Princeton and Berkeley researchers scale empowerment learning
The study uses skill learning to increase agents’ options in complex simulated environments.
Event / data period: 2026-10-09
Published: · Updated:
Researchers at Princeton University and the University of California, Berkeley published a method on 9 October for studying agent empowerment at larger scales. The approach estimates how many executable skills are available from a state, then trains a policy to reach states that offer more options.
In Craftax-Classic, the authors report that agents achieved seven to eight of 22 achievements per rollout after one billion environment steps. The paper also describes behaviors that emerged in simulation, such as building shelter. These results concern specific environments and metrics; the authors do not demonstrate general autonomy or performance in deployed robots or systems.
Data
Study scope
Training budget in the Craftax-Classic experiment; not a real-world measure
| Training environment steps | 1000000000 | 2026-10-09 |
Methodology
Count reported by the paper for its experimental scope; not a national statistic or field result.
Updated:
Subjects and places
Subject proposals
- Science and technologySuggested
- Preserved editorial category/topic; broad candidate onlyarxiv.org ↗
Geographic references
Only source-provided geographic references are shown. Opening an external map is optional. Optional story maps use OpenStreetMap. Opening a map sends your IP address and this site’s address to the map provider; location data remains readable as text without loading the map.
- United States · Place role: Coverage · Place type: Country or area · Location precision: RegionalEvidence
- Explicit preserved editorial country/region; no city or national centroid inferredarxiv.org ↗