HaM-World : Soft-Hamiltonian World Models with Selective Memory for Planning
Haoyun Tang, Haodong Cui, Keyao Xu, Kun Wang, Zhandong Mei · 2026
Read HaM-World : Soft-Hamiltonian World Models with Selective Memory for Planning by Haoyun Tang, Haodong Cui, Keyao Xu, Kun Wang, Zhandong Mei (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-2605-05951.txt.
Description
Haoyun Tang and four co-authors present a world model designed to keep imagined futures stable for planning. World models learn how a scene changes, then let an agent test actions inside the model. Long imagined rollouts often drift off course, and shifts in the setting make the drift worse.
HaM-World pairs selective memory, which sums up past observations and actions, with dynamics shaped by a soft Hamiltonian prior. The design splits the hidden state into a position and momentum part plus a context part for non conservative effects. On six control tasks it ranked first on four and cut rollout error sharply at short horizons. It also held up best under twelve kinds of test disturbance.
Catalog record
| CREATOR | Haoyun Tang, Haodong Cui, Keyao Xu, Kun Wang, Zhandong Mei |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 76 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.05951