Code World Model: Coding Agent as World Brain

Hugging Face Daily Papers Papers
world-model coding-agent language-model video-model executable-code fine-tuning

Summary

The paper introduces Code World Model, a framework that combines language models for reasoning and coding with video models for visual rendering to simulate persistent world evolution in an open-ended manner.

World models aim to simulate how complex environments evolve under actions and events, yet existing video-based world models primarily learn dynamics from visual observations, which reveal outcomes rather than the underlying knowledge, rules, and mechanisms governing world evolution. This makes it difficult to maintain persistent consequences and support coherent, open-ended evolution. We introduce Code World Model, a framework that separates world evolution from visual realization by combining the reasoning and coding capabilities of language models with the generative priors of video models. A coding agent serves as the world brain, reasoning about events and their consequences and generating executable code to maintain persistent world state and perform rule-consistent evolution. To connect executable state with visual generation, we introduce a proxy representation that encodes frame-wise spatiotemporal constraints and is compiled into a proxy video, which conditions a video model to render high-fidelity visual observations. We further develop data pipelines for constructing aligned proxy-observation pairs from gameplay and real-world videos. After fine-tuning on paired gameplay data, MiniMax-H3 follows proxy-based spatiotemporal specifications from simple interactive worlds built by the coding agent while preserving rich visual details and dynamics. These results demonstrate the potential of combining code for persistent world evolution with video models for flexible visual realization, providing a new path toward open-ended world models.
Original Article
View Cached Full Text

Cached at: 08/27/26, 03:18 AM

Paper page - Code World Model: Coding Agent as World Brain

Source: https://huggingface.co/papers/2608.25927

Abstract

Code World Model separates persistent world dynamics from visual rendering by using a language model to generate executable state updates and a video model to render observations from proxy representations.

World modelsaim to simulate how complex environments evolve under actions and events, yet existingvideo-based world modelsprimarily learn dynamics from visual observations, which reveal outcomes rather than the underlying knowledge, rules, and mechanisms governing world evolution. This makes it difficult to maintain persistent consequences and support coherent, open-ended evolution. We introduce Code World Model, a framework that separates world evolution from visual realization by combining the reasoning and coding capabilities oflanguage modelswith the generative priors ofvideo models. Acoding agentserves as the world brain, reasoning about events and their consequences and generatingexecutable codeto maintain persistent world state and perform rule-consistent evolution. To connect executable state with visual generation, we introduce aproxy representationthat encodes frame-wisespatiotemporal constraintsand is compiled into a proxy video, which conditions avideo modelto render high-fidelity visual observations. We further develop data pipelines for constructing aligned proxy-observation pairs from gameplay and real-world videos. Afterfine-tuningon paired gameplay data, MiniMax-H3 follows proxy-based spatiotemporal specifications from simple interactive worlds built by thecoding agentwhile preserving rich visual details and dynamics. These results demonstrate the potential of combining code for persistent world evolution withvideo models for flexible visual realization, providing a new path towardopen-ended world models.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2608\.25927

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.25927 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.25927 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.25927 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Coding Agent Is Good As World Simulator

arXiv cs.AI

This paper presents an agentic framework that uses coding agents to generate physically plausible world simulations from natural language prompts, outperforming video-based models in physical accuracy and instruction fidelity.

Qwen-AgentWorld: Language World Models for General Agents

Hacker News Top

Qwen-AgentWorld introduces language world models for agentic environments, covering seven domains with long chain-of-thought reasoning. The work includes a new benchmark, AgentWorldBench, and shows that world modeling improves downstream agent performance.

Quo Vadis, World Modeling?

Hugging Face Daily Papers

This paper proposes an agent-centric paradigm for world modeling, introducing 'Agent-Centric Interactive World Proxies' and organizing them into six functional forms across three progressive levels of agent improvement, offering a roadmap for building world proxies that empower agents to plan, learn, and evolve.

Coding agents think ahead of time

Hacker News Top

This paper investigates how language models under coding agents internally represent the evolving program during iterative editing. The authors find that linear probes can decode program properties (e.g., parsing, test pass rate) from residual streams, and surprisingly, these representations predict future outcomes before the agent materializes edits, revealing a 'latent programming horizon'.