mlsanityData to 7 Oct 2026

Fast autoregressive visual generation

97 papers0 this weekMedium confidence in this grouping: mostly one topic, with blurry edgesMedium confidence in this grouping: mostly one topic, with blurry edges · About

Keywords: generation, video, autoregressive, decoding, image

Key facts. Fast autoregressive visual generation: 97 papers on arXiv since 2023 (data through 7 Oct 2026). 16 of its 97 mapped papers (16%) cite “Accelerating Auto-regressive Text-to-Image Generation with Training-free Speculative Jacobi Decoding” (2024, arXiv 2410.01699). 14 of its 97 mapped papers (14%) cite “LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decoding” (2024, arXiv 2410.03355). 6 new papers 24 Sep–7 Oct.

What’s new

24 Sep – 7 Oct 2026

6 new papers, 24 Sep – 7 Oct; 0 this week.

  1. Earlier · 6
  2. LOCI: Spatial Linear Memory for Streaming World Models
  3. GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
  4. DeCoPrune: Efficient KV-Cache Pruning for Autoregressive Video Diffusion via Denoising Consistency
  5. ReCaVSR: One-Step Streaming Diffusion Video Super-Resolution with Recycled Latents and Learned Cache Routing
  6. WorldAttention: An Efficient Attention Architecture for Interactive Video World Models
  7. In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion

Foundations

Most cited by papers here

  1. Accelerating Auto-regressive Text-to-Image Generation with Training-free Speculative Jacobi Decoding2024cited by 16 here
  2. LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decoding2024cited by 14 here
  3. On Speculative Decoding for Multimodal Large Language Models2024cited by 12 here
  4. Deep Forcing: Training-Free Long Video Generation with Deep Sink and Participative Compression2025cited by 12 here
  5. ZipAR: Parallel Auto-regressive Image Generation through Spatial Locality2024cited by 10 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2022Fast Inference from Transformers via Speculative Decodingcited by 42 here
  2. 2023Accelerating Large Language Model Decoding with Speculative Samplingcited by 31 here
  3. 2023Accelerating Transformer Inference for Translation via Parallel Decodingcited by 9 here
  4. 2024EAGLE: Speculative Sampling Requires Rethinking Feature Uncertaintycited by 29 here
  5. 2024Break the Sequential Dependency of LLM Inference Using Lookahead Decodingcited by 15 here
  6. 2024EAGLE-2: Faster Inference of Language Models with Dynamic Draft Treescited by 28 here
  7. 2025EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Testcited by 13 here
  8. 2025ViSpec: Accelerating Vision-Language Models with Vision-Aware Speculative Decodingcited by 10 here
16 more it builds on, not in this chain
  1. 2022Speculative Decoding: Exploiting Speculative Execution for Accelerating Seq2seq Generationcited by 8 here
  2. 2023SpecInfer: Accelerating Generative Large Language Model Serving with Tree-based Speculative Inference and Verificationcited by 12 here
  3. 2023Scissorhands: Exploiting the Persistence of Importance Hypothesis for LLM KV Cache Compression at Test Timecited by 8 here
  4. 2023H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Modelscited by 20 here
  5. 2023SpecTr: Fast Speculative Decoding via Optimal Transportcited by 8 here
  6. 2024Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Headscited by 35 here
  7. 2024KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cachecited by 10 here
  8. 2024On Speculative Decoding for Multimodal Large Language Modelscited by 12 here
  9. 2024SnapKV: LLM Knows What You are Looking for Before Generationcited by 16 here
  10. 2024PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funnelingcited by 10 here
  11. 2024Ada-KV: Optimizing KV Cache Eviction by Adaptive Budget Allocation for Efficient LLM Inferencecited by 10 here
  12. 2024Accelerating Auto-regressive Text-to-Image Generation with Training-free Speculative Jacobi Decodingcited by 16 here
  13. 2024LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decodingcited by 14 here
  14. 2024ZipAR: Parallel Auto-regressive Image Generation through Spatial Localitycited by 10 here
  15. 2025FastVAR: Linear Visual Autoregressive Modeling via Cached Token Pruningcited by 8 here
  16. 2025Deep Forcing: Training-Free Long Video Generation with Deep Sink and Participative Compressioncited by 12 here

The 24 most cited of 290.

Browse

30 of 97, ranked 6–35

  1. 6ViSpec: Accelerating Vision-Language Models with Vision-Aware Speculative Decoding202510
  2. 7FastVAR: Linear Visual Autoregressive Modeling via Cached Token Pruning20258
  3. 8SpecVLM: Enhancing Speculative Decoding of Video LLMs via Verifier-Guided Token Pruning20258
  4. 9Head-Aware KV Cache Compression for Efficient Visual Autoregressive Modeling20257
  5. 10DREAM: Drafting with Refined Target Features and Entropy-Adaptive Cross-Attention Fusion for Multimodal Speculative Decoding20257
  6. 11LANTERN++: Enhancing Relaxed Speculative Decoding with Static Tree Drafting for Visual Auto-regressive Models20256
  7. 12Memory-Efficient Visual Autoregressive Modeling with Scale-Aware KV Cache Compression20256
  8. 13Grouped Speculative Decoding for Autoregressive Image Generation20256
  9. 14SANA-Video: Efficient Video Generation with Block Linear Diffusion Transformer20256
  10. 15Spec-LLaVA: Accelerating Vision-Language Models with Dynamic Tree-Based Speculative Decoding20255
All 30
  1. 16RandAR: Decoder-only Autoregressive Visual Generation in Random Orders20245
  2. 17Past- and Future-Informed KV Cache Policy with Salience Estimation in Autoregressive Video Diffusion20264
  3. 18Quant VideoGen: Auto-Regressive Long Video Generation via 2-Bit KV-Cache Quantization20264
  4. 19Autoregressive Image Generation with Randomized Parallel Decoding20254
  5. 20Speculative Decoding Reimagined for Multimodal Large Language Models20254
  6. 21Locality-aware Parallel Decoding for Efficient Autoregressive Image Generation20254
  7. 22Continuous Speculative Decoding for Autoregressive Image Generation20243
  8. 23Multi-Scale Local Speculative Decoding for Image Generation20262
  9. 24Fast Autoregressive Video Diffusion and World Models with Temporal Cache Compression and Sparse Attention20262
  10. 25Sparrow: Text-Anchored Window Attention with Visual-Semantic Glimpsing for Speculative Decoding in Video LLMs20262
  11. 26MMSpec: Benchmarking Speculative Decoding for Vision-Language Models20262
  12. 27ParallelVLM: Lossless Video-LLM Acceleration with Visual Alignment Aware Parallel Speculative Decoding20262
  13. 28PackForcing: Short Video Training Suffices for Long Video Sampling and Long Context Inference20262
  14. 29TIGER: Text-Conditioned Visual Gated Routing with Acceptance Alignment for Multimodal Speculative Decoding20262
  15. 30SpecFLASH: A Latent-Guided Semi-autoregressive Speculative Decoding Framework for Efficient Multimodal Generation20252
  16. 31Make It Efficient: Dynamic Sparse Attention for Autoregressive Image Generation20252
  17. 32SpecVLM: Fast Speculative Decoding in Vision-Language Models20252
  18. 33HiViS: Hiding Visual Tokens from the Drafter for Speculative Decoding in Vision-Language Models20252
  19. 34PackCache: A Training-Free Acceleration Method for Unified Autoregressive Video Generation via Compact KV-Cache20261
  20. 35HIPPO: Accelerating Video Large Language Models Inference via Holistic-aware Parallel Speculative Decoding20261