UCB exploration via Q-ensembles

OpenAI Blog Papers

Summary

OpenAI presents a novel exploration strategy for deep reinforcement learning using ensembles of Q-functions with upper-confidence bounds (UCB), demonstrating significant performance improvements on the Atari benchmark.

No content available
Original Article
View Cached Full Text

Cached at: 04/20/26, 02:45 PM

# UCB exploration via Q-ensembles Source: [https://openai.com/index/ucb-exploration-via-q-ensembles/](https://openai.com/index/ucb-exploration-via-q-ensembles/) OpenAI## Abstract We show how an ensemble of Q\*\-functions can be leveraged for more effective exploration in deep reinforcement learning\. We build on well established algorithms from the bandit setting, and adapt them to the Q\-learning setting\. We propose an exploration strategy based on upper\-confidence bounds \(UCB\)\. Our experiments show significant gains on the Atari benchmark\.

Similar Articles

Benchmarking safe exploration in deep reinforcement learning

OpenAI Blog

OpenAI proposes standardizing constrained RL as the formalism for safe exploration and introduces Safety Gym, a benchmark suite for evaluating safe deep RL algorithms in high-dimensional continuous control tasks with safety constraints.

EvoCUA-1.5: Online Reinforcement Learning for Multi-turn Computer-Use Agents

arXiv cs.AI

EvoCUA-1.5 introduces an online reinforcement learning framework for multi-turn computer-use agents, achieving a 63.2% success rate on OSWorld-Verified and outperforming comparable open-weight models up to 35B parameters through step-level policy optimization and dynamic curriculum learning.

Learning Montezuma’s Revenge from a single demonstration

OpenAI Blog

OpenAI demonstrates a method for training a reinforcement learning agent to play Montezuma's Revenge from a single human demonstration, addressing the challenge of sparse rewards through curriculum learning and careful hyperparameter tuning. The approach achieves strong performance on the notoriously difficult Atari game while showing generalization limitations on other titles.