{"url":"/sota/image-generation-on-imagenet-64x64","task":{"name":"Image Generation","url":"/task/image-generation","note":null},"dataset":{"name":"ImageNet 64x64","url":null},"category":"Computer Vision","categories":["Computer Vision","Medical","Miscellaneous","Natural Language Processing"],"category_note":null,"description":"**Image Generation** (synthesis) is the task of generating new images from an existing dataset.\r\n\r\n- **Unconditional generation** refers to generating samples unconditionally from the dataset, i.e. $p(y)$\r\n- **[Conditional image generation](/task/conditional-image-generation)** (subtask) refers to generating samples conditionally from the dataset, based on a label, i.e. $p(y|x)$.\r\n\r\nIn this section, you can find state-of-the-art leaderboards for **unconditional generation**. For conditional  generation, and other types of image generations, refer to the subtasks.\r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [StyleGAN](https://github.com/NVlabs/stylegan) )</span>","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["FID","Bits per dim","NFE","Inception Score","KID"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"FID":"lower","Bits per dim":null,"NFE":null,"Inception Score":"higher","KID":"lower"}},"counts":{"rows":65,"rows_with_code":52,"rows_with_paper_page":65,"rows_dated":65,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"SIMS","metrics":{"FID":"0.92","NFE":"126"},"uses_additional_data":false,"paper_date":"2024-08-29","paper":"/paper/self-improving-diffusion-models-with","paper_url":"https://arxiv.org/abs/2408.16333v1","paper_title":"Self-Improving Diffusion Models with Synthetic Data","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"EDM2-S+DDO","metrics":{"FID":"0.97","NFE":"63"},"uses_additional_data":false,"paper_date":"2025-03-03","paper":"/paper/direct-discriminative-optimization-your-1","paper_url":"https://arxiv.org/abs/2503.01103v2","paper_title":"Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator","code":"https://github.com/nvlabs/ddo","n_code_links":1,"syntology":null},{"rank_in_archive_order":3,"model":"Uni-Instruct","metrics":{"FID":"1.02","NFE":"1"},"uses_additional_data":false,"paper_date":"2025-05-27","paper":"/paper/uni-instruct-one-step-diffusion-model-through","paper_url":"https://arxiv.org/abs/2505.20755","paper_title":"Uni-Instruct: One-step Diffusion Model through Unified Diffusion Divergence Instruction","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"SiDA-EDM","metrics":{"FID":"1.11","NFE":"1"},"uses_additional_data":false,"paper_date":"2024-10-19","paper":"/paper/adversarial-score-identity-distillation","paper_url":"https://arxiv.org/abs/2410.14919v4","paper_title":"Adversarial Score identity Distillation: Rapidly Surpassing the Teacher in One Step","code":"https://github.com/mingyuanzhou/sid","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":1}},{"rank_in_archive_order":5,"model":"GDD-I","metrics":{"FID":"1.16","NFE":"1"},"uses_additional_data":false,"paper_date":"2024-05-31","paper":"/paper/diffusion-models-are-innate-one-step","paper_url":"https://arxiv.org/abs/2405.20750v2","paper_title":"Diffusion Models Are Innate One-Step Generators","code":"https://github.com/Zyriix/GDD","n_code_links":1,"syntology":null},{"rank_in_archive_order":6,"model":"PaGoDA","metrics":{"FID":"1.21","Inception Score":"76.47","NFE":"1"},"uses_additional_data":false,"paper_date":"2024-05-23","paper":"/paper/pagoda-progressive-growing-of-a-one-step","paper_url":"https://arxiv.org/abs/2405.14822v2","paper_title":"PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher","code":"https://github.com/sony/pagoda","n_code_links":1,"syntology":{"n_ran":15,"n_unverified":2,"n_samples":17,"n_pointer_only_licence":0}},{"rank_in_archive_order":7,"model":"DisCo-Diff","metrics":{"FID":"1.22"},"uses_additional_data":false,"paper_date":"2024-07-03","paper":"/paper/disco-diff-enhancing-continuous-diffusion","paper_url":"https://arxiv.org/abs/2407.03300v1","paper_title":"DisCo-Diff: Enhancing Continuous Diffusion Models with Discrete Latents","code":"https://github.com/gcorso/disco-diffdock","n_code_links":1,"syntology":{"n_ran":6,"n_unverified":0,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"RIN","metrics":{"FID":"1.23"},"uses_additional_data":false,"paper_date":"2022-12-22","paper":"/paper/scalable-adaptive-computation-for-iterative","paper_url":"https://arxiv.org/abs/2212.11972v2","paper_title":"Scalable Adaptive Computation for Iterative Generation","code":"https://github.com/google-research/pix2seq","n_code_links":2,"syntology":{"n_ran":21,"n_unverified":8,"n_samples":29,"n_pointer_only_licence":1}},{"rank_in_archive_order":9,"model":"GDD","metrics":{"FID":"1.42","NFE":"1"},"uses_additional_data":false,"paper_date":"2024-05-31","paper":"/paper/diffusion-models-are-innate-one-step","paper_url":"https://arxiv.org/abs/2405.20750v2","paper_title":"Diffusion Models Are Innate One-Step Generators","code":"https://github.com/Zyriix/GDD","n_code_links":1,"syntology":null},{"rank_in_archive_order":10,"model":"SCT","metrics":{"FID":"1.47","NFE":"2"},"uses_additional_data":false,"paper_date":"2024-10-24","paper":"/paper/stable-consistency-tuning-understanding-and","paper_url":"https://arxiv.org/abs/2410.18958v3","paper_title":"Stable Consistency Tuning: Understanding and Improving Consistency Models","code":"https://github.com/G-U-N/Stable-Consistency-Tuning","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"CDM","metrics":{"FID":"1.48"},"uses_additional_data":false,"paper_date":"2021-05-30","paper":"/paper/cascaded-diffusion-models-for-high-fidelity","paper_url":"https://arxiv.org/abs/2106.15282v3","paper_title":"Cascaded Diffusion Models for High Fidelity Image Generation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":12,"model":"StyleGAN-XL","metrics":{"FID":"1.51","NFE":"1"},"uses_additional_data":false,"paper_date":"2022-02-01","paper":"/paper/stylegan-xl-scaling-stylegan-to-large-diverse","paper_url":"https://arxiv.org/abs/2202.00273v2","paper_title":"StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets","code":"https://github.com/autonomousvision/stylegan-xl","n_code_links":2,"syntology":{"n_ran":14,"n_unverified":5,"n_samples":19,"n_pointer_only_licence":0}},{"rank_in_archive_order":13,"model":"SiD","metrics":{"FID":"1.524","NFE":"1"},"uses_additional_data":false,"paper_date":"2024-04-05","paper":"/paper/score-identity-distillation-exponentially","paper_url":"https://arxiv.org/abs/2404.04057v3","paper_title":"Score identity Distillation: Exponentially Fast Distillation of Pretrained Diffusion Models for One-Step Generation","code":"https://github.com/mingyuanzhou/sid","n_code_links":2,"syntology":{"n_ran":9,"n_unverified":6,"n_samples":15,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"TCM","metrics":{"FID":"1.62","NFE":"2"},"uses_additional_data":false,"paper_date":"2024-10-18","paper":"/paper/truncated-consistency-models","paper_url":"https://arxiv.org/abs/2410.14895v2","paper_title":"Truncated Consistency Models","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":15,"model":"ECM-XL","metrics":{"FID":"1.67","NFE":"2"},"uses_additional_data":false,"paper_date":"2024-06-20","paper":"/paper/consistency-models-made-easy","paper_url":"https://arxiv.org/abs/2406.14548v2","paper_title":"Consistency Models Made Easy","code":"https://github.com/locuslab/ect","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":16,"model":"CAF","metrics":{"FID":"1.69","Inception Score":"62.03","NFE":"2"},"uses_additional_data":false,"paper_date":"2024-11-01","paper":"/paper/constant-acceleration-flow-1","paper_url":"https://arxiv.org/abs/2411.00322v1","paper_title":"Constant Acceleration Flow","code":"https://github.com/mlvlab/CAF","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":17,"model":"CTM","metrics":{"FID":"1.73","Inception Score":"64.29","NFE":"2"},"uses_additional_data":false,"paper_date":"2023-10-01","paper":"/paper/consistency-trajectory-models-learning","paper_url":"https://arxiv.org/abs/2310.02279v3","paper_title":"Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion","code":"https://github.com/sony/ctm","n_code_links":2,"syntology":{"n_ran":14,"n_unverified":4,"n_samples":18,"n_pointer_only_licence":0}},{"rank_in_archive_order":18,"model":"ADM (dropout)","metrics":{"FID":"2.07"},"uses_additional_data":false,"paper_date":"2021-05-11","paper":"/paper/diffusion-models-beat-gans-on-image-synthesis","paper_url":"https://arxiv.org/abs/2105.05233v4","paper_title":"Diffusion Models Beat GANs on Image Synthesis","code":"https://github.com/openai/guided-diffusion","n_code_links":21,"syntology":{"n_ran":27,"n_unverified":23,"n_samples":50,"n_pointer_only_licence":16}},{"rank_in_archive_order":19,"model":"LEGO","metrics":{"FID":"2.16","Inception Score":"78.7"},"uses_additional_data":false,"paper_date":"2023-10-10","paper":"/paper/learning-stackable-and-skippable-lego-bricks","paper_url":"https://arxiv.org/abs/2310.06389v3","paper_title":"Learning Stackable and Skippable LEGO Bricks for Efficient, Reconfigurable, and Variable-Resolution Diffusion Modeling","code":"https://github.com/JegZheng/LEGODiffusion","n_code_links":1,"syntology":{"n_ran":11,"n_unverified":6,"n_samples":17,"n_pointer_only_licence":0}},{"rank_in_archive_order":20,"model":"TarFlow","metrics":{"Bits per dim":"2.99","FID":"2.9"},"uses_additional_data":false,"paper_date":"2024-12-09","paper":"/paper/normalizing-flows-are-capable-generative","paper_url":"https://arxiv.org/abs/2412.06329v2","paper_title":"Normalizing Flows are Capable Generative Models","code":"https://github.com/apple/ml-tarflow","n_code_links":3,"syntology":{"n_ran":1,"n_unverified":3,"n_samples":4,"n_pointer_only_licence":1}},{"rank_in_archive_order":21,"model":"Improved DDPM","metrics":{"Bits per dim":"3.53","FID":"2.92"},"uses_additional_data":false,"paper_date":"2021-02-18","paper":"/paper/improved-denoising-diffusion-probabilistic-1","paper_url":"https://arxiv.org/abs/2102.09672v1","paper_title":"Improved Denoising Diffusion Probabilistic Models","code":"https://github.com/neonbjb/tortoise-tts","n_code_links":18,"syntology":{"n_ran":9,"n_unverified":2,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":22,"model":"2-rectified flow++ (NFE=2)","metrics":{"FID":"3.64"},"uses_additional_data":false,"paper_date":"2024-05-30","paper":"/paper/improving-the-training-of-rectified-flows","paper_url":"https://arxiv.org/abs/2405.20320v2","paper_title":"Improving the Training of Rectified Flows","code":"https://github.com/sangyun884/rfpp","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":23,"model":"2-rectified flow++ (NFE=1)","metrics":{"FID":"4.31"},"uses_additional_data":false,"paper_date":"2024-05-30","paper":"/paper/improving-the-training-of-rectified-flows","paper_url":"https://arxiv.org/abs/2405.20320v2","paper_title":"Improving the Training of Rectified Flows","code":"https://github.com/sangyun884/rfpp","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":24,"model":"CD (Diffusion + Distillation, NFE=2)","metrics":{"FID":"4.70","NFE":"2"},"uses_additional_data":false,"paper_date":"2023-03-02","paper":"/paper/consistency-models","paper_url":"https://arxiv.org/abs/2303.01469v2","paper_title":"Consistency Models","code":"https://github.com/openai/consistency_models","n_code_links":15,"syntology":{"n_ran":28,"n_unverified":29,"n_samples":57,"n_pointer_only_licence":8}},{"rank_in_archive_order":25,"model":"CD (Diffusion + Distillation, NFE=1)","metrics":{"FID":"6.20","NFE":"1"},"uses_additional_data":false,"paper_date":"2023-03-02","paper":"/paper/consistency-models","paper_url":"https://arxiv.org/abs/2303.01469v2","paper_title":"Consistency Models","code":"https://github.com/openai/consistency_models","n_code_links":15,"syntology":{"n_ran":28,"n_unverified":29,"n_samples":57,"n_pointer_only_licence":8}},{"rank_in_archive_order":26,"model":"CT (Direct Generation, NFE=2)","metrics":{"FID":"11.1","NFE":"2"},"uses_additional_data":false,"paper_date":"2023-03-02","paper":"/paper/consistency-models","paper_url":"https://arxiv.org/abs/2303.01469v2","paper_title":"Consistency Models","code":"https://github.com/openai/consistency_models","n_code_links":15,"syntology":{"n_ran":28,"n_unverified":29,"n_samples":57,"n_pointer_only_licence":8}},{"rank_in_archive_order":27,"model":"CT (Direct Generation, NFE=1)","metrics":{"FID":"13.0","NFE":"1"},"uses_additional_data":false,"paper_date":"2023-03-02","paper":"/paper/consistency-models","paper_url":"https://arxiv.org/abs/2303.01469v2","paper_title":"Consistency Models","code":"https://github.com/openai/consistency_models","n_code_links":15,"syntology":{"n_ran":28,"n_unverified":29,"n_samples":57,"n_pointer_only_licence":8}},{"rank_in_archive_order":28,"model":"FM","metrics":{"Bits per dim":"3.31","FID":"14.45"},"uses_additional_data":false,"paper_date":"2022-10-06","paper":"/paper/flow-matching-for-generative-modeling","paper_url":"https://arxiv.org/abs/2210.02747v2","paper_title":"Flow Matching for Generative Modeling","code":"https://github.com/shivammehta25/Matcha-TTS","n_code_links":5,"syntology":{"n_ran":1,"n_unverified":3,"n_samples":4,"n_pointer_only_licence":3}},{"rank_in_archive_order":29,"model":"CLR-GAN","metrics":{"FID":"20.27"},"uses_additional_data":false,"paper_date":"2024-09-30","paper":"/paper/clr-gan-improving-gans-stability-and-quality","paper_url":"https://link.springer.com/chapter/10.1007/978-3-031-73232-4_12","paper_title":"CLR-GAN: Improving GANs Stability and Quality via Consistent Latent Representation and Reconstruction","code":"https://github.com/Petecheco/CLR-GAN","n_code_links":1,"syntology":null},{"rank_in_archive_order":30,"model":"PGMGAN","metrics":{"FID":"21.73"},"uses_additional_data":false,"paper_date":"2021-04-02","paper":"/paper/partition-guided-gans","paper_url":"https://arxiv.org/abs/2104.00816v2","paper_title":"Partition-Guided GANs","code":"https://github.com/alisadeghian/PGMGAN","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":31,"model":"GLIDE + CLIP + CLS + CLS-FREE","metrics":{"FID":"29.184","Inception Score":"34.952","KID":"3.766"},"uses_additional_data":false,"paper_date":"2022-10-20","paper":"/paper/composing-ensembles-of-pre-trained-models-via","paper_url":"https://arxiv.org/abs/2210.11522v1","paper_title":"Composing Ensembles of Pre-trained Models via Iterative Consensus","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":32,"model":"GLIDE + CLS-FREE","metrics":{"FID":"29.219","Inception Score":"25.926","KID":"5.325"},"uses_additional_data":false,"paper_date":"2022-10-20","paper":"/paper/composing-ensembles-of-pre-trained-models-via","paper_url":"https://arxiv.org/abs/2210.11522v1","paper_title":"Composing Ensembles of Pre-trained Models via Iterative Consensus","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":33,"model":"GLIDE + CLIP","metrics":{"FID":"30.462","Inception Score":"25.017","KID":"6.174"},"uses_additional_data":false,"paper_date":"2022-10-20","paper":"/paper/composing-ensembles-of-pre-trained-models-via","paper_url":"https://arxiv.org/abs/2210.11522v1","paper_title":"Composing Ensembles of Pre-trained Models via Iterative Consensus","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":34,"model":"GLIDE + CLS","metrics":{"FID":"30.871","Inception Score":"22.077"},"uses_additional_data":false,"paper_date":"2022-10-20","paper":"/paper/composing-ensembles-of-pre-trained-models-via","paper_url":"https://arxiv.org/abs/2210.11522v1","paper_title":"Composing Ensembles of Pre-trained Models via Iterative Consensus","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":35,"model":"NFDM","metrics":{"Bits per dim":"3.2"},"uses_additional_data":false,"paper_date":"2024-04-19","paper":"/paper/neural-flow-diffusion-models-learnable","paper_url":"https://arxiv.org/abs/2404.12940v2","paper_title":"Neural Flow Diffusion Models: Learnable Forward Process for Improved Diffusion Modelling","code":"https://github.com/GrigoryBartosh/neural_diffusion","n_code_links":1,"syntology":null},{"rank_in_archive_order":36,"model":"BSI","metrics":{"Bits per dim":"3.22"},"uses_additional_data":false,"paper_date":"2025-02-11","paper":"/paper/generative-modeling-with-bayesian-sample","paper_url":"https://arxiv.org/abs/2502.07580v2","paper_title":"Generative Modeling with Bayesian Sample Inference","code":"https://github.com/martenlienen/bsi","n_code_links":1,"syntology":null},{"rank_in_archive_order":37,"model":"Efficient-VDVAE","metrics":{"Bits per dim":"3.30 (different downsampling)"},"uses_additional_data":false,"paper_date":"2022-03-25","paper":"/paper/efficient-vdvae-less-is-more","paper_url":"https://arxiv.org/abs/2203.13751v2","paper_title":"Efficient-VDVAE: Less is more","code":"https://github.com/Rayhane-mamah/Efficient-VDVAE","n_code_links":1,"syntology":null},{"rank_in_archive_order":38,"model":"DenseFlow-74-10","metrics":{"Bits per dim":"3.35 (different downsampling)"},"uses_additional_data":false,"paper_date":"2021-06-08","paper":"/paper/densely-connected-normalizing-flows","paper_url":"https://arxiv.org/abs/2106.04627v3","paper_title":"Densely connected normalizing flows","code":"https://github.com/matejgrcic/DenseFlow","n_code_links":4,"syntology":null},{"rank_in_archive_order":39,"model":"NDM","metrics":{"Bits per dim":"3.35"},"uses_additional_data":false,"paper_date":"2023-10-12","paper":"/paper/neural-diffusion-models","paper_url":"https://arxiv.org/abs/2310.08337v3","paper_title":"Neural Diffusion Models","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":40,"model":"VDM","metrics":{"Bits per dim":"3.40"},"uses_additional_data":false,"paper_date":"2021-07-01","paper":"/paper/variational-diffusion-models","paper_url":"https://arxiv.org/abs/2107.00630v6","paper_title":"Variational Diffusion Models","code":"https://github.com/google-research/vdm","n_code_links":5,"syntology":{"n_ran":29,"n_unverified":9,"n_samples":38,"n_pointer_only_licence":2}},{"rank_in_archive_order":41,"model":"Combiner-Axial","metrics":{"Bits per dim":"3.42"},"uses_additional_data":false,"paper_date":"2021-07-12","paper":"/paper/combiner-full-attention-transformer-with","paper_url":"https://arxiv.org/abs/2107.05768v2","paper_title":"Combiner: Full Attention Transformer with Sparse Computation Cost","code":"https://github.com/google-research/google-research","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":2,"n_samples":3,"n_pointer_only_licence":1}},{"rank_in_archive_order":42,"model":"Routing Transformer","metrics":{"Bits per dim":"3.43"},"uses_additional_data":false,"paper_date":"2020-03-12","paper":"/paper/efficient-content-based-sparse-attention-with-1","paper_url":"https://arxiv.org/abs/2003.05997v5","paper_title":"Efficient Content-Based Sparse Attention with Routing Transformers","code":"https://github.com/lucidrains/local-attention","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":2}},{"rank_in_archive_order":43,"model":"Sparse Transformer 59M (strided)","metrics":{"Bits per dim":"3.44"},"uses_additional_data":false,"paper_date":"2019-04-23","paper":"/paper/190410509","paper_url":"http://arxiv.org/abs/1904.10509v1","paper_title":"Generating Long Sequences with Sparse Transformers","code":"https://github.com/mistralai/mistral-src","n_code_links":7,"syntology":{"n_ran":5,"n_unverified":1,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":44,"model":"MRCNF","metrics":{"Bits per dim":"3.44"},"uses_additional_data":false,"paper_date":"2021-06-15","paper":"/paper/multi-resolution-continuous-normalizing-flows","paper_url":"https://arxiv.org/abs/2106.08462v5","paper_title":"Multi-Resolution Continuous Normalizing Flows","code":"https://github.com/voletiv/mrcnf","n_code_links":1,"syntology":null},{"rank_in_archive_order":45,"model":"Hourglass","metrics":{"Bits per dim":"3.44"},"uses_additional_data":false,"paper_date":"2021-10-26","paper":"/paper/hierarchical-transformers-are-more-efficient","paper_url":"https://arxiv.org/abs/2110.13711v2","paper_title":"Hierarchical Transformers Are More Efficient Language Models","code":"https://github.com/labmlai/annotated_deep_learning_paper_implementations","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":2,"n_samples":4,"n_pointer_only_licence":2}},{"rank_in_archive_order":46,"model":"Combiner-Mixture","metrics":{"Bits per dim":"3.504"},"uses_additional_data":false,"paper_date":"2021-07-12","paper":"/paper/combiner-full-attention-transformer-with","paper_url":"https://arxiv.org/abs/2107.05768v2","paper_title":"Combiner: Full Attention Transformer with Sparse Computation Cost","code":"https://github.com/google-research/google-research","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":2,"n_samples":3,"n_pointer_only_licence":1}},{"rank_in_archive_order":47,"model":"SPN","metrics":{"Bits per dim":"3.52"},"uses_additional_data":false,"paper_date":"2018-12-04","paper":"/paper/generating-high-fidelity-images-with-subscale","paper_url":"http://arxiv.org/abs/1812.01608v1","paper_title":"Generating High Fidelity Images with Subscale Pixel Networks and Multidimensional Upscaling","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":48,"model":"Very Deep VAE","metrics":{"Bits per dim":"3.52"},"uses_additional_data":false,"paper_date":"2020-11-20","paper":"/paper/very-deep-vaes-generalize-autoregressive-1","paper_url":"https://arxiv.org/abs/2011.10650v2","paper_title":"Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images","code":"https://github.com/openai/vdvae","n_code_links":8,"syntology":{"n_ran":8,"n_unverified":1,"n_samples":9,"n_pointer_only_licence":4}},{"rank_in_archive_order":49,"model":"PixelCNN","metrics":{"Bits per dim":"3.57"},"uses_additional_data":false,"paper_date":"2016-12-24","paper":"/paper/pixelcnn-models-with-auxiliary-variables-for","paper_url":"http://arxiv.org/abs/1612.08185v4","paper_title":"PixelCNN Models with Auxiliary Variables for Natural Image Modeling","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":50,"model":"Gated PixelCNN (van den Oord et al., [2016c])","metrics":{"Bits per dim":"3.57"},"uses_additional_data":false,"paper_date":"2016-06-16","paper":"/paper/conditional-image-generation-with-pixelcnn","paper_url":"http://arxiv.org/abs/1606.05328v2","paper_title":"Conditional Image Generation with PixelCNN Decoders","code":"https://github.com/openai/pixel-cnn","n_code_links":14,"syntology":{"n_ran":4,"n_unverified":10,"n_samples":14,"n_pointer_only_licence":3}},{"rank_in_archive_order":51,"model":"Performer (12 layers)","metrics":{"Bits per dim":"3.636"},"uses_additional_data":false,"paper_date":"2020-09-30","paper":"/paper/rethinking-attention-with-performers","paper_url":"https://arxiv.org/abs/2009.14794v4","paper_title":"Rethinking Attention with Performers","code":"https://github.com/tensorflow/models/tree/master/official/nlp/modeling","n_code_links":7,"syntology":{"n_ran":9,"n_unverified":7,"n_samples":16,"n_pointer_only_licence":6}},{"rank_in_archive_order":52,"model":"Flow++","metrics":{"Bits per dim":"3.69"},"uses_additional_data":false,"paper_date":"2019-02-01","paper":"/paper/flow-improving-flow-based-generative-models","paper_url":"https://arxiv.org/abs/1902.00275v2","paper_title":"Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Design","code":"https://github.com/aravind0706/flowpp","n_code_links":4,"syntology":{"n_ran":0,"n_unverified":9,"n_samples":9,"n_pointer_only_licence":0}},{"rank_in_archive_order":53,"model":"MaCow (Var)","metrics":{"Bits per dim":"3.69"},"uses_additional_data":false,"paper_date":"2019-02-12","paper":"/paper/macow-masked-convolutional-generative-flow","paper_url":"https://arxiv.org/abs/1902.04208v5","paper_title":"MaCow: Masked Convolutional Generative Flow","code":"https://github.com/XuezheMax/wolf","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":3,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":54,"model":"Parallel Multiscale","metrics":{"Bits per dim":"3.7"},"uses_additional_data":false,"paper_date":"2017-03-10","paper":"/paper/parallel-multiscale-autoregressive-density","paper_url":"http://arxiv.org/abs/1703.03664v1","paper_title":"Parallel Multiscale Autoregressive Density Estimation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":55,"model":"MALI","metrics":{"Bits per dim":"3.71"},"uses_additional_data":false,"paper_date":"2021-02-09","paper":"/paper/mali-a-memory-efficient-and-reverse-accurate-1","paper_url":"https://arxiv.org/abs/2102.04668v2","paper_title":"MALI: A memory efficient and reverse accurate integrator for Neural ODEs","code":"https://github.com/juntang-zhuang/TorchDiffEqPack","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":2,"n_samples":4,"n_pointer_only_licence":0}},{"rank_in_archive_order":56,"model":"Reformer (12 layers)","metrics":{"Bits per dim":"3.710"},"uses_additional_data":false,"paper_date":"2020-01-13","paper":"/paper/reformer-the-efficient-transformer-1","paper_url":"https://arxiv.org/abs/2001.04451v2","paper_title":"Reformer: The Efficient Transformer","code":"https://github.com/huggingface/transformers","n_code_links":10,"syntology":{"n_ran":6,"n_unverified":2,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":57,"model":"Performer (6 layers)","metrics":{"Bits per dim":"3.719"},"uses_additional_data":false,"paper_date":"2020-09-30","paper":"/paper/rethinking-attention-with-performers","paper_url":"https://arxiv.org/abs/2009.14794v4","paper_title":"Rethinking Attention with Performers","code":"https://github.com/tensorflow/models/tree/master/official/nlp/modeling","n_code_links":7,"syntology":{"n_ran":9,"n_unverified":7,"n_samples":16,"n_pointer_only_licence":6}},{"rank_in_archive_order":58,"model":"Reformer (6 layers)","metrics":{"Bits per dim":"3.740"},"uses_additional_data":false,"paper_date":"2020-01-13","paper":"/paper/reformer-the-efficient-transformer-1","paper_url":"https://arxiv.org/abs/2001.04451v2","paper_title":"Reformer: The Efficient Transformer","code":"https://github.com/huggingface/transformers","n_code_links":10,"syntology":{"n_ran":6,"n_unverified":2,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":59,"model":"MaCow (Unf)","metrics":{"Bits per dim":"3.75"},"uses_additional_data":false,"paper_date":"2019-02-12","paper":"/paper/macow-masked-convolutional-generative-flow","paper_url":"https://arxiv.org/abs/1902.04208v5","paper_title":"MaCow: Masked Convolutional Generative Flow","code":"https://github.com/XuezheMax/wolf","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":3,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":60,"model":"Residual Flow","metrics":{"Bits per dim":"3.757"},"uses_additional_data":false,"paper_date":"2019-06-06","paper":"/paper/residual-flows-for-invertible-generative","paper_url":"https://arxiv.org/abs/1906.02735v6","paper_title":"Residual Flows for Invertible Generative Modeling","code":"https://github.com/rtqichen/residual-flows","n_code_links":4,"syntology":{"n_ran":5,"n_unverified":3,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":61,"model":"Glow (Kingma and Dhariwal, 2018)","metrics":{"Bits per dim":"3.81"},"uses_additional_data":false,"paper_date":"2018-07-09","paper":"/paper/glow-generative-flow-with-invertible-1x1","paper_url":"http://arxiv.org/abs/1807.03039v2","paper_title":"Glow: Generative Flow with Invertible 1x1 Convolutions","code":"https://github.com/openai/glow","n_code_links":27,"syntology":{"n_ran":75,"n_unverified":54,"n_samples":129,"n_pointer_only_licence":44}},{"rank_in_archive_order":62,"model":"Axial Transformer (6 layers)","metrics":{"Bits per dim":"4.032"},"uses_additional_data":false,"paper_date":"2019-12-20","paper":"/paper/axial-attention-in-multidimensional-1","paper_url":"https://arxiv.org/abs/1912.12180v1","paper_title":"Axial Attention in Multidimensional Transformers","code":"https://github.com/lucidrains/axial-attention","n_code_links":3,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":1}},{"rank_in_archive_order":63,"model":"Logsparse (6 layers)","metrics":{"Bits per dim":"4.351"},"uses_additional_data":false,"paper_date":"2019-06-29","paper":"/paper/enhancing-the-locality-and-breaking-the","paper_url":"https://arxiv.org/abs/1907.00235v3","paper_title":"Enhancing the Locality and Breaking the Memory Bottleneck of Transformer on Time Series Forecasting","code":"https://github.com/AIStream-Peelout/flow-forecast","n_code_links":2,"syntology":null},{"rank_in_archive_order":64,"model":"CTM (NFE 1)","metrics":{"Inception Score":"70.38","NFE":"1"},"uses_additional_data":false,"paper_date":"2023-10-01","paper":"/paper/consistency-trajectory-models-learning","paper_url":"https://arxiv.org/abs/2310.02279v3","paper_title":"Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion","code":"https://github.com/sony/ctm","n_code_links":2,"syntology":{"n_ran":14,"n_unverified":4,"n_samples":18,"n_pointer_only_licence":0}},{"rank_in_archive_order":65,"model":"GLIDE +CLS","metrics":{"KID":"7.952"},"uses_additional_data":false,"paper_date":"2022-10-20","paper":"/paper/composing-ensembles-of-pre-trained-models-via","paper_url":"https://arxiv.org/abs/2210.11522v1","paper_title":"Composing Ensembles of Pre-trained Models via Iterative Consensus","code":null,"n_code_links":0,"syntology":null}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,795 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9581,"papers_checked":6795,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2785},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":41,"rows_with_any_sample_ran":38,"distinct_papers_with_graph_line":32,"distinct_papers_with_any_sample_ran":30,"samples_over_distinct_papers":{"n_ran":318,"n_unverified":199,"n_samples":517,"n_pointer_only_licence":102,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":433,"n_unverified":304,"n_samples":737,"n_pointer_only_licence":134,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}