推出 Science One 框架,这是一个实验性研究原型,旨在通过原生构建可验证的证据链来消除模型幻觉;同时推出 CoE Audit,这是一套用于评估 AI 生成论文完整性的自动化协议。
快速链接
大语言模型(LLM)正越来越多地不仅被用作编码助手,还被用作能够执行端到端科学研究工作流的自主智能体。近期的系统(例如 Sakana 的 AI-Scientist、AutoResearchClaw、DeepScientist、AI-Researcher)能够综述文献、提出假设、执行实验,并撰写可与人类撰写论文相媲美的完整稿件。然而,随着这些 AI 生成稿件表面质量的提升,一个关键的结构性问题浮现出来:可验证性。由于当前的自主研究流水线以迭代方式生成文本,任何阶段引入的错误都会被放大。一些现有系统会生成不存在的引用,表现出所描述方法与实际代码之间的不一致,并报告无法从所提供代码中完全复现的实验分数。
在我们的论文中,我们通过引入证据链(Chain-of-Evidence,CoE)来应对这一问题,这是一个面向 AI 驱动研究的全新可验证性框架。我们以 Science One Framework 和 CoE Audit 来实例化 CoE:前者是一个自主研究原型,能够原生地构建并维护证据链;后者是一套自动化评估指标,用于衡量 AI 生成论文相对于其底层代码和证据的完整性。我们的结果表明,基线系统会幻觉出高达 21% 的参考文献,并且经常出现代码与文本不一致的情况,而 Science One Framework 则实现了零幽灵参考文献和完全可验证的得分,同时在 MLE-Bench 和 Parameter-Golf 等前沿基准上达到了 SOTA 性能。
证据链:一个面向可验证研究的框架
CoE 是一个概念框架,它定义了什么样的研究成果才值得信赖,正如 ACID 定义了什么样的数据库事务才是可靠的。该框架并不规定如何构建研究智能体,而是规定其输出必须具备哪些属性。它遵循一条原则,包含两个方面:研究成果中的每一项主张都必须携带一条有记录的证据链(完整性),并且每条证据链都必须真正支持其所附着的该项主张(正确性)。一项主张可以是一条参考文献、一个报告的数字、一段方法描述或一个结论,它必须链接回相应的证据,例如一篇同行评审论文、一行实验日志、实际运行的代码或结果表格。
一条幻觉参考文献指向了一篇并不存在的论文。一个无法复现的分数在代码重新运行后并未再次出现。一段描述有误的方法声称论文中采用某种算法,而代码实现的却是另一种。每一项都是其证据链条断裂的主张;CoE Audit 让这些断裂变得可测量。
Science One Framework
为了证明可验证的 AI 研究能够在不牺牲问题求解性能的前提下实现,我们设计了 Science One Framework。与以往那些先生成论文、再试图回溯性地关联事实的智能体不同,Science One Framework 通过三个主要模块在构建之初就实例化了 CoE 框架:
- 问题调查器(文献接地):为防止幻觉参考文献,Science One Framework 通过 Semantic Scholar API 构建引文图谱。它针对每个主题读取最多 100 篇全文 PDF,以生成结构化的研究简报。最终论文中的每一条参考文献都源自这一接地的 API 调用,从而完全消除了对模型记忆的依赖。
- 发现引擎(并行探索-利用):Science One Framework 在多个并行分支上系统地探索和利用想法。在每个隔离的循环中,一个 Solver 智能体实现一个解决方案,一个针对具体任务的评估器为其打分。表现优异的分支会被迭代优化,所有评估器的原始输出都会被汇编成一份严格的、只读的记录。
- 论文撰写器与论断验证器:在生成稿件之前,Science One Framework 会为每一条事实性论断构建结构化表示,并附带一个内联证据标签,将其绑定到特定的工作区产物。一个专门的论断验证器会对照每条论断所声明的来源进行核查。对于超出其证据支撑范围的论断,会与来源进行调和——以更保守的方式重新表述,而非直接删除,从而使论文与工作本身所能支撑的内容保持一致。

Science One Framework 流水线。问题调查器通过检索到的 PDF 为文献提供依据。发现模块探索并评估解决方案。论文撰写与验证模块撰写并验证论文,由论断验证器确保所有论断与其证据来源相符。
CoE Audit:衡量可验证性
为了严格评估 Science One Framework 原型相对于最先进基线(例如 Sakana AI 的 AI Scientist v2、AutoResearchClaw、DeepScientist、AI-Researcher)的表现,我们开发了 CoE Audit。这一事后评估协议充当自动化取证审查员,对生成的产物(论文、解决方案、代码和参考文献)运行四项严格的完整性检查:
- 分数验证:从论文中提取所报告的分数,并将其与对提交代码进行的完全独立重跑结果进行比较。
- 规范违反: 检查解决方案代码,确保它真正解决了任务,而不是利用评估指标漏洞或读取真实答案文件。
- 参考文献验证: 将每一条参考文献条目与学术 API 进行交叉核对,以捕捉不存在的幽灵参考文献。
- 方法-代码一致性: 使用 LLM 评审员将论文的方法部分与代码进行并排比较,以确保文本忠实描述了所实现的算法。

CoE Audit 框架及其四项核心诚信检查概述。
结果
我们将 CoE Audit 应用于五个系统优化任务(Prism、Cloudcast、EPLB、LLM-SQL 和 事务调度)中生成的 75 篇论文,这些任务来自 Automated Design of Research Systems(ADRS)基准。
Science One 框架在可验证性方面显著优于现有基线。CoE Audit 对每个系统应用同一套独立协议,对照实时学术数据库重新核查每一条参考文献,在该协议下,Science One 框架在全部四项诚信检查中均处于领先。
它的参考文献没有一条是虚构的:每一条都指向一篇真实、可检索的论文,而基线的模型幻觉率高达 21%,因为 Problem Investigator 会检索每一条参考文献,而不是凭记忆生成。它还实现了完美的评分验证,以及最高的方法—代码一致性。
相比之下,基线系统常常声称使用了复杂算法(如“混合神经符号求解器”),而它们提交的代码却只是一个简单、确定性的启发式方法。

五个系统的 CoE Audit 结果。
关键在于,实施严格的可验证性并未损害该智能体的科学能力。Science One 框架在全部五项 ADRS 任务上达到或超过了人类专家表现,并在其中两项任务(Cloudcast 和 EPLB)上取得了所有系统中的最佳总分。

五个 MLE-Bench 任务及 Parameter Golf 上求解器性能的对比。
为测试其泛化能力,我们将 Science One 框架部署到六项高度复杂的外部任务上:
- MLE-Bench:在涵盖医学影像、细粒度识别和 3D 感知的五项高难度 Kaggle 竞赛中,Science One Framework 斩获两枚金牌(其中包括在基线系统完全失败的 3D 目标检测任务中取得获胜分数)和两枚银牌。
- Parameter-Golf:我们在一个对硬件和文件大小有严格限制的实时 LLM 训练竞赛中测试了 Science One Framework。基线系统未能产出有效提交,而 Science One Framework 成功遵守了所有约束,并取得了最先进的分数(截至 2026 年 4 月 27 日)。值得注意的是,Science One Framework 发现的是真正且新颖的算法技术,而不仅仅是调整表面的超参数。
展望未来
随着自主研究系统规模不断扩大,以解决日益困难的科学问题,仅靠求解器质量将不再足以区分它们。真正区分其产出的是,所得到的研究是否可信。我们的发现表明,可验证性必须被视为一等架构约束。通过在主张产生之时构建证据链,而不是事后试图重建依据,The Science One Framework 证明 AI 智能体能够产出严谨、可信且极具竞争力的科学研究。
我们希望 Chain-of-Evidence 框架及其审计能够成为社区的有价值工具,帮助我们继续构建下一代 AI 科学家。
致谢
我们要感谢 Bhavana Dalvi Mishra、Jiefeng Chen、Chun-Liang Li、Palash Goyal、Mihir Parmar、Yiwen Song、Yale Song、Raj Sinha、Parthasarathy Ranganathan、Burak Gokturk 和 Jinsung Yoon 对本项工作做出的宝贵贡献。
免责声明
Science One Framework 是一个实验性研究原型,并非可用于生产的工具。
Introducing the Science One Framework, an experimental research prototype designed to eliminate hallucinations by natively building verifiable evidence chains, and CoE Audit, an automated protocol to evaluate the integrity of AI-generated papers.
Quick links
Large language models (LLMs) are increasingly being deployed not just as coding assistants but as autonomous agents capable of conducting end-to-end scientific research workflows. Recent systems (e.g., Sakana’s AI-Scientist, AutoResearchClaw, DeepScientist, AI-Researcher) can review literature, formulate hypotheses, execute experiments and write complete manuscripts that are comparable to human-authored papers. However, as the surface-level quality of these AI-generated manuscripts improves, a critical structural problem has emerged: verifiability. Because current autonomous research pipelines generate text iteratively, errors introduced at any stage are amplified. Some existing systems can generate non-existent citations, exhibit misalignments between the described methods and the actual code, and report experimental scores that are not fully reproducible from the provided code.
In our paper, we tackle this problem by introducing Chain-of-Evidence (CoE), a new verifiability framework for AI-driven research. We instantiate CoE with the Science One Framework, an autonomous research prototype that natively builds and maintains evidence chains, and the CoE Audit, a set of automated evaluation metrics that measures the integrity of AI-generated papers against their underlying code and evidence. Our results show that baseline systems hallucinate up to 21% of their references and frequently misalign their code and text, whereas the Science One Framework achieves zero phantom references and fully verifiable scores while achieving state-of-the-art performance on frontier benchmarks like MLE-Bench and Parameter-Golf.
Chain-of-Evidence: A framework for verifiable research
The CoE is a conceptual framework that defines what makes a research artifact trustworthy, much as ACID defines what makes a database transaction reliable. Rather than prescribing how to build a research agent, the framework specifies the properties its outputs must have. It follows a single principle with two halves: every claim in a research artifact must carry a recorded evidence chain (completeness), and each chain must genuinely support the claim it is attached to (correctness). A claim may be a reference, a reported number, a method description, or a conclusion, that must link back to corresponding evidence, such as a peer-reviewed paper, an experimental log line, the code that actually ran, or the results table.
A hallucinated reference points to a paper that does not exist. An unreproducible score does not reappear when the code is re-run. A misdescribed method claims one algorithm in the paper while the code implements another. Each is a claim whose chain back to its evidence is broken; the CoE Audit makes these breaks measurable.
The Science One Framework
To demonstrate that verifiable AI research is possible without sacrificing problem-solving performance, we designed the Science One Framework. Unlike previous agents that generate a paper and try to link facts retroactively, the Science One Framework instantiates the CoE framework by construction through three main modules:
- Problem investigator (literature grounding): To prevent hallucinated references, the Science One Framework builds a citation graph via the Semantic Scholar API. It reads up to 100 full-text PDFs per topic to produce a structured research brief. Every reference in the final paper originates from this grounded API call entirely eliminating reliance on model memory.
- Discovery engine (parallel explore-exploit): The Science One Framework systematically explores and exploits ideas across multiple parallel branches. In each isolated cycle, a Solver agent implements a solution and a task-specific evaluator scores it. High-performing branches are iteratively refined, and all raw evaluator outputs are compiled into a strict, read-only record.
- Paper writer and claim verifier: Before rendering the manuscript, the Science One Framework builds a structured representation of every factual claim with an inline evidence tag binding it to a specific workspace artifact. A dedicated Claim Verifier checks every claim against its declared source. Claims that outrun their evidence are reconciled with the source — restated conservatively rather than removed, keeping the paper aligned with what the work supports.

The Science One Framework pipeline. Problem Investigator grounds literature via retrieved PDFs. Discovery module explores and evaluates solutions. Paper writing & verification module writes and verifies the paper with a Claim Verifier ensuring all claims match their evidence source.
CoE Audit: Measuring verifiability
To rigorously evaluate the Science One Framework prototype against state-of-the-art baselines (e.g., Sakana AI's AI Scientist v2, AutoResearchClaw, DeepScientist, AI-Researcher), we developed the CoE Audit. This post-hoc evaluation protocol acts as an automated forensic reviewer, running four strict integrity checks on the generated artifacts (paper, solution, code and references):
- Score verification: Extracts the reported score from the paper and compares it against a completely independent re-run of the submitted code.
- Specification violation: Inspects the solution code to ensure it actually solves the task rather than exploiting the evaluator metric or reading ground-truth answer files.
- Reference verification: Cross-checks every bibliography entry against academic APIs to catch non-existent phantom references.
- Method-code alignment: Uses LLM judges to compare the paper's method section side-by-side with the code to ensure the text faithfully describes the implemented algorithm.

Overview of the CoE Audit framework and its four core integrity checks.
Results
We applied the CoE Audit to 75 papers generated across five systems-optimization tasks (Prism, Cloudcast, EPLB, LLM-SQL, and transaction scheduling) from the Automated Design of Research Systems (ADRS) benchmark.
The Science One Framework significantly outperformed existing baselines in verifiability. The CoE Audit applies the same independent protocol to every system, re-checking each reference against live scholarly databases, and under it, Science One Framework led on all four integrity checks. None of its references were phantom: every one pointed to a real, retrievable paper, compared to hallucination rates as high as 21% in baselines, because the Problem Investigator retrieves every reference rather than generating it from memory. It also achieved perfect score verification and the highest method-code alignment. In contrast, baseline systems frequently described sophisticated algorithms (like "hybrid neuro-symbolic solvers") when their submitted code was a simple, deterministic heuristic.

CoE Audit results across five systems.
Crucially, implementing strict verifiability did not compromise the agent's scientific capabilities. The Science One Framework matched or exceeded human expert performance on all five ADRS tasks, achieving the best overall score among all systems on two of them (Cloudcast and EPLB).

Comparison of solver performance across five MLE-Bench tasks and Parameter Golf.
To test its generalization, we deployed the Science One Framework on six highly complex external tasks:
- MLE-Bench: Across five difficult Kaggle competitions spanning medical imaging, fine-grained recognition and 3D perception, Science One Framework achieved two Gold Medals (including a winning score on 3D Object Detection where baselines failed entirely) and two Silver Medals.
- Parameter-Golf: We tested Science One Framework in a live LLM-training competition with strict hardware and file-size constraints. While baseline systems failed to produce valid submissions, Science One Framework successfully adhered to all constraints and achieved a state-of-the-art score (as of April 27, 2026). Note that Science One Framework discovers genuine and novel algorithmic techniques rather than just tweaking superficial hyperparameters.
Looking forward
As autonomous research systems scale to solve increasingly difficult scientific problems, solver quality alone will no longer be enough to differentiate them. What will separate their outputs is whether the resulting research can be trusted. Our findings demonstrate that verifiability must be treated as a first-class architectural constraint. By building evidence chains at the time a claim is produced rather than attempting to reconstruct grounding after the fact, The Science One Framework demonstrates that AI agents can produce rigorous, trustworthy and highly competitive scientific research. We hope the Chain-of-Evidence framework and its audit will serve as valuable tools for the community as we continue to build the next generation of AI scientists.
Acknowledgements
We would like to thank Bhavana Dalvi Mishra, Jiefeng Chen, Chun-Liang Li, Palash Goyal, Mihir Parmar, Yiwen Song, Yale Song, Raj Sinha, Parthasarathy Ranganathan, Burak Gokturk and Jinsung Yoon for their valuable contributions to this work.
Disclaimer
The Science One Framework is an experimental research prototype, not production-ready tools.