AI 玩新游戏，看 80 次就能学会规则
我们总以为 AI 学新游戏规则要海量数据；这篇发现，一个 3500 万参数的小模型，在没见过的游戏里，只要看大约 80 次某个机制发生，就能以 80% 以上的准确率预测它接下来会怎么改变画面——而且是在一局游戏内、边玩边学，不需要提前训练。研究者把游戏规则拆成 69 种可标注的「机制」，让模型每看一步就更新自己，结果五分之四的机制都能学会，而如果不让它实时更新，只有七分之一能学会。这暗示：AI 适应新环境的关键可能不是模型更大，而是「边看边改自己」的能力。它不是你明天能用上的东西，但对「AI 能不能像人一样快速上手新任务」这个问题，是个实打实的信号。
📄 原文摘要(英文)
World models enable agents to plan and learn from imagined experience, but in an unfamiliar environment they must learn its mechanics from limited interaction. A mechanic is a rule for how an action or an interaction changes part of the environment. Each occurrence of a mechanic that the model observes is one exposure. We ask how many exposures a world model needs to predict the effects of a new environment's mechanics. We address this question on ARC-AGI-3, a benchmark of interactive games defined by their programs. We introduce MechEval, which labels 69 recurring mechanics across 20 games from these programs. It scores each prediction on the cells the mechanic changes at each exposure. As a test bed, we build WARC (World model for ARC-AGI-3), a 35M-parameter vision transformer that is pretrained on other games and updates all of its parameters after each transition. On held-out games, WARC learns to predict four in five mechanics with at least 80% accuracy on the cells they change and keeps that accuracy to the end of the recordings. Half of them reach that accuracy within about 80 exposures, in a single play of the game, and without updates only one in seven does at all. In closed-loop play on the public games, WARC-Agent, which pairs WARC with a fixed graph explorer, triples the explorer's official efficiency score from 0.15 to 0.46, mostly through test-time training. On the hidden games of the Kaggle leaderboard, our entry scores 0.20.