mlsanityData to 14 Oct 2026

Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models

arXiv 2609.00355 ↗First posted 31 Aug 2026cs.AI cs.CL cs.CVcited by 0

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

also in 1 more
Abstract (arXiv)

Speculative decoding accelerates generation without changing its output, but on vision-language models (VLMs) a self-reinforcing cycle holds it back. Because an autoregressive drafter pays a sequential pass for each drafted token, it must stay small and can ill afford to attend to the image at each pass. Prior work therefore compresses or hides the image, leaving the drafter weakest on the text the image determines. We present GLANCE, a one-pass block drafter that breaks this cycle on an unmodified VLM target. Its block-diffusion head drafts a whole block in one forward pass over the target's already fused vision-language states, reading the multimodal context once, however deep the draft. The target verifies a wide candidate tree in one pass and commits exactly its greedy output. In one production engine at a fixed round budget, GLANCE decodes up to 3.05 times faster than autoregressive decoding and outpaces the production EAGLE3-VL head on average and by about 11% on grounded tasks. An entropy law explains when drafting pays, predicting the longest accepted blocks on grounded tasks, where the target's next-token entropy is lowest. Our code is available at https://github.com/js-lee-AI/GLANCE.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 46 of its 72 references that are in the map. First posted on arXiv 31 Aug 2026.

Builds on

46 of 72 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025ViSpec: Accelerating Vision-Language Models with Vision-Aware Speculative Decoding17 Sep
  2. 2025Spec-LLaVA: Accelerating Vision-Language Models with Dynamic Tree-Based Speculative Decoding15 Sep
  3. 2025DREAM: Drafting with Refined Target Features and Entropy-Adaptive Cross-Attention Fusion for Multimodal Speculative Decoding25 May
  4. 2025SpecFLASH: A Latent-Guided Semi-autoregressive Speculative Decoding Framework for Efficient Multimodal Generation19 May
  5. 2024On Speculative Decoding for Multimodal Large Language Models13 Apr

Other references · 41

  1. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  2. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  3. 2025Qwen2.5-VL Technical Reportcited by 5,587
  4. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  5. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
  6. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answeringcited by 1,919
  7. 2019Towards VQA Models That Can Readcited by 1,718
  8. 2022Fast Inference from Transformers via Speculative Decodingcited by 1,653
  9. 2022ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoningcited by 1,479
  10. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,317
  11. 2023SGLang: Efficient Execution of Structured Language Model Programscited by 1,141
  12. 2023Accelerating Large Language Model Decoding with Speculative Samplingcited by 1,006
29 more
  1. 2024Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Headscited by 843
  2. 2024WildChat: 1M ChatGPT Interaction Logs in the Wildcited by 629
  3. 2024EAGLE: Speculative Sampling Requires Rethinking Feature Uncertaintycited by 578
  4. 2024EAGLE-2: Faster Inference of Language Models with Dynamic Draft Treescited by 370
  5. 2023SpecInfer: Accelerating Generative Large Language Model Serving with Tree-based Speculative Inference and Verificationcited by 365
  6. 2025EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Testcited by 331
  7. 2024Break the Sequential Dependency of LLM Inference Using Lookahead Decodingcited by 298
  8. 2024Unlocking Efficiency in Large Language Model Inference: A Comprehensive Survey of Speculative Decodingcited by 269
  9. 2026DFlash: Block Diffusion for Flash Speculative Decodingcited by 135
  10. 2024Hydra: Sequentially-Dependent Draft Heads for Medusa Decodingcited by 117
  11. 2024Sequoia: Scalable, Robust, and Hardware-aware Speculative Decodingcited by 103
  12. 2024SpecDec++: Boosting Speculative Decoding via Adaptive Candidate Lengthscited by 75
  13. 2024OPT-Tree: Speculative Decoding with Adaptive Draft Tree Structurecited by 67
  14. 2024Speculative Diffusion Decoding: Accelerating Language Generation through Diffusioncited by 57
  15. 2025DiffuSpec: Unlocking Diffusion Language Models for Speculative Decodingcited by 42
  16. 2024A Theoretical Perspective for Speculative Decoding Algorithmcited by 29
  17. 2026Accelerating Speculative Decoding with Block Diffusion Draft Treescited by 29
  18. 2024AdaEDL: Early Draft Stopping for Speculative Decoding of Large Language Models via an Entropy-based Lower Bound on Token Acceptance Probabilitycited by 25
  19. 2025SpecVLM: Enhancing Speculative Decoding of Video LLMs via Verifier-Guided Token Pruningcited by 20
  20. 2026D-PACE: Dynamic Position-Aware Cross-Entropy for Parallel Speculative Draftingcited by 17
  21. 2026SpecForge: A Flexible and Efficient Open-Source Training Framework for Speculative Decodingcited by 16
  22. 2025DEER: Draft with Diffusion, Verify with Autoregressive Modelscited by 9
  23. 2026Fast-dVLM: Efficient Block-Diffusion VLM via Direct Conversion from Autoregressive VLMcited by 9
  24. 2026xPress: Parallel Refinement for Diffusion Drafters in Speculative Decodingcited by 8
  25. 2026Cost-Aware Diffusion Draft Trees for Speculative Decodingcited by 6
  26. 2026DARTree: Speculative Diffusion Decoding with Autoregressive Draft Treescited by 6
  27. 2026Fast-dDrive: Efficient Block-Diffusion VLM for Autonomous Drivingcited by 5
  28. 2026Acceptance Dynamics Across Cognitive Domains in Speculative Decodingcited by 1
  29. 2026WhiFlash: Accelerating Speculative Decoding with Token-Level Cross-Paradigm Routingcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026SP2ec: Adaptive Self-Speculative Decoding for Vision-Language Models
  2. 2026HAWK: Rethinking Multimodal Drafting for Speculative Decoding
  3. 2026LongSpark: Efficient speculative decoding with a fixed-cost parallel drafter
  4. 2025HiViS: Hiding Visual Tokens from the Drafter for Speculative Decoding in Vision-Language Models
  5. 2026MMSpec: Benchmarking Speculative Decoding for Vision-Language Models
  6. 2026CASCADE: Context-Aware Relaxation for Speculative Image Decoding
  7. 2026Is Multimodal Speculative Decoding Ready for Diffusion-Based Parallel Drafting? A Survey and Empirical Diagnosis
  8. 2025MASSV: Multimodal Adaptation and Self-Data Distillation for Speculative Decoding of Vision-Language Models