TY  - RPRT
TI  - Smaller Abstract State Spaces Enable Cross-Scale Generalization in Reinforcement Learning
AU  - Nasehatul Mustakim
AU  - Lucas Lehnert
PY  - 2026
UR  - https://arxiv.org/abs/2605.20272
ID  - 2605.20272
ER  -