ended research, which requires robust skills in intent recognition, long-horizon decision-making, and cross-source verification. To address this, we introduce Step-DeepResearch, a cost-effective, end-to-end agent. We propose a Data Synthesis Strategy Based on Atomic Capabilities to reinforce planning and report writing, combined with a progressive training path from agentic mid-training to SFT and RL. Enhanced by a Checklist-style Judger, this approach significantly improves robustness. Furthermore, to bridge the evaluation gap in the Chinese domain, we establish ADR-Bench for realistic deep research scenarios. Experimental results show that Step-DeepResearch (32B) scores 61.4% on Scale AI Research Rubrics. On ADR-Bench, it significantly outperforms comparable models and rivals SOTA closed-source models like OpenAI and Gemini DeepResearch. These findings prove that refined training enables medium-sized models to achieve expert-level capabilities at industry-leading cost-efficiency.