Predictive representations: Building blocks of intelligence

W Carvalho, MS Tomov, W de Cothi, C Barry… - Neural …, 2024 - direct.mit.edu
Adaptive behavior often requires predicting future events. The theory of reinforcement
learning prescribes what kinds of predictive representations are useful and how to compute …

Learning Temporal Distances: Contrastive Successor Features Can Provide a Metric Structure for Decision-Making

V Myers, C Zheng, A Dragan, S Levine… - arXiv preprint arXiv …, 2024 - arxiv.org
Temporal distances lie at the heart of many algorithms for planning, control, and
reinforcement learning that involve reaching goals, allowing one to estimate the transit time …

Reinforcement Learning: An Overview

K Murphy - arXiv preprint arXiv:2412.05265, 2024 - arxiv.org
This manuscript gives a big-picture, up-to-date overview of the field of (deep) reinforcement
learning and sequential decision making, covering value-based RL, policy-gradient …

Accelerating Goal-Conditioned RL Algorithms and Research

M Bortkiewicz, W Pałucki, V Myers, T Dziarmaga… - arXiv preprint arXiv …, 2024 - arxiv.org
Self-supervision has the potential to transform reinforcement learning (RL), paralleling the
breakthroughs it has enabled in other areas of machine learning. While self-supervised …

DeMoBot: Deformable Mobile Manipulation with Vision-based Sub-goal Retrieval

Y Zhang, W Yang, J Pajarinen - arXiv preprint arXiv:2408.15919, 2024 - arxiv.org
Imitation learning (IL) algorithms typically distill experience into parametric behavior policies
to mimic expert demonstrations. Despite their effectiveness, previous methods often struggle …

A Single Goal is All You Need: Skills and Exploration Emerge from Contrastive RL without Rewards, Demonstrations, or Subgoals

G Liu, M Tang, B Eysenbach - arXiv preprint arXiv:2408.05804, 2024 - arxiv.org
In this paper, we present empirical evidence of skills and directed exploration emerging from
a simple RL algorithm long before any successful trials are observed. For example, in a …

Can a MISL Fly? Analysis and Ingredients for Mutual Information Skill Learning

C Zheng, J Tuyls, J Peng, B Eysenbach - arXiv preprint arXiv:2412.08021, 2024 - arxiv.org
Self-supervised learning has the potential of lifting several of the key challenges in
reinforcement learning today, such as exploration, representation learning, and reward …

Exploration by Learning Diverse Skills through Successor State Measures

PAL Tolguenec, Y Besse… - arXiv preprint arXiv …, 2024 - arxiv.org
The ability to perform different skills can encourage agents to explore. In this work, we aim to
construct a set of diverse skills which uniformly cover the state space. We propose a …

A Single Goal is All You Need

G Liu, M Tang, B Eysenbach - Intrinsically-Motivated and Open-Ended … - openreview.net
In this paper, we present empirical evidence of skills and directed exploration emerging from
a simple RL algorithm long before any successful trials are observed. For example, in a …