Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planning

TL;DR

This paper introduces discrete diffusion models, specifically Multi-Granularity Diffusion Modeling (MGDM), as a novel approach that significantly improves complex reasoning and planning tasks over traditional autoregressive models, without relying on search techniques.

Contribution

The paper proposes MGDM, a new diffusion-based framework that prioritizes subgoals by difficulty, enabling better learning of complex subgoals and outperforming autoregressive models on reasoning tasks.

Findings

MGDM achieves 91.5% accuracy on Countdown.

MGDM reaches 100% accuracy on Sudoku.

Outperforms autoregressive models without search techniques.

Abstract

Autoregressive language models, despite their impressive capabilities, struggle with complex reasoning and long-term planning tasks. We introduce discrete diffusion models as a novel solution to these challenges. Through the lens of subgoal imbalance, we demonstrate how diffusion models effectively learn difficult subgoals that elude autoregressive approaches. We propose Multi-Granularity Diffusion Modeling (MGDM), which prioritizes subgoals based on difficulty during learning. On complex tasks like Countdown, Sudoku, and Boolean Satisfiability Problems, MGDM significantly outperforms autoregressive models without using search techniques. For instance, MGDM achieves 91.5\% and 100\% accuracy on Countdown and Sudoku, respectively, compared to 45.8\% and 20.7\% for autoregressive models. Our work highlights the potential of diffusion-based approaches in advancing AI capabilities for sophisticated language understanding and problem-solving tasks. All associated codes are available at \href{https://github.com/HKUNLP/diffusion-vs-ar}{https://github.com/HKUNLP/diffusion-vs-ar}.