Video World Models · 2026-10-05
World Editing: Intervening on Executable Worlds at Increasing Depth
Max Ku, Nok-Kan Law, Yu-Chien Tang, Shih-Ying Yeh, Ping Nie, Andy Zheng, Tat Hei Lai, Fei-Yueh Chen, Nikko Yu, Wei-Chieh Sun, Suzy Huang, Chiao-Wei Hsu, Chih-Chuan Huang, Chak-Wing Mak, Ho Yin Sam Ng, Edisy Kin Wai Chan, Min-Hung Chen, Ho Kei Cheng
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
The authors define world editing as intervening on an existing executable world while preserving unrelated properties, and introduce intervention depth (property→entity→dynamics→system). They build IGMWorld/IGMBench: 110 Minecraft/Terraria modding tasks (over 1.1K executable state/behavior criteria) with executability, behavioral, preservation, and visual checks. Frontier coding agents reach up to 78.2% task success (strict WSR) and 94.8% criterion pass rate (CPR), with reliability decreasing as depth increases.
Why it matters
World editing is presented as a capability distinct from world generation and gameplay: the goal is to deliberately change an already-executable environment while preserving what should remain unchanged. By grounding evaluation in deterministic build/load/executable behavior plus visual checks, IGMBench/IGMWorld provide a practical way to study how editing reliability changes with intervention depth and where failures occur (mainly after executability, in behavior).
Method
- Formulate world editing as intervening on an existing executable world W to produce W' that realizes requested Δk while preserving unrelated properties; define intervention depth by coupling across property/entity/dynamics/system levels.
- Instantiate with IGMWorld (edit→build→run→observe loop with hidden executable verification) and IGMBench (110 Minecraft/Terraria tasks with ~1.1K state/behavior criteria).
- Evaluate via deterministic executability (build/load gates), world-edit correctness (state/behavior checks + regression preservation checks), and visual consistency (TPIPS-based semantic/style checks).
Limitation
Preservation under deeper interventions is only partially tested because targeted regression checks have limited coverage and are concentrated at L1.
Abstract (from arXiv)
Interactive world models are increasingly capable of generating environments and acting within them, yet deliberately editing an existing executable world remains underexplored. We formulate world editing as intervening on an existing world while preserving properties that should remain unchanged, and introduce intervention depth as an axis describing how strongly an edit couples world entities, dynamics, and systems. We instantiate this capability through industry-grade game modding and introduce IGMWorld, together with IGMBench, a benchmark of 110 tasks and over 1.1K executable state and behavioral criteria across Minecraft and Terraria. The tasks span property, entity, dynamics, and system interventions and are evaluated through deterministic executability, behavioral, preservation, and visual checks. Frontier coding agents already exhibit substantial world-editing capability: the strongest configuration solves 78.2% of tasks under a strict task-level criterion, while criterion-level performance reaches 94.8%. Reliability generally decreases with intervention depth, and this pattern persists even among tasks with similar numbers of evaluation criteria. Most failed edits still build and load successfully, suggesting that the main difficulty is making the edited world behave as requested. Visual consistency remains a separate weakness, with all evaluated configurations below 50% joint visual pass rate. These results show that world editing is a distinct capability from world generation and interaction, and that executable games provide a practical testbed for studying it.
Related papers
- ExeWorldBench: Benchmarking Code-Based Executable 3D World Modeling
- WorldStream: Streaming Interactive 3D Worlds with Hierarchical Planning and Persistent State
- Oneira: From Open-Ended Generation to Open-World Interaction in Video World Models
- MultiWorldBench: Do Independently Controlled Views Describe One Shared World?
- Generative World Simulation with DAO and JING
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-01 · Indexed by Pub-AI: 2026-10-05 UTC
World Editing: Intervening on Executable Worlds at Increasing Depth
Preprint posted: 2026-10-01 · Original record ↗
- arxiv
- 2610.02331
Explicit source versions (1)
- v1 · 2026-10-01
- Preprint available
Source date: 2026-10-01 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Interactive world models are increasingly capable of generating environments and acting within them, yet deliberately editing an existing executable world remains underexplored. We formulate world editing as intervening on an existing world while preserving properties that should remain unchanged, and introduce intervention depth as an axis describing how strongly an edit couples world entities, dynamics, and systems. We instantiate this capability through industry-grade game modding and introduce IGMWorld, together with IGMBench, a benchmark of 110 tasks and over 1.1K executable state and behavioral criteria across Minecraft and Terraria. The tasks span property, entity, dynamics, and system interventions and are evaluated through deterministic executability, behavioral, preservation, and visual checks. Frontier coding agents already exhibit substantial world-editing capability: the strongest configuration solves 78.2% of tasks under a strict task-level criterion, while criterion-level performance reaches 94.8%. Reliability generally decreases with intervention depth, and this pattern persists even among tasks with similar numbers of evaluation criteria. Most failed edits still build and load successfully, suggesting that the main difficulty is making the edited world behave as requested. Visual consistency remains a separate weakness, with all evaluated configurations below 50% joint visual pass rate. These results show that world editing is a distinct capability from world generation and interaction, and that executable games provide a practical testbed for studying it.