Learning to Theorize the World from Observation
Doojin Baek, Gyubin Lee, Junyeob Baek, Hosung Lee, Sungjin Ahn · 2026
Read Learning to Theorize the World from Observation by Doojin Baek, Gyubin Lee, Junyeob Baek, Hosung Lee, Sungjin Ahn (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-03413.txt.
Description
Doojin Baek, Gyubin Lee, Junyeob Baek, Hosung Lee, and Sungjin Ahn propose learning explicit theories of the world from raw observation, rather than only learning to predict what comes next. Human children build internal theories of how things work before mature language, and the authors take that as the model for machine understanding.
Their Neural Theorizer induces latent programs in a learned language of thought and runs them through a shared transition model. Because theories are compositional programs, their parts recombine to explain new phenomena. Experiments show explanation-driven generalization, where observations are understood through the programs that could generate them.
Catalog record
| CREATOR | Doojin Baek, Gyubin Lee, Junyeob Baek, Hosung Lee, Sungjin Ahn |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 97 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.03413