papers

Publications (29)

cs.LG2020

Behaviour Suite for Reinforcement Learning

Ian Osband, Yotam Doron, Matteo Hessel +11

This paper introduces the Behaviour Suite for Reinforcement Learning, or bsuite for short. bsuite is a collection of carefully-designed experiments that investigate core capabiliti…

cs.LG2022

Muesli: Combining Improvements in Policy Optimization

Matteo Hessel, Ivo Danihelka, Fabio Viola +6

We propose a novel policy update that combines regularized policy optimization with model learning as an auxiliary loss. The update (henceforth Muesli) matches MuZero's state-of-th…

cs.LG2021

Self-Consistent Models and Values

Gregory Farquhar, Kate Baumli, Zita Marinho +4

Learned models of the environment provide reinforcement learning (RL) agents with flexible ways of making predictions about the environment. In particular, models enable planning,…

cs.LG2018

Multi-task Deep Reinforcement Learning with PopArt

Matteo Hessel, Hubert Soyer, Lasse Espeholt +3

The reinforcement learning community has made great strides in designing algorithms capable of exceeding human performance on specific tasks. These algorithms are mostly trained on…

stat.ML2021

A Self-Tuning Actor-Critic Algorithm

Tom Zahavy, Zhongwen Xu, Vivek Veeriah +5

Reinforcement learning algorithms are highly sensitive to the choice of hyperparameters, typically requiring significant manual effort to identify hyperparameters that perform well…

cs.LG2019

When to use parametric models in reinforcement learning?

Hado van Hasselt, Matteo Hessel, John Aslanides

We examine the question of when and how parametric models are most useful in reinforcement learning. In particular, we look at commonalities and differences between parametric mode…

cs.LG2020

Meta-Gradient Reinforcement Learning with an Objective Discovered Online

Zhongwen Xu, Hado van Hasselt, Matteo Hessel +3

Deep reinforcement learning includes a broad family of algorithms that parameterise an internal representation, such as a value function or policy, by a deep neural network. Each a…

cs.LG2018

Unicorn: Continual Learning with a Universal, Off-policy Agent

Daniel J. Mankowitz, Augustin Žídek, André Barreto +7

Some real-world domains are best characterized as a single task, but for others this perspective is limiting. Instead, some tasks continually grow in complexity, in tandem with the…

cs.AI2018

Deep Reinforcement Learning and the Deadly Triad

Hado van Hasselt, Yotam Doron, Florian Strub +3

We know from reinforcement learning theory that temporal difference learning can fail in certain cases. Sutton and Barto (2018) identify a deadly triad of function approximation, b…

cs.AI2017

Rainbow: Combining Improvements in Deep Reinforcement Learning

Matteo Hessel, Joseph Modayil, Hado van Hasselt +7

The deep reinforcement learning community has made several independent improvements to the DQN algorithm. However, it is unclear which of these extensions are complementary and can…

cs.LG2019

General non-linear Bellman equations

Hado van Hasselt, John Quan, Matteo Hessel +3

We consider a general class of non-linear Bellman equations. These open up a design space of algorithms that have interesting properties, which has two potential advantages. First,…

stat.ML2025

DataRater: Meta-Learned Dataset Curation

Dan A. Calian, Gregory Farquhar, Iurii Kemaev +9

The quality of foundation models depends heavily on their training data. Consequently, great efforts have been put into dataset curation. Yet most approaches rely on manual tuning…

cs.LG2021

Discovering Reinforcement Learning Algorithms

Junhyuk Oh, Matteo Hessel, Wojciech M. Czarnecki +4

Reinforcement learning (RL) algorithms update an agent's parameters according to one of several possible rules, discovered manually through years of research. Automating the discov…

cs.LG2017

The Predictron: End-To-End Learning and Planning

David Silver, Hado van Hasselt, Matteo Hessel +8

One of the key challenges of artificial intelligence is to learn models that are effective in the context of planning. In this document we introduce the predictron architecture. Th…

cs.CL2025

Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

Gheorghe Comanici, Eric Bieber, Mike Schaekermann +3431

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our…

cs.LG2018

Scaling shared model governance via model splitting

Miljan Martic, Jan Leike, Andrew Trask +3

Currently the only techniques for sharing governance of a deep learning model are homomorphic encryption and secure multiparty computation. Unfortunately, neither of these techniqu…

cs.LG2018

Distributed Prioritized Experience Replay

Dan Horgan, John Quan, David Budden +4

We propose a distributed architecture for deep reinforcement learning at scale, that enables agents to learn effectively from orders of magnitude more data than previously possible…

cs.LG2021

Podracer architectures for scalable Reinforcement Learning

Matteo Hessel, Manuel Kroiss, Aidan Clark +5

Supporting state-of-the-art AI research requires balancing rapid prototyping, ease of use, and quick iteration, with the ability to deploy experiments at a scale traditionally asso…

cs.LG2018

Observe and Look Further: Achieving Consistent Performance on Atari

Tobias Pohlen, Bilal Piot, Todd Hester +10

Despite significant advances in the field of deep Reinforcement Learning (RL), today's algorithms still fail to learn human-level policies consistently over a set of diverse tasks…

cs.LG2021

Emphatic Algorithms for Deep Reinforcement Learning

Ray Jiang, Tom Zahavy, Zhongwen Xu +4

Off-policy learning allows us to learn about possible policies of behavior from experience generated by a different behavior policy. Temporal difference (TD) learning algorithms ca…

cs.LG2016

Learning values across many orders of magnitude

Hado van Hasselt, Arthur Guez, Matteo Hessel +2

Most learning algorithms are not invariant to the scale of the function that is being approximated. We propose to adaptively normalize the targets used in learning. This is useful…

cs.AI2019

Discovery of Useful Questions as Auxiliary Tasks

Vivek Veeriah, Matteo Hessel, Zhongwen Xu +6

Arguably, intelligent agents ought to be able to discover their own questions so that in learning answers for them they learn unanticipated useful knowledge and skills; this depart…

cs.LG2016

Dueling Network Architectures for Deep Reinforcement Learning

Ziyu Wang, Tom Schaul, Matteo Hessel +3

In recent years there have been many successes of using deep representations in reinforcement learning. Still, many of these applications use conventional architectures, such as co…

cs.LG2021

Expected Eligibility Traces

Hado van Hasselt, Sephora Madjiheurem, Matteo Hessel +3

The question of how to determine which states and actions are responsible for a certain outcome is known as the credit assignment problem and remains a central research question in…

cs.LG2019

Off-Policy Actor-Critic with Shared Experience Replay

Simon Schmitt, Matteo Hessel, Karen Simonyan

We investigate the combination of actor-critic reinforcement learning algorithms with uniform large-scale experience replay and propose solutions for two challenges: (a) efficient…

cs.LG2019

On Inductive Biases in Deep Reinforcement Learning

Matteo Hessel, Hado van Hasselt, Joseph Modayil +1

Many deep reinforcement learning algorithms contain inductive biases that sculpt the agent's objective and its interface to the environment. These inductive biases can take many fo…

cs.AI2020

What Can Learned Intrinsic Rewards Capture?

Zeyu Zheng, Junhyuk Oh, Matteo Hessel +5

The objective of a reinforcement learning agent is to behave so as to maximise the sum of a suitable scalar function of state: the reward. These rewards are typically given and imm…

cs.LG2019

Transfer in Deep Reinforcement Learning Using Successor Features and Generalised Policy Improvement

André Barreto, Diana Borsa, John Quan +6

The ability to transfer skills across tasks has the potential to scale up reinforcement learning (RL) agents to environments currently out of reach. Recently, a framework based on…

cs.LG2021

Discovery of Options via Meta-Learned Subgoals

Vivek Veeriah, Tom Zahavy, Matteo Hessel +6

Temporal abstractions in the form of options have been shown to help reinforcement learning (RL) agents learn faster. However, despite prior work on this topic, the problem of disc…