Position: Profiling Game Worlds by Transition Complexity

Abstract: Game world modeling (GWM) and reinforcement learning (RL) are often confounded because research papers rarely quantify how difficult the underlying transition prediction problem is at the declared interface (pixels/tokens/latents with finite history).

摘要： 游戏世界建模（GWM）和强化学习（RL）的研究经常被混淆，因为研究论文很少量化在既定接口（具有有限历史的像素/标记/潜在变量）下，底层转换预测问题的难度。

We propose the Transition Complexity Profile (TCP): a small, reproducible set of metrics that characterizes an environment’s (or gameplay dataset’s) induced transition kernel by (i) intrinsic one-step branching, (ii) interaction-induced uncertainty and opponent influence when observable, and (iii) temporal/spatial dependency span via standardized probe curves.

我们提出了转换复杂度概况（Transition Complexity Profile, TCP）：这是一套小巧且可复现的指标，通过以下三个维度来表征环境（或游戏数据集）诱导的转换核：(i) 内在的单步分支，(ii) 可观测时的交互诱导不确定性与对手影响，以及 (iii) 通过标准化探测曲线得出的时空依赖跨度。

TCP is reported with an explicit reference distribution, protocol stochasticity, and a versioned measurement budget (sampling/resampling and fixed probe compute), enabling comparable numbers across benchmarks.

TCP 的报告包含明确的参考分布、协议随机性以及版本化的测量预算（采样/重采样和固定的探测计算），从而使得不同基准测试之间的数值具有可比性。

We outline how common game families and modern “neural game engine” domains populate this landscape and call for TCP to become standard benchmark metadata and a required statistic in GWM and RL papers.

我们概述了常见的游戏系列和现代“神经游戏引擎”领域如何填充这一图景，并呼吁将 TCP 纳入标准基准测试元数据，并使其成为 GWM 和 RL 论文中的必要统计指标。