斯坦福发布Paper2Agent：将论文转为可复现结果的AI Agent
Stanford Researchers Release Paper2Agent: Turning Research Papers Into AI Agents That Reproduce Results and Run on New Data
Nature发表的硬核工程实践，提供从论文到可执行Agent的完整工作流与验证标准，对科研自动化极具参考价值。
Computational papers ship code that readers must clone, install, configure and debug. That cost keeps useful methods locked inside PDFs. A Stanford team led by Jiacheng Miao and James Zou proposes a fix. Paper2Agent was published in Nature on 16 September 2026. It converts a paper and its codebase into a Model Context Protocol (MCP) server. Any MCP-compatible agent, such as Claude Code, can then run the paper’s methods through natural language. The authors describe the result as a virtual corresponding author.
计算类论文提供的代码要求读者必须克隆、安装、配置和调试。这种成本使得许多有用的方法被锁定在 PDF 文件中。由 Jiacheng Miao 和 James Zou 领导的斯坦福大学团队提出了一种解决方案。Paper2Agent 于 2026 年 9 月 16 日发表在《自然》（Nature）杂志上。它将论文及其代码库转换为模型上下文协议（MCP）服务器。任何兼容 MCP 的代理，例如 Claude Code，都可以通过自然语言运行论文中的方法。作者将这一结果描述为“虚拟通讯作者”。
Is it deployable? Yes. The code is MIT-licensed and installs as a skill for Claude Code or Codex. Prebuilt AlphaGenome, Scanpy and TISSUE servers run on Hugging Face Spaces. A hosted version is also available at paper2agent.ai.
它是否可部署？是的。该代码采用 MIT 许可证授权，并可作为技能安装在 Claude Code 或 Codex 中。预构建的 AlphaGenome、Scanpy 和 TISSUE 服务器可在 Hugging Face Spaces 上运行。此外，paper2agent.ai 也提供了托管版本。
How the Pipeline Works
工作流程原理
Paper2Agent runs on Claude Code’s agent SDK. A central orchestrator dispatches specialized sub-agents through 6 steps:
Paper2Agent 基于 Claude Code 的代理 SDK 运行。一个中央协调器通过 6 个步骤调度专门的子代理：
- Locate and download the codebase.
- An environment manager builds an isolated virtual environment.
- A tutorial scanner indexes usable tutorials.
- A tutorial executor runs them end to end and records reference outputs.
- A tool extractor turns tutorials into parameterized MCP tools, and a test verifier validates them.
- The orchestrator assembles validated tools into 1 MCP server.
- 定位并下载代码库。
- 环境管理器构建隔离的虚拟环境。
- 教程扫描器索引可用的教程。
- 教程执行器端到端地运行这些教程并记录参考输出。
- 工具提取器将教程转化为参数化的 MCP 工具，测试验证器对其进行验证。
- 协调器将经过验证的工具组装成 1 个 MCP 服务器。
The validation gate is strict. A tool passes only when expected files appear and numbers match within 3%. Figures must also match references by perceptual hash, with Hamming distance under 20. The verifier gets up to 6 attempts per function. Tools that keep failing are excluded from the final server.
验证关卡非常严格。只有当预期文件出现且数值误差在 3% 以内时，工具才算通过。图表还必须通过感知哈希与参考文献匹配，汉明距离需低于 20。每个函数最多允许验证器尝试 6 次。持续失败的工具将被排除在最终服务器之外。
Each server exposes 3 components. MCP tools wrap the paper’s methods as executable functions: MCP resources hold the manuscript, code links, datasets and figures. MCP prompts encode multi-step workflows, such as the correct Scanpy preprocessing order. The research team used Claude Sonnet 4 for all Paper2Agent applications.
每个服务器暴露 3 个组件。MCP 工具将论文的方法封装为可执行函数；MCP 资源包含手稿、代码链接、数据集和图表。MCP 提示词编码了多步工作流，例如正确的 Scanpy 预处理顺序。研究团队在所有 Paper2Agent 应用中均使用了 Claude Sonnet 4。
Interactive Explainer
交互式解释器
AlphaGenome Agent Results
AlphaGenome 代理结果
For AlphaGenome, Paper2Agent built 22 tools in about 45 minutes for US $14. All 22 passed validation without human intervention. The team compared the agent with Claude Code plus repository access (Claude + Repo) and Biomni.
对于 AlphaGenome，Paper2Agent 在约 45 分钟内构建了 22 个工具，花费 14 美元。所有 22 个工具均在无人干预的情况下通过了验证。团队将该代理与 Claude Code 加上仓库访问权限（Claude + Repo）以及 Biomni 进行了比较。
| Benchmark | Paper2Agent | Claude + Repo | Biomni |
|---|---|---|---|
| 15 tutorial-derived queries | 98.7 ± 1.3% | 82.7 ± 3.4% | 37.3 ± 4.0% |
| 15 novel queries | 100.0 ± 0.0% | 78.7 ± 4.4% | 56.0 ± 3.4% |
| 30 open-ended queries | 82.7 ± 2.4% | 56.7 ± 2.3% | 72.2 ± 2.2% |
| 基准测试 | Paper2Agent | Claude + Repo | Biomni |
|---|---|---|---|
| 15 个源自教程的查询 | 98.7 ± 1.3% | 82.7 ± 3.4% | 37.3 ± 4.0% |
| 15 个新颖查询 | 100.0 ± 0.0% | 78.7 ± 4.4% | 56.0 ± 3.4% |
| 30 个开放式查询 | 82.7 ± 2.4% | 56.7 ± 2.3% | 72.2 ± 2.2% |
Results span 5 runs, graded by 2 human experts with 96.7% inter-rater agreement. On tutorial queries, median runtime fell 1.9× versus Claude + Repo and 3.1× versus Biomni. The gains persisted when the baseline was upgraded to Claude Opus 4.6.
结果基于 5 次运行，由 2 位人类专家评分，评分者间一致性为 96.7%。在教程查询中，与 Claude + Repo 相比，中位运行时间缩短了 1.9 倍，与 Biomni 相比缩短了 3.1 倍。当基线升级为 Claude Opus 4.6 时，这些优势依然保持。
The agent also re-examined an LDL cholesterol variant, chr1:109274968:G>T. It ranked SORT1 as the likely causal gene. The original AlphaGenome paper emphasized CELSR2 and PSRC1. GTEx shows significant liver eQTLs for all 3 genes. The research team say this shows how hard causal gene assignment is at such loci.
该代理还重新检查了 LDL 胆固醇变异体 chr1:109274968:G>T。它将 SORT1 列为最可能的因果基因。原始的 AlphaGenome 论文强调了 CELSR2 和 PSRC1。GTEx 显示这 3 个基因在肝脏中均存在显著的 eQTL。研究团队表示，这表明在此类位点进行因果基因分配的困难程度。
Scanpy, TISSUE and Scale Tests
Scanpy、TISSUE 和 Scale Tests
The Scanpy agent received 7 validated tools in about 45 minutes for US $13. On 4 public datasets, it matched human researchers on cell counts, gene counts and top marker genes. A TISSUE agent reproduced human results on spatial transcriptomics data.
Scanpy 代理在约 45 分钟内获得了 7 个经过验证的工具，费用为 13 美元。在 4 个公共数据集上，它在细胞计数、基因计数和顶级标记基因方面与人类研究人员的表现相当。一个 TISSUE 代理在空间转录组数据上重现了人类的结果。
Scale tests covered 3 corpora with no manual cleanup:
规模测试涵盖了 3 个无需人工清理的语料库：
- 100 bioRxiv computational biology papers: 74 were agentified, and 593 of 599 proposed tools passed validation.
- 300 questions: Paper2Agent scored 91.2%, versus 80.3% (Sonnet 4) and 86.3% (Sonnet 4.6) for Claude + Repo.
- Cost per query: US $0.20 and 1.6 minutes, compared with US $0.38 and 4.3 minutes.
- 10 non-biology papers, including TabPFN, SAM 2 and SAELens: 98.1% accuracy on 42 execution tasks.
- 26 data-focused papers: resource layer 89.0% versus 82.0% for browser use, 34× cheaper and 15× faster.
- 100 篇 bioRxiv 计算生物学论文：其中 74 篇被转化为代理，599 个提出的工具中有 593 个通过了验证。
- 300 个问题：Paper2Agent 得分为 91.2%，而 Claude + Repo 的得分分别为 Sonnet 4 的 80.3% 和 Sonnet 4.6 的 86.3%。
- 每次查询成本：0.20 美元和 1.6 分钟，相比之下为 0.38 美元和 4.3 分钟。
- 10 篇非生物学论文，包括 TabPFN、SAM 2 和 SAELens：在 42 个执行任务上的准确率为 98.1%。
- 26 篇以数据为主的论文：资源层准确率为 89.0%，而浏览器使用为 82.0%，成本低 34 倍，速度快 15 倍。
Paper2Agent also rejected 100% of out-of-scope queries in a permuted benchmark. It recovered from injected dependency, file-path, typo and deprecated API failures.
Paper2Agent 在一个置换基准测试中也拒绝了 100% 超出范围的查询。它从注入的依赖关系、文件路径、拼写错误和已弃用 API 失败中恢复过来。
Paper Agents Collaborating
论文代理协作
The research team connected 3 agents: AlphaGenome, an MPRA-coupled scCRISPRi screen and a CD4+ T cell Perturb-seq dataset. AlphaGenome flagged GPR137 at psoriasis locus rs887314, with an RNA-seq quantile score of 0.997. The AI co-scientist proposed 10 validation strategies, and a researcher picked signature correlation.
研究团队连接了 3 个代理：AlphaGenome、一个耦合 MPRA 的 scCRISPRi 筛选以及一个 CD4+ T 细胞 Perturb-seq 数据集。AlphaGenome 在银屑病位点 rs887314 处标记了 GPR137，其 RNA-seq 分位数得分为 0.997。AI 共同科学家提出了 10 种验证策略，研究人员选择了签名相关性。
Only GPR137 knockdown matched the CRE perturbation signature. The match appeared under stimulation: Spearman 0.613 at Stim8hr and 0.630 at Stim48hr. BAD and 3 other candidates showed no significant correlation. A second study paired AlphaGenome with an ADHD GWAS and nominated rs1626703 among 209 candidates. That hypothesis still needs experimental validation.
只有 GPR137 敲低与 CRE 扰动签名相匹配。这种匹配出现在刺激条件下：Stim8hr 时 Spearman 相关系数为 0.613，Stim48hr 时为 0.630。BAD 和其他 3 个候选者显示出无显著相关性。第二项研究将 AlphaGenome 与 ADHD GWAS 配对，并在 209 个候选者中提名了 rs1626703。该假设仍需实验验证。
Key Takeaways
关键要点
- Paper2Agent converts papers and repos into tested MCP servers with tools, resources and prompts.
- The AlphaGenome agent took about 45 minutes, cost US $14, and scored 100% on novel queries.
- 74 of 100 bioRxiv papers were agentified, with 593 of 599 tools validated.
- 3 paper agents jointly supported GPR137 as the probable psoriasis causal gene.
- The code is MIT-licensed, with prebuilt MCP servers on Hugging Face Spaces.
- Paper2Agent 将论文和仓库转换为带有工具、资源和提示词的可测试 MCP 服务器。
- AlphaGenome 代理耗时约 45 分钟，花费 14 美元，在新颖查询中的得分为 100%。
- 100篇bioRxiv论文中有74篇被Agent化，599个工具中有593个得到验证。
- 3个论文Agent共同支持GPR137作为可能的银屑病致病基因。
- 代码采用MIT许可证，Hugging Face Spaces上提供预构建的MCP服务器。
Check out the Paper and Repo. All credit goes to the researcher of this project. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.
查看论文和仓库。所有功劳归于本项目的研究者。也欢迎在Twitter上关注我们，别忘了加入我们拥有15万+成员的ML SubReddit，并订阅我们的新闻通讯。等等！你在Telegram上吗？现在你也可以加入我们的Telegram群组。
Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us
需要与我们合作推广你的GitHub仓库、Hugging Face页面、产品发布或网络研讨会等？请联系我们
The post Stanford Researchers Release Paper2Agent: Turning Research Papers Into AI Agents That Reproduce Results and Run on New Data appeared first on MarkTechPost.
本文最初发表于MarkTechPost：斯坦福研究人员发布Paper2Agent：将研究论文转化为可复现结果并在新数据上运行的AI Agent。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力