Step-DeepResearch Technical Report
By: Chen Hu , Haikuo Du , Heng Wang and more
Potential Business Impact:
Helps AI agents find information better and cheaper.
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.
Similar Papers
DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agents
Computation and Language
Tests AI that writes research reports like a human.
Deep Research: A Systematic Survey
Computation and Language
Lets computers research and answer hard questions.
FinDeepResearch: Evaluating Deep Research Agents in Rigorous Financial Analysis
Computation and Language
Tests how well AI can analyze company money.