Monte Carlo Tree Diffusion for System 2 Planning

TL;DR

This paper introduces Monte Carlo Tree Diffusion (MCTD), a novel planning framework that combines diffusion models with Monte Carlo Tree Search to improve long-horizon planning performance.

Contribution

MCTD integrates diffusion models with MCTS, enabling iterative evaluation and refinement of plans, which enhances scalability and solution quality in complex planning tasks.

Findings

MCTD outperforms diffusion baselines on long-horizon tasks.

Higher inference-time computation improves solution quality.

MCTD effectively balances exploration and exploitation.

Abstract

Diffusion models have recently emerged as a powerful tool for planning. However, unlike Monte Carlo Tree Search (MCTS)-whose performance naturally improves with inference-time computation scaling-standard diffusion-based planners offer only limited avenues for the scalability. In this paper, we introduce Monte Carlo Tree Diffusion (MCTD), a novel framework that integrates the generative strength of diffusion models with the adaptive search capabilities of MCTS. Our method reconceptualizes denoising as a tree-structured process, allowing partially denoised plans to be iteratively evaluated, pruned, and refined. By selectively expanding promising trajectories while retaining the flexibility to revisit and improve suboptimal branches, MCTD achieves the benefits of MCTS such as controlling exploration-exploitation trade-offs within the diffusion framework. Empirical results on challenging long-horizon tasks show that MCTD outperforms diffusion baselines, yielding higher-quality solutions as inference-time computation increases.