Discrete Diffusion Models Exploit Asymmetry to Solve Lookahead Planning Tasks

• URL: http://arxiv.org/abs/2602.19980v1
• Date: Mon, 23 Feb 2026 15:47:27 GMT
• Title: Discrete Diffusion Models Exploit Asymmetry to Solve Lookahead Planning Tasks
• Authors: Itamar Trainin, Shauli Ravfogel, Omri Abend, Amir Feder,
• Abstract summary: We show that Non-Autoregressive (NAR) models learn to solve planning tasks by utilizing future tokens to decode backwards.<n>We report that both AR and NAR models are able to achieve perfect accuracy on the lookahead task.
• Score: 31.338692248735892
• License: http://creativecommons.org/licenses/by/4.0/
• 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.

Related papers

• Deep Neural Networks as Iterated Function Systems and a Generalization Bound [2.7920304852537536]
  We show that two important deep architectures can be viewed as, or canonically associated with, place-dependent IFS.<n>We derive a Wasserstein bound for generative modeling that controls the collage-type approximation error between the data distribution and its image.
  arXiv  Detail & Related papers  (2026-01-27T07:32:49Z)
• Architecture Decoupling Is Not All You Need For Unified Multimodal Model [64.19284951218098]
  We propose Attention Interaction Alignment (AIA) loss, which explicitly learns Task-Specific multimodal interaction patterns during training.<n>AIA not only refines cross-modal attention patterns, but also boosts both generation and understanding performance.
  arXiv  Detail & Related papers  (2025-11-27T17:55:25Z)
• Embodied-R: Collaborative Framework for Activating Embodied Spatial Reasoning in Foundation Models via Reinforcement Learning [58.86928947970342]
  Embodied-R is a framework combining large-scale Vision-Language Models for perception and small-scale Language Models for reasoning.<n>After training on only 5k embodied video samples, Embodied-R with a 3B LM matches state-of-the-art multimodal reasoning models.<n>Embodied-R also exhibits emergent thinking patterns such as systematic analysis and contextual integration.
  arXiv  Detail & Related papers  (2025-04-17T06:16:11Z)
• Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining [74.83412846804977]
  Reinforcement learning (RL)-based fine-tuning has become a crucial step in post-training language models.<n>We present a systematic end-to-end study of RL fine-tuning for mathematical reasoning by training models entirely from scratch.
  arXiv  Detail & Related papers  (2025-04-10T17:15:53Z)
• Transformers Use Causal World Models in Maze-Solving Tasks [49.67445252528868]
  We identify World Models in transformers trained on maze-solving tasks.<n>We find that it is easier to activate features than to suppress them.<n> positional encoding schemes appear to influence how World Models are structured within the model's residual stream.
  arXiv  Detail & Related papers  (2024-12-16T15:21:04Z)
• Investigating the Impact of Model Complexity in Large Language Models [3.7919508292745676]
  Large Language Models (LLMs) based on the pre-trained fine-tuning paradigm have become pivotal in solving natural language processing tasks. In this paper, we focus on autoregressive LLMs and propose to employ Hidden Markov Models (HMMs) to model them.
  arXiv  Detail & Related papers  (2024-10-01T13:53:44Z)
• A Fixed-Point Approach for Causal Generative Modeling [20.88890689294816]
  We propose a novel formalism for describing Structural Causal Models (SCMs) as fixed-point problems on causally ordered variables.<n>We establish the weakest known conditions for their unique recovery given the topological ordering (TO)
  arXiv  Detail & Related papers  (2024-04-10T12:29:05Z)
• TSNAT: Two-Step Non-Autoregressvie Transformer Models for Speech Recognition [69.68154370877615]
  The non-autoregressive (NAR) models can get rid of the temporal dependency between the output tokens and predict the entire output tokens in at least one step. To address these two problems, we propose a new model named the two-step non-autoregressive transformer(TSNAT) The results show that the TSNAT can achieve a competitive performance with the AR model and outperform many complicated NAR models.
  arXiv  Detail & Related papers  (2021-04-04T02:34:55Z)
• A Study of Non-autoregressive Model for Sequence Generation [147.89525760170923]
  Non-autoregressive (NAR) models generate all the tokens of a sequence in parallel. We propose knowledge distillation and source-target alignment to bridge the gap between AR and NAR models.
  arXiv  Detail & Related papers  (2020-04-22T09:16:09Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.