Step-DeepResearch Technical Report

Chen Hu, Haikuo Du, Heng Wang, Lin Lin, Mingrui Chen, Peng Liu, Ruihang Miao, Tianchi Yue, Wang You, Wei Ji, Wei Yuan, Wenjin Deng, Xiaojian Yuan, Xiaoyun Zhang, Xiangyu Liu, Xikai Liu, Yanming Xu, Yicheng Cao, Yifei Zhang, Yongyao Wang, Yubo Shu, Yurong Zhang, Yuxiang Zhang, Zheng Gong, Zhichao Chang, Binyan Li, Dan Ma, Furong Jia, Hongyuan Wang, Jiayu Liu, Jing Bai, Junlan Liu, Manjiao Liu, Na Wang, Qiuping Wu, Qinxin Du, Shiwei Li, Wen Sun, Yifeng Gong, Yonglin Chen, Yuling Zhao, Ziqi Ren, Zixuan Wang, Aihu Zhang, Brian Li, Buyun Ma, Kang An, Li Xie, Mingliang Li, Pan Li, Shidong Yang, Xi Chen, Xiaojia Liu, Yuchu Luo, Yuan Song, YuanHao Ding, Yuanwei Liang, Zexi Li, Zhaoning Zhang, Zixin Zhang, Binxing Jiao, Daxin Jiang, Jiansheng Chen, Jing Li, Xiangyu Zhang, Yibo ZhuarXiv 2025 · DOI · arXiv

Introduction

Modern AI agents have excelled at multi-hop search and information retrieval, yet a fundamental gap remains: "Search is not research." While current systems can crawl the web for facts, they often struggle with intent decomposition, cross-source verification, and the synthesis of evidence into coherent, defensible arguments. Step-DeepResearch addresses these challenges by internalizing an "expert cognitive loop" into a 32B-parameter model, achieving performance parity with proprietary giants like OpenAI and Gemini Deep Research at a fraction of the operational cost.

Methodology: Atomic Capability Decomposition

The core of Step-DeepResearch is the decomposition of complex research into four "Atomic Capabilities." Rather than training on raw token sequences, the model is optimized to execute specific cognitive actions:

  1. Planning & Task Decomposition: Utilizing "Reverse Engineering Synthesis," the model learns to work backward from high-quality reports to original queries and structured plans.
  2. Deep Search & Information Seeking: Graph-based synthesis and multi-document "topology walks" enable the model to navigate complex information landscapes.
  3. Reflection & Verification: A closed-loop "Error-Reflection" process allows the model to learn from failed search trajectories and self-correct.
  4. Report Generation: Focuses on domain-specific styles and rigorous citation alignment.

Three-Stage Training Pipeline

The model undergoes a rigorous training process:

  • Stage 1: Agentic Mid-training: Context scaling from 32K to 128K tokens while injecting atomic capabilities.
  • Stage 2: Post-training SFT: Connecting capabilities into end-to-end trajectories, prioritizing the "correct and shortest" paths for maximum efficiency.
  • Stage 3: Reinforcement Learning (RL): PPO training using a "Rubrics Judge" reward model. A strict binary reward mapping is used to prevent the exploitation of "partially correct" signals.

Experimental Setup

The model was evaluated on two primary benchmarks:

  • ResearchRubrics: A diverse set of 101 domain tasks.
  • ADR-Bench (Application-driven Deep Research Benchmark): A new Chinese-language benchmark featuring 110 queries across 9 domains (e.g., Law, Finance, STEM) using an Elo-style human rating protocol.

Results and Discussion

Step-DeepResearch demonstrates that medium-sized models (32B) can rival much larger 70B+ or proprietary models when training focuses on the reasoning process.

Performance and Cost-Efficiency

On the ResearchRubrics benchmark, Step-DeepResearch achieved a score of 61.42, outperforming leading open-source models like Kimi-k2-thinking (56.17) and trailing only Gemini DeepResearch (63.69). Crucially, this performance is achieved with unprecedented efficiency.

Cost-Efficiency on ResearchRubrics

As shown in the cost-efficiency analysis, Step-DeepResearch operates at less than 0.50 RMB per report—less than 1/10th the cost of OpenAI or Gemini Deep Research.

Domain-Specific Excellence

In human evaluations on the ADR-Bench, Step-DeepResearch maintained a consistently high win rate across multiple dimensions, including content depth and information completeness.

ADR-Bench Performance Comparison

The model tied for the top spot in "Technical Documentation" and "Citation Quality," proving its utility in professional, high-stakes environments.

Conclusion

Step-DeepResearch proves that high-quality synthetic data and a focus on "atomic" reasoning capabilities can elevate medium-scale models to expert-level performance. While some limitations remain—such as "temporal confusion" and slight gaps in specialized STEM reasoning compared to the largest models—the efficiency and robustness of this approach mark a significant step toward autonomous, cost-effective research agents. Future work will explore multi-agent paradigms to further reduce hallucinations and improve auditability.

Summary updated

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.
BibTeX
@misc{hu2025step,
  author = {Chen Hu and Haikuo Du and Heng Wang and Lin Lin and Mingrui Chen and Peng Liu and Ruihang Miao and Tianchi Yue and Wang You and Wei Ji and Wei Yuan and Wenjin Deng and Xiaojian Yuan and Xiaoyun Zhang and Xiangyu Liu and Xikai Liu and Yanming Xu and Yicheng Cao and Yifei Zhang and Yongyao Wang and Yubo Shu and Yurong Zhang and Yuxiang Zhang and Zheng Gong and Zhichao Chang and Binyan Li and Dan Ma and Furong Jia and Hongyuan Wang and Jiayu Liu and Jing Bai and Junlan Liu and Manjiao Liu and Na Wang and Qiuping Wu and Qinxin Du and Shiwei Li and Wen Sun and Yifeng Gong and Yonglin Chen and Yuling Zhao and Yuxuan Lin and Ziqi Ren and Zixuan Wang and Aihu Zhang and Brian Li and Buyun Ma and Kang An and Li Xie and Mingliang Li and Pan Li and Shidong Yang and Xi Chen and Xiaojia Liu and Yuchu Luo and Yuan Song and YuanHao Ding and Yuanwei Liang and Zexi Li and Zhaoning Zhang and Zixin Zhang and Binxing Jiao and Daxin Jiang and Jiansheng Chen and Jing Li and Xiangyu Zhang and Yibo Zhu},
  title = {{Step-DeepResearch Technical Report}},
  year = {2025},
  eprint = {2512.20491},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2512.20491},
}
Figure