mlsanityData to 14 Oct 2026

Don't Settle Too Early: Self-Reflective Remasking for Diffusion Language Models

arXiv 2509.23653 ↗First posted 28 Sep 2025cs.CLcited by 37

On the map: Diffusion models → Discrete and text diffusion

Abstract (arXiv)

Mask-based Diffusion Language Models (DLMs) struggle to revise incorrect tokens: once a token is generated, it typically remains fixed. The key challenge is to identify potential errors in the inputs. In this paper, we propose emphunderlineRemasking-underlineenabled underlineDiffusion Language Model (RemeDi, a mask-based DLM that introduces remasking as another fundamental mechanism, enabling more flexible text refinement in diffusion-based text generation. To achieve this, RemeDi jointly predicts token distributions and per-token confidence scores at each step. The confidence scores determine which tokens to be unmasked after the current step, allowing the model to identify tokens with low quality and remask them. These remasked tokens can be resampled with richer context in subsequent steps. We design a remask-aware pipeline to train this ability, including supervised fine-tuning which teaches the model to detect and remask incorrect tokens in addition to predict mask tokens, and reinforcement learning which optimizes full generation trajectories toward higher rewards. Experiments show that RemeDi achieves the state-of-the-art results among open-source DLMs on multiple datasets.

Key facts. Cited by 37 papers in the map (data through 14 Oct 2026). Placed in Discrete and text diffusion, one of 14 directions in Diffusion models. Builds on 45 of its 50 references that are in the map. First posted on arXiv 28 Sep 2025.

Builds on

45 of 50 references.

Closest ancestors · 12

Same direction, newest first

  1. 2025Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcing8 Aug
  2. 2025Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference4 Aug
  3. 2025wd1: Weighted Policy Optimization for Reasoning in Diffusion Language Models7 Jul
  4. 2025CtrlDiff: Boosting Large Diffusion Language Models with Dynamic Block Prediction and Controllable Generation20 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. 2025Generalized Interpolating Discrete Diffusion6 Mar
  8. 2025Remasking Discrete Diffusion Models with Inference-Time Scaling1 Mar
  9. 2025Large Language Diffusion Models14 Feb
  10. 2024Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data6 Jun
  11. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution25 Oct
  12. 2023A Reparameterized Discrete Diffusion Model for Text Generation11 Feb

Other references · 33

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  5. 2023Adding Conditional Control to Text-to-Image Diffusion Modelscited by 5,503
  6. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  7. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,886
  8. 2019HellaSwag: Can a Machine Really Finish Your Sentence?cited by 4,100
  9. 2023Let's Verify Step by Stepcited by 3,789
  10. 2021Program Synthesis with Large Language Modelscited by 3,637
  11. 2023Code Llama: Open Foundation Models for Codecited by 2,139
  12. 2023Instruction-Following Evaluation for Large Language Modelscited by 1,228
21 more
  1. 2024DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligencecited by 1,181
  2. 2024Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluatorscited by 897
  3. 2023MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Modelscited by 735
  4. 2024DeepSeek LLM: Scaling Open-Source Language Models with Longtermismcited by 642
  5. 2023MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuningcited by 487
  6. 2022A Continuous Time Framework for Discrete Denoising Modelscited by 451
  7. 2024Skywork-Reward: Bag of Tricks for Reward Modeling in LLMscited by 327
  8. 2025LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Modelscited by 285
  9. 2025DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoningcited by 280
  10. 2025Skywork-Reward-V2: Scaling Preference Data Curation via Human-AI Synergycited by 214
  11. 2025d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learningcited by 186
  12. 2024OpenCoder: The Open Cookbook for Top-Tier Code Large Language Modelscited by 130
  13. 2025KodCode: A Diverse, Challenging, and Verifiable Synthetic Dataset for Codingcited by 119
  14. 2025LeetCodeDataset: A Temporal Dataset for Robust Evaluation and Efficient Training of Code LLMscited by 70
  15. 2025Edit Flows: Flow Matching with Edit Operationscited by 55
  16. 2025Infinity Instruct: Scaling Instruction Selection and Synthesis to Enhance Language Modelscited by 50
  17. 2024The Perfect Blend: Redefining RLHF with Mixture of Judgescited by 43
  18. 2025Esoteric Language Models: A Family of Any-Order Diffusion LLMscited by 34
  19. 2025Set Block Decoding is a Language Model Inference Acceleratorcited by 23
  20. 2025Unifying Autoregressive and Diffusion-Based Sequence Generationcited by 17
  21. 2025A Comprehensive Review on Noise Control of Diffusion Modelcited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 5; 2026: 32.

37 papers.

Sort:

By area:

  1. 2025Fine-Tuning Masked Diffusion for Provable Self-Correctioncited by 39
  2. 2025Learning Unmasking Policies for Diffusion Language Modelscited by 30
  3. 2025Diffusion LLM with Native Variable Generation Lengths: Let [EOS] Lead the Waycited by 22
  4. 2026Learn from Your Mistakes: Self-Correcting Masked Diffusion Modelscited by 13
  5. 2025Fast-Decoding Diffusion Language Models via Progress-Aware Confidence Schedulescited by 11
  6. 2026CORE: Context-Robust Remasking for Diffusion Language Modelscited by 10
  7. 2026DUEL: Exact Likelihood for Masked Diffusion via Deterministic Unmaskingcited by 9
  8. 2025Guided Star-Shaped Masked Diffusioncited by 5
  9. 2026Provably adaptive sampling with uniform and remasking discrete diffusion modelscited by 4
  10. 2026Search or Accelerate: Confidence-Switched Position Beam Search for Diffusion Language Modelscited by 3
  11. 2026Thinking Diffusion: Penalize and Guide Visual-Grounded Reasoning in Diffusion Multimodal Language Modelscited by 3
  12. 2026Training-Free Self-Correction for Multimodal Masked Diffusion Modelscited by 2
25 more
  1. 2026Reasoning on the Manifold: Bidirectional Consistency for Self-Verification in Diffusion Language Modelscited by 2
  2. 2026Edit-Based Refinement for Parallel Masked Diffusion Language Modelscited by 2
  3. 2026Forward-Free Diffusion Language Models with BPTT-Free Looped Refinementcited by 2
  4. 2026Self-Generated Error Training for Token Editing in Diffusion Language Modelscited by 2
  5. 2026Don't Commit Alone: Joint Token Commitment in Diffusion Language Modelscited by 2
  6. 2026BackPlay: Head-Only Look-Back Self-Correction for Diffusion Language Modelscited by 1
  7. 2026Reversible Diffusion Decoding for Diffusion Language Modelscited by 1
  8. 2026Remask, Don't Replace: Token-to-Mask Refinement in Diffusion Language Modelscited by 1
  9. 2026The Path Matters: Learning a Token-Commitment Policy for Diffusion Language Modelscited by 1
  10. 2026Re-evaluating Confidence Remasking in Masked Diffusion Language Modelscited by 1
  11. 2026Concurrent Image Understanding and Generation: Self-Correcting Coupled Markov Jump Processescited by 1
  12. 2026CaRE Compute-aware Remasking Evaluation Protocol for Masked Diffusion Language Modelscited by 1
  13. 2026From Interface to Inference: Eliciting Any-Order Inference from Any-Order Modelscited by 1
  14. 2026Reconsidering Positional Supervision in Masked Diffusion Language Model Trainingcited by 0
  15. 2026Controllable Accent Normalization via Discrete Diffusioncited by 0
  16. 2026Focus on the Core: Empowering Diffusion Large Language Models by Self-Contrastcited by 0
  17. 2026Targeted Remasking: Replacing Token Editing with Token-to-Mask Refinement in Discrete Diffusion Language Modelscited by 0
  18. 2026Adaptive Order Policies for Masked Diffusioncited by 0
  19. 2026Multi-Turn Reflective Masking Elicits Reasoning in Mask Diffusion Modelscited by 0
  20. 2026Archer: Adaptive Reuse of Cached Hidden States for Efficient Rollback in Diffusion Language Modelscited by 0
  21. 2026Continuous Interaction Diffusion: A Diffusion-Native Runtime for Asynchronous Tool-Augmented Reasoningcited by 0
  22. 2026A Ticket from Marginals to Joints: Coupled-Noise Distillation for One-Step Block Generation in Diffusion Language Modelscited by 0
  23. 2026LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Modelscited by 0
  24. 2026Blackboard Intelligence Can Surpass Autoregressive on Globally Constrained Problemscited by 0
  25. 2026ITC-MoE: Importance-guided Token-aware Compression for MoE Diffusion Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026NAVIRA: Decoupled Stochastic Remasking for Masked Diffusion Language Models
  2. 2026CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models
  3. 2025Review, Remask, Refine (R3): Process-Guided Block Diffusion for Text Generation
  4. 2026Beyond Masks: Efficient, Flexible Diffusion Language Models via Deletion-Insertion Processes
  5. 2025STaRR: Spatial-Temporal Token-Dynamics-Aware Responsive Remasking for Diffusion Language Models
  6. 2026Revise, Don't Freeze: Sampler-Matched Training for Self-Correcting Masked Diffusion Language Models
  7. 2025Saber: An Efficient Sampling with Adaptive Acceleration and Backtracking Enhanced Remasking for Diffusion Language Model
  8. 2026Just on Time: Token-Level Early Stopping for Diffusion Language Models