We are happy to share that two papers from LiteRL have been accepted at the International Conference on Learning Representations (ICLR 2026). Both tackle the same question from different angles: how can deep reinforcement learning agents learn value functions quickly without giving up stability?
Use the Online Network If You Can: Towards Fast and Stable Reinforcement Learning
A. Hendawy, H. Metternich, T. Vincent, M. Kallel, J. Peters, C. D’Eramo
We introduce MINTO, a one-line change to the bootstrapping target that takes the minimum between the target and the online network. It learns faster than using a target network alone, while staying stable. Read the paper →
Bridging the Performance Gap Between Target-Free and Target-Based Reinforcement Learning with Iterated Q-Learning
T. Vincent, Y. Tripathi, T. Faust, Y. Oren, J. Peters, C. D’Eramo
We present iS-QL, which keeps only a copy of the last layer as a target network and learns several Bellman updates in parallel. It matches target-based methods while using a single Q-network. Read the paper →
Congratulations to all authors! See all our work on the Publications page.