发表机构
Tsinghua University; Renmin University of China; Institute of Information Engineering, CAS; USTC; Peking University; University of Macau; The Australian National University（清华大学; 中国人民大学; 中国科学院信息工程研究所; 中国科学技术大学; 北京大学; 澳门大学; 澳大利亚国立大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
SciLENS是一款完全本地化的强化学习驱动自主智能体框架，通过双层基础设施处理约1200万条学术记录，在6个科学基准上性能优于开源基线，与GPT-5.2和Gemini-3.0-pro相当，可用于科学文献合成。
AI 中文摘要
科学文献合成智能体日益依赖专有在线服务，这限制了研究的可复现性、隐私性及离线部署能力。为应对这一挑战，我们推出SciLENS（Scientific Localized Evidence Navigation and Synthesis，科学本地化证据导航与合成），这是一个完全本地化的自主智能体框架，运行在索引了约1200万条学术记录的双层基础设施上。SciLENS率先将结构可视化作为推理循环中的可操作工具，使智能体能够将复杂的引用拓扑结构压缩为经验证的数据驱动图表，从而缓解宏观合成过程中的上下文枯竭问题。为在无需人工标注的情况下训练该智能体，我们开发了一条自动化数据合成流水线，从引用知识图谱中提取多跳子图，并通过20个前沿模型之间的跨模型共识对其进行验证。随后，该智能体通过反向分解评分策略进行对齐，该策略为早期规划提供细粒度过程奖励，并对证据依据提出严格要求。在涵盖标准问答、引用准确性、事实推理及结构合成的6个科学基准上开展的评估表明，SciLENS的性能显著优于开源基线，且达到与GPT-5.2和Gemini-3.0-pro相当的水平。我们的源代码和数据已在该httpsURL发布。
英文摘要
Scientific literature synthesis agents increasingly rely on proprietary online services, limiting reproducibility, privacy, and offline deployment. To address this challenge, we introduce SciLENS Scientific Localized Evidence Navigation and Synthesis), a fully local autonomous agent framework operating on a dual-tier infrastructure indexing approximately 12 million academic records. SciLENS pioneers the integration of structural visualization as an actionable tool within the reasoning loop, enabling the agent to compress complex citation topologies into validated data-driven charts and thereby mitigate context exhaustion during macro-level synthesis. To train the agent without human annotation, we develop an automated data synthesis pipeline that extracts multi-hop subgraphs from a citation knowledge graph, verified by cross-model consensus among 20 frontier models. The agent is subsequently aligned through a reverse-decomposition rubric strategy that provides fine-grained process rewards for early planning and strict evidence grounding. Evaluations across six scientific benchmarks encompassing standard QA, citation accuracy, factual reasoning, and structural synthesis demonstrate that SciLENS significantly outperforms open-source baselines and achieves performance comparable to GPT-5.2 and Gemini-3.0-pro. Our source code and data are released at https://github.com/LQgdwind/SciLENS.
