@taesiri\n\t's contributions. It appears the latest version has been updated on arXiv. Could you please help update the corresponding paper version on Hugging Face?\" \n","updatedAt":"2025-12-25T04:43:31.677Z","author":{"_id":"648183bfea00b120773198ba","avatarUrl":"/avatars/b69d04bba11e393d1a59e98653d5cfa6.svg","fullname":"huchen","name":"ccchen1006","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":4,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9011975526809692},"editors":["ccchen1006"],"editorAvatarUrls":["/avatars/b69d04bba11e393d1a59e98653d5cfa6.svg"],"reactions":[],"isReport":false,"parentCommentId":"694b513c6c32517d14f0fbb5"}},{"id":"694cc70a2b3837395056041c","author":{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","fullname":"taesiri","name":"taesiri","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":235,"isUserFollowing":false},"createdAt":"2025-12-25T05:09:30.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Hey @ccchen1006 \nWhich part of this page needs to be updated? When I click the PDF, it shows the latest version (December 24). Isn’t that already the most recent one?\n","html":"
Hey \n\n@ccchen1006\n\t
Which part of this page needs to be updated? When I click the PDF, it shows the latest version (December 24). Isn’t that already the most recent one?
Could you please add the GitHub link? Thanks a lot!
https://github.com/stepfun-ai/StepDeepResearch
Done!
\n","updatedAt":"2025-12-25T09:13:24.425Z","author":{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","fullname":"taesiri","name":"taesiri","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":235,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.6595590114593506},"editors":["taesiri"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg"],"reactions":[{"reaction":"👍","users":["ccchen1006"],"count":1}],"isReport":false,"parentCommentId":"694b513c6c32517d14f0fbb5"}}]},{"id":"694f0f4c785ac06cb1d374bd","author":{"_id":"65243980050781c16f234f1f","avatarUrl":"/avatars/743a009681d5d554c27e04300db9f267.svg","fullname":"Avi","name":"avahal","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":3,"isUserFollowing":false},"createdAt":"2025-12-26T22:42:20.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"arXiv lens breakdown of this paper 👉 https://arxivlens.com/PaperView/Details/step-deepresearch-technical-report-1855-b4852a8e\n- Executive Summary\n- Detailed Breakdown\n- Practical Applications","html":"arXiv lens breakdown of this paper 👉 https://arxivlens.com/PaperView/Details/step-deepresearch-technical-report-1855-b4852a8e
\n- \n
- Executive Summary \n
- Detailed Breakdown \n
- Practical Applications \n
Papers
arxiv:2512.20491
Step-DeepResearch Technical Report
Abstract
Step-DeepResearch, an end-to-end agent enhanced with a data synthesis strategy and progressive training, achieves expert-level capabilities in deep research scenarios, outperforming established models.
AI-generated summary
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.
Community
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.
Many thanks for @taesiri 's contributions. It appears the latest version has been updated on arXiv. Could you please help update the corresponding paper version on Hugging Face?"
arXiv lens breakdown of this paper 👉 https://arxivlens.com/PaperView/Details/step-deepresearch-technical-report-1855-b4852a8e
- Executive Summary
- Detailed Breakdown
- Practical Applications
Models citing this paper 0
No model linking this paper
Cite arxiv.org/abs/2512.20491 in a model README.md to link it from this page.
Datasets citing this paper 0
No dataset linking this paper
Cite arxiv.org/abs/2512.20491 in a dataset README.md to link it from this page.
Spaces citing this paper 0
No Space linking this paper
Cite arxiv.org/abs/2512.20491 in a Space README.md to link it from this page.