Mechanisms of Misgeneralization in Physical Sequence Modeling
Kento Nishi, Raphael Tang, Karun Kumar, Core Francisco Park † \dagger, Hidenori Tanaka † \dagger · 2026
Read Mechanisms of Misgeneralization in Physical Sequence Modeling by Kento Nishi, Raphael Tang, Karun Kumar, Core Francisco Park † \dagger, Hidenori Tanaka † \dagger (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-2605-20299.txt.
Description
Kento Nishi and four co-workers name physical misgeneralization, a flaw in motion-planning models. Maze robots get trips spread flat over a set length range, to cap power use. Each trip can look sound alone, yet the spread across all trips comes out wrong, as small local slips pass through the physics measure and skew it. A data deviation kernel gauges those slips and foretells which traits gain or lose weight. It works in toy, maze, and double-pendulum tests, and points to one kernel fix.
Catalog record
| CREATOR | Kento Nishi, Raphael Tang, Karun Kumar, Core Francisco Park † \dagger, Hidenori Tanaka † \dagger |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 75652 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.20299