mlsanityData to 14 Oct 2026

A Tale of Two Temperatures: Simple, Efficient, and Diverse Sampling from Diffusion Language Models

arXiv 2604.09921 ↗First posted 10 Apr 2026cs.LGcited by 5

On the map: Diffusion models → Discrete and text diffusion

Abstract (arXiv)

Much work has been done on designing fast and accurate sampling for diffusion language models (dLLMs). However, these efforts have largely focused on the tradeoff between speed and quality of individual samples; how to additionally ensure diversity across samples remains less well understood. In this work, we show that diversity can be increased by using softened, tempered versions of familiar confidence-based remasking heuristics, retaining their computational benefits and offering simple implementations. We motivate this approach by introducing an idealized formal model of fork tokens and studying the impact of remasking on the expected entropy at the forks. Empirically, the proposed tempered heuristics close the exploration gap (pass@k) between existing confidence-based and autoregressive sampling, hence outperforming both when controlling for cost (pass@NFE). We further study how the increase in diversity translates to downstream post-training and test-time compute scaling. Overall, our findings demonstrate that simple, efficient, and diverse sampling from dLLMs is possible.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Placed in Discrete and text diffusion, one of 14 directions in Diffusion models. Builds on 52 of its 56 references that are in the map. First posted on arXiv 10 Apr 2026.

Builds on

52 of 56 references.

Closest ancestors · 27

Same direction, newest first

  1. 2026Free Lunch for Pass@k? Low Cost Diverse Sampling for Diffusion Language Models5 Mar
  2. 2026DUEL: Exact Likelihood for Masked Diffusion via Deterministic Unmasking2 Mar
  3. 2026The Design Space of Tri-Modal Masked Diffusion Models25 Feb
  4. 2026Time-Annealed Perturbation Sampling: Diverse Generation for Diffusion Language Models30 Jan
  5. 2026Improving Diffusion Language Model Decoding through Joint Search in Generation Order and Token Space28 Jan
  6. 2025LLaDA2.0: Scaling Up Diffusion Language Models to 100B10 Dec
  7. 2025Learning Unmasking Policies for Diffusion Language Models9 Dec
  8. 2025KLASS: KL-Guided Fast Inference in Masked Diffusion Models7 Nov
  9. 2025SPG: Sandwiched Policy Gradient for Masked Diffusion Language Models10 Oct
  10. 2025Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization9 Oct
  11. 2025SDAR: A Synergistic Diffusion-AutoRegression Paradigm for Scalable Sequence Generation7 Oct
  12. 2025DiFFPO: Training Diffusion LLMs to Reason Fast and Furious via Reinforcement Learning2 Oct
15 more
  1. 2025d2: Improving Reasoning in Diffusion Language Models via Trajectory Likelihood Estimation25 Sep
  2. 2025Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models8 Sep
  3. 2025Dream 7B: Diffusion Large Language Models21 Aug
  4. 2025Accelerated Sampling from Masked Diffusion Models via Entropy Bounded Unmasking30 May
  5. 2025Reinforcing the Diffusion Chain of Lateral Thought with Diffusion Language Models15 May
  6. 2025Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models12 Mar
  7. 2025Large Language Diffusion Models14 Feb
  8. 2025Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions10 Feb
  9. 2024Scaling up Masked Diffusion Models on Text24 Oct
  10. 2024Masked Diffusion Models are Secretly Time-Agnostic Masked Models and Exploit Inaccurate Categorical Sampling4 Sep
  11. 2024Simple and Effective Masked Diffusion Language Models11 Jun
  12. 2024Simplified and Generalized Masked Diffusion for Discrete Data6 Jun
  13. 2024Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data6 Jun
  14. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution25 Oct
  15. 2021Autoregressive Diffusion Models5 Oct

Other references · 25

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  2. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  3. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  4. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,058
  5. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  6. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  7. 2021Program Synthesis with Large Language Modelscited by 3,637
  8. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
  9. 2022MaskGIT: Masked Generative Image Transformercited by 1,099
  10. 2025Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?cited by 982
  11. 2021Argmax Flows and Multinomial Diffusion: Learning Categorical Distributionscited by 574
  12. 2025Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoningcited by 559
13 more
  1. 2025The Entropy Mechanism of Reinforcement Learning for Reasoning Language Modelscited by 485
  2. 2022A Continuous Time Framework for Discrete Denoising Modelscited by 451
  3. 2025Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decodingcited by 420
  4. 2025Reasoning with Exploration: An Entropy Perspectivecited by 281
  5. 2025MMaDA: Multimodal Large Diffusion Language Modelscited by 250
  6. 2023Is Self-Repair a Silver Bullet for Code Generation?cited by 197
  7. 2025d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learningcited by 186
  8. 2019SPoC: Search-based Pseudocode to Codecited by 176
  9. 2022Concrete Score Matching: Generalized Score Matching for Discrete Datacited by 148
  10. 2024AceMath: Advancing Frontier Math Reasoning with Post-Training and Reward Modelingcited by 77
  11. 2024Forking Paths in Neural Text Generationcited by 34
  12. 2025Low-probability Tokens Sustain Exploration in Reinforcement Learning with Verifiable Rewardcited by 25
  13. 2026Entropy-Preserving Reinforcement Learningcited by 11

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Re-evaluating Confidence Remasking in Masked Diffusion Language Modelscited by 1
  2. 2026From Interface to Inference: Eliciting Any-Order Inference from Any-Order Modelscited by 1
  3. 2026Low-Confidence Remasking Traps Flexibility: Realizing Arbitrary-Order Potential for Diverse Rollouts in Diffusion LLMscited by 0
  4. 2026Self-Repulsive Sampling for Diffusion Language Modelscited by 0
  5. 2026Fork-dLLM: Avoiding the Flexibility Trap in Diffusion Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Diffusion Language Models are Provably Optimal Parallel Samplers
  2. 2026Adaptation to Intrinsic Dependence in Diffusion Language Models
  3. 2026Scaling Beyond Masked Diffusion Language Models
  4. 2025Breaking AR's Sampling Bottleneck: Provable Acceleration via Diffusion Language Models
  5. 2026S3: Stratified Scaling Search for Test-Time in Diffusion Language Models
  6. 2026LESS Is More: Mutual-Stability Sampling for Diffusion Language Models
  7. 2026CaRE Compute-aware Remasking Evaluation Protocol for Masked Diffusion Language Models
  8. 2026CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models