M3PO: Massively Multi-Task Model-Based Policy Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Narendra, Aditya, Makarov, Dmitry, Panov, Aleksandr
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908425331408896
author Narendra, Aditya
Makarov, Dmitry
Panov, Aleksandr
author_facet Narendra, Aditya
Makarov, Dmitry
Panov, Aleksandr
contents We introduce Massively Multi-Task Model-Based Policy Optimization (M3PO), a scalable model-based reinforcement learning (MBRL) framework designed to address sample inefficiency in single-task settings and poor generalization in multi-task domains. Existing model-based approaches like DreamerV3 rely on pixel-level generative models that neglect control-centric representations, while model-free methods such as PPO suffer from high sample complexity and weak exploration. M3PO integrates an implicit world model, trained to predict task outcomes without observation reconstruction, with a hybrid exploration strategy that combines model-based planning and model-free uncertainty-driven bonuses. This eliminates the bias-variance trade-off in prior methods by using discrepancies between model-based and model-free value estimates to guide exploration, while maintaining stable policy updates through a trust-region optimizer. M3PO provides an efficient and robust alternative to existing model-based policy optimization approaches and achieves state-of-the-art performance across multiple benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2506_21782
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle M3PO: Massively Multi-Task Model-Based Policy Optimization
Narendra, Aditya
Makarov, Dmitry
Panov, Aleksandr
Machine Learning
Robotics
We introduce Massively Multi-Task Model-Based Policy Optimization (M3PO), a scalable model-based reinforcement learning (MBRL) framework designed to address sample inefficiency in single-task settings and poor generalization in multi-task domains. Existing model-based approaches like DreamerV3 rely on pixel-level generative models that neglect control-centric representations, while model-free methods such as PPO suffer from high sample complexity and weak exploration. M3PO integrates an implicit world model, trained to predict task outcomes without observation reconstruction, with a hybrid exploration strategy that combines model-based planning and model-free uncertainty-driven bonuses. This eliminates the bias-variance trade-off in prior methods by using discrepancies between model-based and model-free value estimates to guide exploration, while maintaining stable policy updates through a trust-region optimizer. M3PO provides an efficient and robust alternative to existing model-based policy optimization approaches and achieves state-of-the-art performance across multiple benchmarks.
title M3PO: Massively Multi-Task Model-Based Policy Optimization
topic Machine Learning
Robotics
url https://arxiv.org/abs/2506.21782