mlsanityData to 14 Oct 2026

DiffuMask: Diffusion Language Model for Token-level Prompt Pruning

arXiv 2604.06627 ↗First posted 8 Apr 2026cs.CLcited by 1

On the map: LLM reasoning → Test-time compute and efficient thinking

also in 2 more
Abstract (arXiv)

In-Context Learning and Chain-of-Thought prompting improve reasoning in large language models (LLMs). These typically come at the cost of longer, more expensive prompts that may contain redundant information. Prompt compression based on pruning offers a practical solution, yet existing methods rely on sequential token removal which is computationally intensive. We present DiffuMask, a diffusion-based framework integrating hierarchical shot-level and token-level pruning signals, that enables rapid and parallel prompt pruning via iterative mask prediction. DiffuMask substantially accelerates the compression process via masking multiple tokens in each denoising step. It offers tunable control over retained content, preserving essential reasoning context and achieving up to 80% prompt length reduction. Meanwhile, it maintains or improves accuracy across in-domain, out-of-domain, and cross-model settings. Our results show that DiffuMask provides a generalizable and controllable framework for prompt compression, facilitating faster and more reliable in-context reasoning in LLMs.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 27 of its 39 references that are in the map. First posted on arXiv 8 Apr 2026.

Builds on

27 of 39 references.

References · 27

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 9,122
  3. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  4. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,455
  5. 2015Character-level Convolutional Networks for Text Classificationcited by 2,330
  6. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
  7. 2021What Makes Good In-Context Examples for GPT-3?cited by 1,195
  8. 2022Diffusion-LM Improves Controllable Text Generationcited by 1,027
  9. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 745
  10. 2025Dream 7B: Diffusion Large Language Modelscited by 544
  11. 2022DiffuSeq: Sequence to Sequence Text Generation with Diffusion Modelscited by 450
  12. 2023Learning to Compress Prompts with Gist Tokenscited by 390
15 more
  1. 2024Scaling Diffusion Language Models via Adaptation from Autoregressive Modelscited by 274
  2. 2024Scaling up Masked Diffusion Models on Textcited by 199
  3. 2023Walking Down the Memory Maze: Beyond Context Limit through Interactive Readingcited by 133
  4. 2022Prompt Compression and Contrastive Conditioning for Controllability and Toxicity Reduction in Language Modelscited by 97
  5. 2022An Information-theoretic Approach to Prompt Engineering Without Ground Truth Labelscited by 89
  6. 2016Yelp Dataset Challenge: Review Rating Predictioncited by 84
  7. 2024Prompt Compression for Large Language Models: A Surveycited by 67
  8. 2024CompAct: Compressing Retrieved Documents Actively for Question Answeringcited by 64
  9. 2025Diffusion Beats Autoregressive in Data-Constrained Settingscited by 63
  10. 2023OpenICL: An Open-Source Framework for In-context Learningcited by 51
  11. 2023Discrete Prompt Compression with Reinforcement Learningcited by 42
  12. 2024TACO-RL: Task Aware Prompt Compression Optimization with Reinforcement Learningcited by 22
  13. 2025Evolving Prompts In-Context: An Open-ended, Self-replicating Perspectivecited by 8
  14. 2025Can LLMs Narrate Tabular Data? An Evaluation Framework for Natural Language Representations of Text-to-SQL System Outputscited by 7
  15. 2026Barriers to Discrete Reasoning with Transformers: A Survey Across Depth, Exactness, and Bandwidthcited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026JTPRO: A Joint Tool-Prompt Reflective Optimization Framework for Language Agentscited by 4

Similar, not cited

Close in wording, no citation either way

  1. 2026Diffusion Language Models Are Natively Length-Aware
  2. 2026Lexical Prompt Compression for Large Language Models: A Training-Free, Deterministic Pipeline with Empirical Pareto Analysis Across Eleven Task Categories
  3. 2025PromptDistill: Query-based Selective Token Retention in Intermediate Layers for Efficient Large Language Model Inference
  4. 2025Dynamic Compressing Prompts for Efficient Inference of Large Language Models
  5. 2026Unlocking Prompt Infilling Capability for Diffusion Language Models
  6. 2024LanguaShrink: Reducing Token Overhead with Psycholinguistics
  7. 2023Just CHOP: Embarrassingly Simple LLM Compression
  8. 2025Diff-Prompt: Diffusion-Driven Prompt Generator with Mask Supervision