mlsanityData to 7 Oct 2026

Context-Aware Cluster Decoding: Semantic Anchor-Driven Coherence in dMLLMs

arXiv 2608.22367 ↗First posted 23 Aug 2026cs.CLcited by 0

On the map: Multimodal LLMs / vision-language models → Visual token pruning and efficiency

Abstract (arXiv)

Diffusion multimodal large language models (dMLLMs) frequently produce long-form outputs marred by semantic drift and repetition, with quality generally degrading as output length increases. We identify two structural deficiencies in existing decoding methods as primary drivers of these failures: confidence-based scoring ignores decoded-neighbor support, and block partitioning prevents access to high-readiness semantic anchors, together causing tokens to be committed before their local context is sufficiently established. We propose ours (Context-Aware Cluster Decoding), a training-free decoding method that scores each masked position by a multiplicative composite of softmax confidence and neighbor proximity, promoting contextually ready tokens above isolated candidates while suppressing low-confidence positional noise, operating block-free to keep high-readiness anchors globally accessible. ours further applies architecture-aware calibration to handle confidence heterogeneity induced by diverse visual integration strategies. Experiments on three dMLLMs across four benchmarks demonstrate consistent quality gains and hallucination reduction over Original, with larger gains in several longer generation settings, highlighting the importance of neighbor support and visual integration strategy for future dMLLM decoding method design. Our code is openly available at https://github.com/zhaoyk-sysu/CACD-dMLLM.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 38 of its 55 references that are in the map. First posted on arXiv 23 Aug 2026.

Builds on

38 of 55 references.

Closest ancestors · 1

Same direction, newest first

  1. 2026Context Tokens are Anchors: Understanding the Repetition Curse in dMLLMs from an Information Flow Perspective28 Jan

Other references · 37

  1. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answeringcited by 1,881
  2. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contextscited by 1,582
  3. 2018Object Hallucination in Image Captioningcited by 661
  4. 2024Simplified and Generalized Masked Diffusion for Discrete Datacited by 606
  5. 2025Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decodingcited by 404
  6. 2024Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Datacited by 371
  7. 2025Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusionscited by 208
  8. 2024Scaling up Masked Diffusion Models on Textcited by 195
  9. 2025Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inferencecited by 173
  10. 2025LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuningcited by 145
  11. 2025Fast-dLLM v2: Efficient Block-Diffusion LLMcited by 128
  12. 2025Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcingcited by 111
25 more
  1. 2025Accelerating Diffusion LLMs via Adaptive Parallel Decodingcited by 108
  2. 2025Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decodingcited by 101
  3. 2024Beyond Autoregression: Fast LLMs via Self-Distillation Through Timecited by 73
  4. 2025Diffusion Language Models Know the Answer Before Decodingcited by 68
  5. 2025Discrete Diffusion in Large Language and Multimodal Models: A Surveycited by 61
  6. 2024Speculative Diffusion Decoding: Accelerating Language Generation through Diffusioncited by 57
  7. 2025Unified Multimodal Discrete Diffusioncited by 46
  8. 2025Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principlescited by 43
  9. 2025ParallelBench: Understanding the Trade-offs of Parallel Decoding in Diffusion LLMscited by 42
  10. 2025WeDLM: Reconciling Diffusion Language Models with Standard Causal Attention for Fast Inferencecited by 38
  11. 2025AdaBlock-dLLM: Semantic-Aware Diffusion LLM Inference via Adaptive Block Sizecited by 30
  12. 2025CDLM: Consistency Diffusion Language Models For Faster Samplingcited by 20
  13. 2025Empirical Analysis of Decoding Biases in Masked Diffusion Modelscited by 19
  14. 2025Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbonecited by 18
  15. 2025Mask Tokens as Prophet: Fine-Grained Cache Eviction for Efficient dLLM Inferencecited by 17
  16. 2025Parallel Sampling from Masked Diffusion Models via Conditional Independence Testingcited by 13
  17. 2025Beyond Confidence: Adaptive and Coherent Decoding for Diffusion Language Modelscited by 9
  18. 2025Next Semantic Scale Prediction via Hierarchical Diffusion Language Modelscited by 7
  19. 2026Focus-dLLM: Accelerating Long-Context Diffusion LLM Inference via Confidence-Guided Context Focusingcited by 6
  20. 2026Stop the Flip-Flop: Context-Preserving Verification for Fast Revocable Diffusion Decodingcited by 6
  21. 2025Finish First, Perfect Later: Test-Time Token-Level Cross-Validation for Diffusion Large Language Modelscited by 5
  22. 2025D3ToM: Decider-Guided Dynamic Token Merging for Accelerating Diffusion MLLMscited by 5
  23. 2025WavefrontDiffusion: Dynamic Decoding Schedule for Improved Reasoningcited by 4
  24. 2025Review, Remask, Refine (R3): Process-Guided Block Diffusion for Text Generationcited by 3
  25. 2026Breaking Block Boundaries: Anchor-based History-stable Decoding for Diffusion Large Language Modelscited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Information-Guided Frontier Decoding: Contextual Utility-Driven Commitment in dMLLMs
  2. 2026Cluster-Level Attention-Guided Parallel Decoding for Masked Diffusion Language Models
  3. 2026Visual-Redundancy-Controlled Parallel Decoding for Diffusion-Based Multimodal Large Language Models
  4. 2025A Comprehensive Study on Visual Token Redundancy for Discrete Diffusion-based Multimodal Large Language Models
  5. 2026Seeing the End at Step Zero: Accelerating Diffusion MLLMs via MLP Sparsity-Aware Truncation
  6. 2026Visual Information-Guided Parallel Decoding for Diffusion Multimodal Large Language Models
  7. 2026DAWN: Dependency-Aware Fast Inference for Diffusion LLMs
  8. 2026Focus on the Core: Empowering Diffusion Large Language Models by Self-Contrast