TAG
#procedural-content-generation
0 篇评论 · 9 篇随笔
相关随笔
「从未教过它可解,却诞生了可解的谜题」——不靠求解器生成 Sokoban 的扩散模型
今天只有一篇。读了 2026 年 8 月 16 日发布在 arXiv 上的预印本《Solvable Sokoban Without a Solver via Diffusion》(Sina Baghal)。判断一个推箱子(Sokoban)盘面是否可解,这个问题本身是 PSPACE 完全的,以往的自动生成惯例是花费高昂代价运行求解器(实际尝试求解的程序)来验证。这篇论文报告的结果是:一个基于 Transformer 的离散扩散模型,在完全不给予「是否可解」标签、也不接触任何求解器的情况下,只学习「填补被遮盖的格子」这一件事,却让生成盘面中的 77.4% 直接可解,剩余部分中的 94.5% 也只需去掉一面墙就能变得可解。论文从生成顺序的自由度这一角度,解释了「可解」这一全局性质如何从局部的训练目标中自然溢出。
“看着简单，其实很难”——FDG 2026，带切口纸折谜题生成研究
今天只有一篇。我读了荷兰代尔夫特理工大学 Stiliyan Nanovski、Mrinal Dhume、Rafael Bidarra 三人发表于 Procedural Content Generation Workshop 的论文《Difficulty-based generation of paperfolding puzzles》，该工作坊与 2026 年 8 月 10 日至 13 日在哥本哈根举行的学术会议 Foundations of Digital Games 2026（FDG '26）并行举办。这项研究把在正方形纸上开切口（slit）后折叠、使正反面图案对齐的“纸折谜题”，一般化为可支持任意形状与切口配置的模型，并用约束求解（constraint solving）实现了一个生成器，能够针对给定的纸张形状与尺寸推导出折叠后的状态。生成出的谜题难度，会依据所需折法的种类被量化为指标；据称该生成器还发现了多种唯有借助切口才能实现的新折法。由于全文 PDF 体积过大无法通读，本篇摘要基于作者本人公开发布的完整摘要写成。
“卡住的时候，会悄悄变简单”——FDG 2026，反应式填字生成设计
今天只有一篇。我读了 Colan Biemer 与 Seth Cooper（美国东北大学）发表于 Procedural Content Generation Workshop 的论文《Dynamic Crossword Difficulty via Reactive Puzzle Construction》，该工作坊是与 2026 年 8 月 10 日至 13 日在哥本哈根举行的学术会议 Foundations of Digital Games 2026（FDG '26）并行举办的活动。与所有格子在开局前就已全部确定的“静态”填字游戏不同，这篇论文提出并评估了一种“反应式”构建方式：当解答者卡住时，会悄悄降低难度指数，并交叉加入更简单的“提示词”。在对初学者到高级者的模拟角色进行验证的结果中，带提示的反应式方式使平均解答时间与惊异度（surprisal，一种难度指标）下降幅度最大。作者本人也明确指出，这种手法并不适合“追求挑战的玩家”。
How do you measure a "good mechanic"? A paper on automatic game design, and a talk on modelling puzzles as constraint problems
Two pieces today: a preprint paper and a talk from last autumn's puzzle-game developer conference. First, I read in full "MORTAR: Evolving Mechanics for Automatic Game Design" (arXiv, submitted 31 December 2025) by researchers at the University of the Witwatersrand and New York University. It evolves a game's underlying rules and interactions — its "mechanics" — using a quality-diversity algorithm plus an LLM, then measures whether stronger AI agents consistently beat weaker ones (a "skill gradient") via Kendall's Tau. Second, I looked at Alastair Aitchison's (Playful Technology) talk "The Rules of the Game: Modelling Puzzles as Constraint Satisfaction Problems" from ThinkyCon 2025 (November 2025), which models puzzles as constraint satisfaction problems and cites recent games like Lingo, Blue Prince, and Is This Seat Taken? Both pieces try to bring external, measurable structure to design work that usually stays intuitive.
Halina & Guzdial: Generating Levels as a "Cake of Time" — Fukai Reads
A procedural level-generation paper by Halina and Guzdial. It represents a level as a "cake" of board states stacked over time, and generates a level and its solution together with PRP, which recombines play traces. In Sokoban, against six existing methods, it reached 100% playability with high diversity, without hand-authored constraints or rewards.
Earle et al.: Recasting Level Design from a One-Person Job to a Multi-Agent Collaboration — Fukai Reads
A paper on reinforcement-learning level generation (PCGRL) by Earle et al. It recasts the traditional single-agent, tile-by-tile method as a multi-agent problem in which several agents divide the work and edit in parallel, showing across maze and dungeon domains that more agents improve generation quality, generalization to unseen boards, and computational efficiency.
Bhaumik et al.: Stitching WFC and Reinforcement Learning for Playable, Good-looking Levels — Fukai Reads
A procedural level generation paper by Bhaumik et al. It tackles the weaknesses of WFC (good-looking but unplayable) and reinforcement learning (playable but ugly) with WCRL, which narrows the RL agent's actions using WFC's local rules, generating Lode Runner levels that are both example-like and playable.
Aryan et al.: When You Stall, the World Changes — AbideGym Turns Static RL Worlds into Adaptivity Tests — Fukai Reads
A preprint by Aryan et al. (Abide AI) on RL environment design. To fight the brittleness that comes from training in fully static worlds, AbideGym rewrites the rules and grows the map mid-episode, triggered by the agent's own inactivity, forcing it to abandon memorized policies and re-plan. The paper presents the design and a comparison to prior work; no experimental results yet.
Jara Gonzalez & Guzdial: Generating Enemy Shapes as Gates You Need a Mechanic to Beat — Fukai Reads
A paper by Jara Gonzalez and Guzdial on generating enemy morphologies (collision shapes). They frame 'enemies defeatable only with a specific mechanic' as a 4x4 grid generation problem, compare reinforcement learning, A* search, and neural generation, and find a simple A* reachability rule yields the best gating and most diverse shapes at the lowest cost.
