Agents fail in simple tasks, like crossing a road without driving cars or catching the ball in Pong, if the enemy paddle is easier to beat.
×
Deep Reinforcement Learning Agents are not even close to Human Intelligence
Preprint
ArXiv
RLC 2025 - IBRL Workshop and EWRL 2025
Poster
Corresponding Authors
Quentin Delfosse
Jannis Blüml
Code
HackAtari: Modifications and Game Variations
OCAtari: Object Extraction
JAXAtari: Jax based Atari Learning Environment (Experimental)
If you like our work, feel free to share it.
Share this Paper
https://ml-research.github.io/people/jblueml/papertrees/notintelligent
Copy Link
Close
Share
