mlsanityData to 7 Oct 2026

SDAR-VL: Stable and Efficient Block-wise Diffusion for Vision-Language Understanding

arXiv 2512.14068 ↗First posted 16 Dec 2025cs.CV cs.AIcited by 12

On the map: Multimodal LLMs / vision-language models → Image editing and quality assessment

also in 1 more
Abstract (arXiv)

Block-wise discrete diffusion offers an attractive balance between parallel generation and causal dependency modeling, making it a promising backbone for vision-language modeling. However, its practical adoption has been limited by high training cost, slow convergence, and instability, which have so far kept it behind strong autoregressive (AR) baselines. We present SDAR-VL, the first systematic application of block-wise discrete diffusion to large-scale vision-language understanding (VLU), together with an integrated framework for efficient and stable training. This framework unifies three components: (1) Asynchronous Block-wise Noise Scheduling to diversify supervision within each batch; (2) Effective Mask Ratio Scaling for unbiased loss normalization under stochastic masking; and (3) a Progressive Beta Noise Curriculum that increases effective mask coverage while preserving corruption diversity. Experiments on 21 single-image, multi-image, and video benchmarks show that SDAR-VL consistently improves training efficiency, convergence stability, and task performance over conventional block diffusion. On this evaluation suite, SDAR-VL sets a new state of the art among diffusion-based vision-language models and, under matched settings, matches or surpasses strong AR baselines such as LLaVA-OneVision as well as the global diffusion baseline LLaDA-V, establishing block-wise diffusion as a practical backbone for VLU.

Key facts. Cited by 12 papers in the map (data through 7 Oct 2026). Placed in Image editing and quality assessment, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 34 of its 43 references that are in the map. First posted on arXiv 16 Dec 2025.

Builds on

34 of 43 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Discrete Diffusion in Large Language and Multimodal Models: A Survey16 Jun
  2. 2025Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decoding22 May
  3. 2025LaViDa: A Large Diffusion Language Model for Multimodal Understanding22 May

Other references · 31

  1. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,151
  2. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,044
  3. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answeringcited by 1,881
  4. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,665
  5. 2022ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoningcited by 1,449
  6. 2025InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiencycited by 1,419
  7. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,288
  8. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilitiescited by 1,221
  9. 2025SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Featurescited by 1,127
  10. 2023SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehensioncited by 929
  11. 2024Simple and Effective Masked Diffusion Language Modelscited by 800
  12. 2016A Diagram Is Worth A Dozen Imagescited by 793
19 more
  1. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 712
  2. 2024BLINK: Multimodal Large Language Models Can See but Not Perceivecited by 529
  3. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Datasetcited by 516
  4. 2025Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decodingcited by 404
  5. 2025Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Modelscited by 385
  6. 2025R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalizationcited by 364
  7. 2025MMaDA: Multimodal Large Diffusion Language Modelscited by 241
  8. 2025LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuningcited by 145
  9. 2025dKV-Cache: The Cache for Diffusion Language Modelscited by 140
  10. 2024Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planningcited by 135
  11. 2025Fast-dLLM v2: Efficient Block-Diffusion LLMcited by 128
  12. 2024MAmmoTH-VL: Eliciting Multimodal Reasoning with Instruction Tuning at Scalecited by 125
  13. 2025Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcingcited by 111
  14. 2025Lumina-DiMOO: An Omni Diffusion Large Language Model for Multi-Modal Generation and Understandingcited by 83
  15. 2025Muddit: Liberating Generation Beyond Text-to-Image with a Unified Discrete Diffusion Modelcited by 52
  16. 2025VisualWebInstruct: Scaling up Multimodal Instruction Data through Web Searchcited by 34
  17. 2025Diffusion LLM with Native Variable Generation Lengths: Let [EOS] Lead the Waycited by 22
  18. 2025How Efficient Are Diffusion Language Models? A Critical Examination of Efficiency Evaluation Practicescited by 10
  19. 2025From Denoising to Refining: A Corrective Framework for Vision-Language Diffusion Modelcited by 9

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 12.

12 papers.

  1. 2026DMax: Aggressive Parallel Decoding for dLLMscited by 14
  2. 2026Fast-dVLM: Efficient Block-Diffusion VLM via Direct Conversion from Autoregressive VLMcited by 8
  3. 2026MinerU-Diffusion: Rethinking Document OCR as Inverse Rendering via Diffusion Decodingcited by 7
  4. 2026Seeing the End at Step Zero: Accelerating Diffusion MLLMs via MLP Sparsity-Aware Truncationcited by 4
  5. 2026ECHO: Efficient Chest X-ray Report Generation with One-step Block Diffusioncited by 3
  6. 2026VidLaDA: Bidirectional Diffusion Large Language Models for Efficient Video Understandingcited by 2
  7. 2026STRIDE: When to Speak Meets Sequence Denoising for Streaming Video Understandingcited by 2
  8. 2026BARD: Bridging AutoRegressive and Diffusion Vision-Language Models Via Highly Efficient Progressive Block Merging and Stage-Wise Distillationcited by 2
  9. 2026TACG: Trajectory-Aware Commit Gating for Diffusion Language Model Decodingcited by 2
  10. 2026Block3D: Efficient Text-to-3D Generation via Block-Wise Diffusioncited by 1
  11. 2026LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agentscited by 0
  12. 2026d-OPD: Future-Aware On-Policy Distillation for Block Diffusion Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026BlockVLA: Accelerating Autoregressive VLA via Block Diffusion Finetuning
  2. 2025Vision-Language-Vision Auto-Encoder: Scalable Knowledge Distillation from Diffusion Models
  3. 2025Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies
  4. 2025DiffusionVL: Translating Any Autoregressive Models into Diffusion Vision Language Models
  5. 2026High-Fidelity Text-to-Image Generation from Pre-Trained Vision-Language Models via Distribution-Conditioned Diffusion Decoding
  6. 2026Reliability Challenges in Diffusion Vision-Language Models
  7. 2026Mitigating Mask Prior Drift and Positional Attention Collapse in Large Diffusion Vision-Language Models
  8. 2025Unified Diffusion VLA: Vision-Language-Action Model via Joint Discrete Denoising Diffusion Process