arXiv cs.AI· Doojin Baek, Junyeob Baek, Mingyu Jo, Hosung Lee, Taegu Kang, Gyubin Lee, Sungjin Ahn·· 11 小时前AI 评分39
BabyTheorist：仅从观察中学习世界理论的基准测试
BabyTheorist: A Benchmark for Learning to Theorize the World from Observation Alone
AI 导读
BabyTheorist 是一个用于受控训练和评估“观察式理论学习”的基准，从隐藏的组合程序中生成连续的视觉前后状态观察流，训练时学习者只能看到状态转移，没有程序标签、任务身份或支持集。该基准控制程序语言、程序复杂度课程、高频重复子程序和带跳级的持续训练-测试协议。七个代表性基线均远低于上限，且随课程推进差距扩大。
正文 · 原文
Abstract:How would we know whether a machine truly understands the world? We argue that one key capability is \textit{observational theory learning}: acquiring reusable theories from pure observations and applying them to new situations. Current visual reasoning benchmarks focus primarily on the latter, presenting each task as a few-shot puzzle grouped by a common rule. We introduce BabyTheorist, a benchmark for controlled training and evaluation of observational theory learning. BabyTheorist generates continual streams of before-after visual observations from hidden compositional programs; during training, learners see only the resulting transitions, with no program labels, task identities, or support sets. To support theory learning and evaluation, BabyTheorist controls the program language, a curriculum over program complexity, recurring frequent subprograms, and a continual train-test protocol with level skipping. Seven representative baselines plateau well below ceiling, with the gap widening as the curriculum advances---evidence that BabyTheorist isolates a capability current program-learning approaches do not yet address.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.08177 [cs.AI] |
| (or arXiv:2611.08177v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08177 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Junyeob Baek [view email]
[v1]
Thu, 5 Nov 2026 15:33:32 UTC (7,293 KB)
来源：arXiv cs.AI · arxiv.org