arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.01489cs.AIcs.CL
收藏

SciResearcher: 面向前沿科学推理的深度研究智能体规模化

SciResearcher: Scaling Deep Research Agents for Frontier Scientific Reasoning

HKUST(香港科技大学);CUHK(香港大学);Tencent AI Lab(腾讯AI实验室)

Tianshi Zheng, Rui Wang, Xiyun Li, Kelvin Kiu Wai Tam, Newt Nguyen Kim Hue Nam, Wei Fan, Yangqiu Song, Tianqing Fang

AI总结:

提出SciResearcher框架,通过合成基于学术证据的概念与计算任务并训练智能体,在HLE-Bio/Chem-Gold等基准上达到最优性能。

摘要:

前沿科学推理正迅速成为推动AI智能体在自动化科学发现中的关键基础。深度研究智能体为此挑战提供了有前景的方法。这些模型通过后训练处理信息寻求任务(通常通过知识图谱构建或迭代网页浏览来策划)来发展强大的问题解决能力。然而,这些策略在前沿科学中面临固有局限性,因为领域特定知识分散在稀疏且异构的学术来源中,而问题解决需要远超事实回忆的复杂计算和推理。为弥合这一差距,我们引入了SciResearcher,一个用于前沿科学数据构建的全自动智能体框架。SciResearcher综合了基于学术证据的多样化概念和计算任务,同时激发信息获取、工具集成推理和长程能力。利用策划的数据进行监督微调和智能体强化学习,我们开发了SciResearcher-8B,一个在HLE-Bio/Chem-Gold基准上达到19.46%的智能体基础模型,在其参数规模上建立了新的最先进水平,并超越了多个更大的专有智能体。它在SuperGPQA-Hard-Biology和TRQA-Literature基准上进一步取得了13-15%的绝对提升。总体而言,SciResearcher为前沿科学推理的自动数据构建引入了一种新范式,并为未来的科学智能体提供了一条可扩展的路径。

英文摘要:

Frontier scientific reasoning is rapidly emerging as a key foundation for advancing AI agents in automated scientific discovery. Deep research agents offer a promising approach to this challenge. These models develop robust problem-solving capabilities through post-training on information-seeking tasks, which are typically curated via knowledge graph construction or iterative web browsing. However, these strategies face inherent limitations in frontier science, where domain-specific knowledge is scattered across sparse and heterogeneous academic sources, and problem solving requires sophisticated computation and reasoning far beyond factual recall. To bridge this gap, we introduce SciResearcher, a fully automated agentic framework for frontier-science data construction. SciResearcher synthesizes diverse conceptual and computational tasks grounded in academic evidence, while eliciting information acquisition, tool-integrated reasoning, and long-horizon capabilities. Leveraging the curated data for supervised fine-tuning and agentic reinforcement learning, we develop SciResearcher-8B, an agent foundation model that achieves 19.46% on the HLE-Bio/Chem-Gold benchmark, establishing a new state of the art at its parameter scale and surpassing several larger proprietary agents. It further achieves 13-15% absolute gains on SuperGPQA-Hard-Biology and TRQA-Literature benchmarks. Overall, SciResearcher introduces a new paradigm for automated data construction for frontier scientific reasoning and offers a scalable path toward future scientific agents.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑