Papers
arxiv:2608.27549

Code as Worlds: Agentic Discovery of Executable World Representations for Physical Reasoning

Published on Aug 27
· Submitted by
Jialong Wu
on Aug 31
Authors:
,
,
,
,
,
,
,
,
,
,
,
,
,
,

Abstract

Code-as-World represents physical environments as executable code to enable quantitative reasoning and scalable supervision for vision-language models.

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.

Community

Paper author Paper submitter

Pixels are evidence of the physical world, not its ontology. A pixel-level observation records how the world appears at a particular moment and from a particular viewpoint, but does not directly specify what exists within it, how it is structured, or what governs its evolution. Code-as-World introduces code as executable representations for the physical world and an agentic process for discovering them through iterative simulation and verification. In doing so, it turns raw abundant observations into reusable physical data: explicit states, dynamics, and mechanisms that capture not only what was seen, but the underlying world that could have produced it. This provides scalable physical supervision, enabling our models to achieve state-of-the-art performance on quantitative physical reasoning.

Sign up or log in to comment

Models citing this paper 4

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2608.27549 in a dataset README.md to link it from this page.

Spaces citing this paper 1

Collections including this paper 1