AutoEnv: Automated Environments for Measuring Cross-Environment Agent Learning
Foundation Agents proposes AutoEnv, a framework for automatically generating diverse RL environments and evaluating an agent's ability to learn across them rather than within a single fixed one.
- Publisher
- Foundation Agents
- Year
- 2025
- Venue
- preprint
- Authors
- 16
- Hosting
- External sourcelicense unknown
Cite
Notes
Only stored in your browser.
Introduces 1 artifact - 1 tool
TL;DR
Semantic Scholar
This work proposes AutoEnv, an automated framework that treats environments as factorizable distributions over transitions, observations, and rewards, enabling low-cost generation of heterogeneous worlds and formalizes agent learning as a component-centric process driven by three stages of Selection, Optimization, and Evaluation applied to an improvable agent component.
Artifacts
1Tools