以程式碼為世界：智能體自主發現可執行的世界表徵以進行物理推理

摘要

物理理解與推理依賴於對世界形成緊湊且可泛化的表徵。儘管現代的視覺-語言模型能夠辨識並解釋多樣化的物理事件，但它們往往缺乏對底層機制的顯式表徵——例如物體狀態、物理參數以及支配動力學——而這些正是可靠推理世界如何演化以及如何回應干預所需的。在本工作中，我們提出「代碼即世界」（Code-as-World），一種透過可執行的世界表徵來表示物理世界的範式。透過將物理組成、動態演化與視覺外觀表達為可執行程式碼，Code-as-World 提供了對物理世界的一種緊湊、具量化基礎且可控的抽象。為了從多模態觀測（例如自然語言描述或真實世界影片）中建構此類表徵，我們開發了一種受溯因推理啟發的智能體驅動發現迴圈，其中智能體提出、執行、渲染、驗證並迭代精煉可執行的世界假設。作為一項具體應用，我們使用經過驗證的可執行世界來提供可擴展的物理監督，以訓練視覺-語言模型進行定量物理推理。實驗表明，Code-as-World-VL 在 QuantiPhy 上達到了最先進的性能，並超越了領先的專有模型，凸顯了可執行世界表徵作為物理智慧可擴展基礎的潛力。

English

Physical understanding and reasoning depend on forming compact and generalizable representations of the world. While modern vision-language models can recognize and explain diverse physical events, they often lack explicit representations of the underlying mechanisms-such as object states, physical parameters, and governing dynamics-needed for reliably reasoning how the world evolves and responds to interventions. In this work, we introduce Code-as-World, a paradigm that represents physical worlds through executable world representations. By expressing physical composition, dynamic evolution, and visual appearance as executable code, Code-as-World provides a compact, quantitatively grounded, and controllable abstraction of the physical world. To construct such representations from multimodal observations, such as natural-language descriptions or real-world videos, we develop an agentic discovery loop inspired by abductive reasoning, where an agent proposes, executes, renders, verifies, and iteratively refines executable world hypotheses. As a concrete application, we use verified executable worlds to provide scalable physical supervision for training vision-language models on quantitative physical reasoning. Experiments show that Code-as-World-VL achieves state-of-the-art performance on QuantiPhy and surpasses leading proprietary models, highlighting the potential of executable world representations as a scalable foundation for physical intelligence.