Search papers, labs, and topics across Lattice.
This paper critiques the conventional approach of scaling world models through increased data and compute, arguing for the necessity of a recursive data engine that delivers grounded reward signals. It highlights the limitations of current reward systems in spatial generation, which rely on imprecise proxies like CLIP scores, and contrasts this with the advantages of game development as a robust environment for generating high-quality rewards. The authors introduce Reinforcement Learning with Human-Engine Verification (RLHEV), which leverages executable game environments to provide both dense feedback and human acceptance signals, enabling more effective post-training reinforcement learning.
Game development could revolutionize how we generate high-quality reward signals for spatial world models, moving beyond fuzzy proxies to executable environments.
A common strategy for scaling world models is to train on more crawled video with more compute. We argue that this strategy is inefficient: scaling world models also requires a recursive data engine that offers grounded reward signals. The success of code agents illustrates why this matters. As code is executable, compilers and runtimes can provide high-quality rewards for Reinforcement Learning (RL) post-training of LLMs. By contrast, spatial generation still relies largely on fuzzy proxies such as CLIP scores. These signals are fuzzy and biased, making them hard to support RL post-training. Compared with these, game development provides a missing reward environment for spatial world models. A scene encoded by a game engine is an executable world specification: the engine can efficiently check collision, physics, navigability and bounded playability, while the developer provides the global verification signal by judging whether the scene should be accepted. Game development also provides real-world long-horizon trajectory data for RL post-training. We therefore propose Reinforcement Learning with Human-Engine Verification (RLHEV), a post-training paradigm that combines dense engine signals with implicit human acceptance feedback from the development process.
