This is the pre-proceedings for the RLC 2026. You may expect minor changes.
Reinforcement Learning Journal, vol. 7, 2026, pp. TBD.
Will be presented at the Reinforcement Learning Conference (RLC), Montréal, Quebec, Canada, August 15–17, 2026.
Keywords: Reinforcement Learning, Conditional Computation
While reinforcement learning agents can achieve superhuman performance in many complex tasks, they typically do not become more computationally efficient as they improve. In contrast, humans gradually require less cognitive effort as they become more proficient at a task. If agents could reason about their compute as they learn, could they similarly reduce their computation footprint? If they could, we could have more energy efficient agents or free up compute cycles for other processes like planning. In this paper, we experiment with showing agents the cost of their computation and giving them the ability to control when they use compute. We conduct our experiments on the Arcade Learning Environment, and our results demonstrate that with the same training compute budget, agents that reason about their compute perform better on 75\% of games. Furthermore, these agents use $3$ times less compute on average. We analyze individual games and show where agents gain these efficiencies.
Adrian Orenstein, Jessica Chen, Gwyneth Anne Delos Santos, Bayley Sapara, and Michael Bowling. "Toward Agents That Reason About Their Computation." Reinforcement Learning Journal, vol. 7, 2026, pp. TBD.
BibTeX:@article{orenstein2026toward,
title={Toward Agents That Reason About Their Computation},
author={Adrian Orenstein and Jessica Chen and Gwyneth Anne Delos Santos and Bayley Sapara and Michael Bowling},
journal={Reinforcement Learning Journal},
volume={7},
pages={},
year={2026}
}