跳转至

文章背景与核心概要

使科学研究实现可计算化，需要在机器可读的科学知识与检验实验断言的物理世界之间架起一座桥梁。本文引入了一种新型的物理实验室可计算表示（computable representation of the physical laboratory），该框架构建于定型研究对象（typed research objects）、能力受限的操作（capability-bound operations）以及组合式工作流代数（compositional workflow algebra）的基础之上。

该框架的核心特性包括：充当机器可读知识的物理世界对应物、支持显式依赖与并发的高级工作流逻辑、通过绑定可执行函数技能部署在模块化智能机器人实验室中，以及实现有状态的模拟与验证。最终，这一表示与工程框架建立了一个通用的计算接口，将智能体的推理能力与受限于能力的物理转化过程联系起来，为端到端的自主科学发现铺平了道路。

A computable representation of the physical laboratory enables verifiable workflows

arXiv: 2609.03621 [cs.AI]
Submitted: September 3, 2026
Primary Subject: Artificial Intelligence (cs.AI)
Authors: Xiaobo Li, Luyao Ge, Xiaohui Li, Lulu Guo, Ming Mao, Jiwang Zheng, Wenting Guan, Xin Yang, Yi Luo, Jun Jiang, Linjiang Chen

arXiv: 2609.03621 [cs.AI]
Submitted: September 3, 2026
Primary Subject: Artificial Intelligence (cs.AI)
Authors: Xiaobo Li, Luyao Ge, Xiaohui Li, Lulu Guo, Ming Mao, Jiwang Zheng, Wenting Guan, Xin Yang, Yi Luo, Jun Jiang, Linjiang Chen

Summary

Making science computable requires bridging the gap between machine-readable scientific knowledge and the physical world where experimental claims are tested. This paper introduces a novel computable representation of the physical laboratory, built using typed research objects, capability-bound operations, and a compositional workflow algebra.

Key features of this framework include: * Physical-World Counterpart: Acts as a bridge for machine-readable knowledge, translating workflows into programs that operate over evolving laboratory states. * Advanced Workflow Logic: Supports explicit dependencies, decisions, iteration, and concurrency. * Modular Implementation: Deployed in a modular agentic robotic laboratory by binding formal operations to executable Function Skills. * Stateful Simulation & Verification: Dynamically propagates object transformations and verifies operation preconditions and laboratory constraints before physical dispatch.

Ultimately, this representation and engineering framework establish a general computational interface that links agent reasoning with capability-bound physical transformations, paving the way for end-to-end autonomous scientific discovery.

Making science computable requires bridging the gap between machine-readable scientific knowledge and the physical world where experimental claims are tested. This paper introduces a novel computable representation of the physical laboratory, built using typed research objects, capability-bound operations, and a compositional workflow algebra.

Key features of this framework include: * Physical-World Counterpart: Acts as a bridge for machine-readable knowledge, translating workflows into programs that operate over evolving laboratory states. * Advanced Workflow Logic: Supports explicit dependencies, decisions, iteration, and concurrency. * Modular Implementation: Deployed in a modular agentic robotic laboratory by binding formal operations to executable Function Skills. * Stateful Simulation & Verification: Dynamically propagates object transformations and verifies operation preconditions and laboratory constraints before physical dispatch.

Ultimately, this representation and engineering framework establish a general computational interface that links agent reasoning with capability-bound physical transformations, paving the way for end-to-end autonomous scientific discovery.

References & Citations

References & Citations