Discrete Diffusion Models Exploit Asymmetry to Solve Lookahead Planning Tasks

TL;DR

This paper compares autoregressive and non-autoregressive discrete diffusion models in lookahead planning tasks, revealing that NAR models exploit asymmetry to efficiently solve these tasks with less data and simpler architectures.

Contribution

It uncovers the mechanistic differences enabling NAR models to leverage asymmetry for planning, demonstrating their efficiency over AR models in lookahead tasks.

Findings

NAR models solve planning tasks by decoding backwards using future tokens.

NAR models require exponentially fewer training examples.

AR models often fail to converge without curriculum adjustments.

Abstract

While Autoregressive (AR) Transformer-based Generative Language Models are frequently employed for lookahead tasks, recent research suggests a potential discrepancy in their ability to perform planning tasks that require multi-step lookahead. In this work, we investigate the distinct emergent mechanisms that arise when training AR versus Non-Autoregressive (NAR) models, such as Discrete Diffusion Models (dLLMs), on lookahead tasks. By requiring the models to plan ahead to reach the correct conclusion, we analyze how these two paradigms fundamentally differ in their approach to the problem. We identify a critical asymmetry in planning problems: while forward generation requires complex lookahead at branching junctions, reverse generation is often deterministic. This asymmetry creates an opportunity for NAR models. Through mechanistic analysis of training and inference dynamics, we demonstrate that NAR models learn to solve planning tasks by utilizing future tokens to decode backwards, avoiding the need to learn complex traversal mechanisms entirely. Consequently, we report that both AR and NAR models are able to achieve perfect accuracy on the lookahead task. However, NAR models require exponentially fewer training examples and shallower architectures compared to AR models, which often fail to converge without specific curriculum adjustments.