Policy Rehearsing: Training Generalizable Policies for Reinforcement Learning

C Jia, C Gao, H Yin, F Zhang, XH Chen… - The Twelfth …, 2024 - openreview.net
Human beings can make adaptive decisions in a preparatory manner, ie, by making
preparations in advance, which offers significant advantages in scenarios where both online …

No regrets: Investigating and improving regret approximations for curriculum discovery

A Rutherford, M Beukman, T Willi, B Lacerda… - arxiv preprint arxiv …, 2024 - arxiv.org
What data or environments to use for training to improve downstream performance is a
longstanding and very topical question in reinforcement learning. In particular …

minimax: Efficient Baselines for Autocurricula in JAX

M Jiang, M Dennis, E Grefenstette… - arxiv preprint arxiv …, 2023 - arxiv.org
Unsupervised environment design (UED) is a form of automatic curriculum learning for
training robust decision-making agents to zero-shot transfer into unseen environments. Such …

Environment curriculum generation via large language models

W Liang, S Wang, HJ Wang, O Bastani… - … Conference on Robot …, 2024 - openreview.net
Recent work has demonstrated that a promising strategy for teaching robots a wide range of
complex skills is by training them on a curriculum of progressively more challenging …

The Overcooked Generalisation Challenge

C Ruhdorfer, M Bortoletto, A Penzkofer… - arxiv preprint arxiv …, 2024 - arxiv.org
We introduce the Overcooked Generalisation Challenge (OGC)-the first benchmark to study
agents' zero-shot cooperation abilities when faced with novel partners and levels in the …

Multi-Agent Diagnostics for Robustness via Illuminated Diversity

M Samvelyan, D Paglieri, M Jiang… - arxiv preprint arxiv …, 2024 - arxiv.org
In the rapidly advancing field of multi-agent systems, ensuring robustness in unfamiliar and
adversarial settings is crucial. Notwithstanding their outstanding performance in familiar …

Syllabus: Portable Curricula for Reinforcement Learning Agents

R Sullivan, R Pégoud, AU Rahmen, X Yang… - arxiv preprint arxiv …, 2024 - arxiv.org
Curriculum learning has been a quiet yet crucial component of many of the high-profile
successes of reinforcement learning. Despite this, none of the major reinforcement learning …

Refining Minimax Regret for Unsupervised Environment Design

M Beukman, S Coward, M Matthews, M Fellows… - arxiv preprint arxiv …, 2024 - arxiv.org
In unsupervised environment design, reinforcement learning agents are trained on
environment configurations (levels) generated by an adversary that maximises some …

Adversarial Environment Design via Regret-Guided Diffusion Models

H Chung, J Lee, M Kim, D Kim, S Oh - arxiv preprint arxiv:2410.19715, 2024 - arxiv.org
Training agents that are robust to environmental changes remains a significant challenge in
deep reinforcement learning (RL). Unsupervised environment design (UED) has recently …

Scenario-Based Curriculum Generation for Multi-Agent Autonomous Driving

A Brunnbauer, L Berducci, P Priller, D Nickovic… - arxiv preprint arxiv …, 2024 - arxiv.org
The automated generation of diverse and complex training scenarios has been an important
ingredient in many complex learning tasks. Especially in real-world application domains …