CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference
- ID: ad830db9
- 原文链接: https://arxiv.org/abs/2607.22511
- PDF: https://arxiv.org/pdf/2607.22511v1
- 作者: Jiyuan Tan, Vasilis Syrgkanis
- 日期: 2026-07-24
- 更新: 2026-07-24
- 分类: agents
- 来源类型: article
- 标签: automated-research, lean, formal-verification, causal-inference, self-improving
- 质量评分: 4/5
- 抓取时间: 2026-07-29T06:19:39Z
中文导读
CausalForge结合了Causalean（一个包含7035条机器检查声明的Lean因果推断证明库）和CausalSmith（一个自我改进的Agent流水线，可选题提出结果形式化陈述构造证明并提交人类审查）因为机器检查证明只能验证形式陈述与假设一致，无法确认其是否真实表达科学主张，该系统在核验之外增加了陈述审计环节研究指出LLM审稿人对自动研究产出的评估并不可靠，识别伪造论文的概率接近随机
为什么值得关注
CausalForge结合了Causalean（一个包含7035条机器检查声明的Lean因果推断证明库）和CausalSmith（一个自我改进的Agent流水线，可选题提出结果形式化陈述构造证明并提交人类审查）因为机器检查证明只能验证形式陈述与假设一致...
本文以 arXiv 摘要与条目既有摘要为据：程序性技能/自进化/形式化研究/路由等主题直接对应 Agent 工程与评估实践中的可靠性、可验证性与部署权衡。
关键信息
- 论文标题：CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference
- 作者：Jiyuan Tan, Vasilis Syrgkanis
- arXiv：https://arxiv.org/abs/2607.22511
- 发布时间：2026-07-24
- 更新时间：2026-07-24
- arXiv 分类：stat.ML, cs.AI, cs.LG, econ.EM
- 关联标签：automated-research, lean, formal-verification, causal-inference, self-improving
English Abstract
Automating theoretical research is constrained not only by the generation of candidate results, but also by their reliable evaluation. A common approach is to close the research loop with a large language model (LLM) reviewer. However, such reviewers remain empirically unreliable: they may accept fabricated papers and detect them at rates close to chance (Bad Scientist, 2025). We present CausalForge, a framework for automated theoretical research in causal inference grounded in the Lean proof assistant. CausalForge combines Causalean, a foundational Lean library for causal inference containing 7,035 machine-checked declarations developed with language-model assistance under human design and review, with CausalSmith, a self-improving agentic pipeline that selects research topics, proposes results, formalizes statements, constructs proofs, and presents the resulting artifacts for human inspection. Because a machine-checked proof establishes only that a formal statement follows from its assumptions, not that the statement faithfully captures the intended scientific claim, the pipeline augments kernel verification with a statement audit that compares each formal theorem against the informal claim it is intended to express. We evaluate the system using artifacts produced by completed autonomous research runs. The source code, formal library, and run records are available at https://github.com/Jiyuan-Tan/CausalForge.
English Summary
CausalForge combines Causalean (a Lean proof assistant library with 7,035 machine-checked declarations for causal inference) with CausalSmith (a self-improving agentic pipeline that selects topics, proposes results, formalizes statements, constructs proofs, and presents artifacts for human inspection). Because machine-checked proofs only verify that a statement follows from assumptions--not that it captures the intended scientific claim--the pipeline augments kernel verification with a statement audit comparing formal theorems against their informal claims. The work highlights that LLM reviewers alone are unreliable for evaluating automated research, detecting fabricated papers at rates near chance.
Obsidian Notes
- 内容由
opencli arxiv paper拉取 arXiv 元数据与摘要生成。 - 中文导读与价值判断均锚定在条目已有摘要、论文摘要、作者、日期与分类信息上；未补充论文摘要之外的实验细节。