{"url":"/sota/image-reconstruction-on-imagenet","task":{"name":"Image Reconstruction","url":"/task/image-reconstruction","note":null},"dataset":{"name":"ImageNet","url":"/dataset/imagenet"},"category":"Computer Vision","categories":["Computer Vision","Medical","Miscellaneous"],"category_note":null,"description":null,"description_from":null,"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","LPIPS","PSNR","SSIM"],"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","LPIPS":null,"PSNR":"higher","SSIM":"higher"}},"counts":{"rows":15,"rows_with_code":15,"rows_with_paper_page":15,"rows_dated":15,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"MGVQ (16x16x8)","metrics":{"FID":"0.49","LPIPS":"0.086","PSNR":"24.70","SSIM":"0.787"},"uses_additional_data":false,"paper_date":"2025-07-14","paper":"/paper/mgvq-could-vq-vae-beat-vae-a-generalizable","paper_url":"https://arxiv.org/abs/2507.07997","paper_title":"MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization","code":"https://github.com/MKJia/MGVQ","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"MGVQ (16x16x4)","metrics":{"FID":"0.64","LPIPS":"0.110","PSNR":"23.71","SSIM":"0.755"},"uses_additional_data":false,"paper_date":"2025-07-14","paper":"/paper/mgvq-could-vq-vae-beat-vae-a-generalizable","paper_url":"https://arxiv.org/abs/2507.07997","paper_title":"MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization","code":"https://github.com/MKJia/MGVQ","n_code_links":1,"syntology":null},{"rank_in_archive_order":3,"model":"GigaTok-XL-XXL","metrics":{"FID":"0.79","LPIPS":"0.1947","PSNR":"21.65","SSIM":"0.699"},"uses_additional_data":false,"paper_date":"2025-04-11","paper":"/paper/gigatok-scaling-visual-tokenizers-to-3","paper_url":"https://arxiv.org/abs/2504.08736v1","paper_title":"GigaTok: Scaling Visual Tokenizers to 3 Billion Parameters for Autoregressive Image Generation","code":"https://github.com/SilentView/GigaTok","n_code_links":1,"syntology":{"n_ran":10,"n_unverified":5,"n_samples":15,"n_pointer_only_licence":15}},{"rank_in_archive_order":4,"model":"OptVQ (16x16x8)","metrics":{"FID":"0.91","LPIPS":"0.066","PSNR":"27.57","SSIM":"0.729"},"uses_additional_data":false,"paper_date":"2024-12-19","paper":"/paper/preventing-local-pitfalls-in-vector","paper_url":"https://arxiv.org/abs/2412.15195v1","paper_title":"Preventing Local Pitfalls in Vector Quantization via Optimal Transport","code":"https://github.com/zbr17/OptVQ","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"OptVQ (16x16x4)","metrics":{"FID":"1.00","LPIPS":"0.076","PSNR":"26.59","SSIM":"0.717"},"uses_additional_data":false,"paper_date":"2024-12-19","paper":"/paper/preventing-local-pitfalls-in-vector","paper_url":"https://arxiv.org/abs/2412.15195v1","paper_title":"Preventing Local Pitfalls in Vector Quantization via Optimal Transport","code":"https://github.com/zbr17/OptVQ","n_code_links":1,"syntology":null},{"rank_in_archive_order":6,"model":"IBQ (16x16)","metrics":{"FID":"1.00","LPIPS":"0.2030"},"uses_additional_data":false,"paper_date":"2024-12-03","paper":"/paper/taming-scalable-visual-tokenizer-for","paper_url":"https://arxiv.org/abs/2412.02692v1","paper_title":"Taming Scalable Visual Tokenizer for Autoregressive Image Generation","code":"https://github.com/tencentarc/seed-voken","n_code_links":2,"syntology":{"n_ran":13,"n_unverified":0,"n_samples":13,"n_pointer_only_licence":3}},{"rank_in_archive_order":7,"model":"Mo-VQGAN (16x16x4)","metrics":{"FID":"1.12","LPIPS":"0.113","PSNR":"22.42","SSIM":"0.673"},"uses_additional_data":false,"paper_date":"2022-09-19","paper":"/paper/movq-modulating-quantized-vectors-for-high","paper_url":"https://arxiv.org/abs/2209.09002v1","paper_title":"MoVQ: Modulating Quantized Vectors for High-Fidelity Image Generation","code":"https://github.com/ai-forever/Kandinsky-2","n_code_links":2,"syntology":null},{"rank_in_archive_order":8,"model":"Open-Magvit2 (16x16)","metrics":{"FID":"1.17","PSNR":"21.90"},"uses_additional_data":false,"paper_date":"2024-09-06","paper":"/paper/open-magvit2-an-open-source-project-toward","paper_url":"https://arxiv.org/abs/2409.04410v2","paper_title":"Open-MAGVIT2: An Open-Source Project Toward Democratizing Auto-regressive Visual Generation","code":"https://github.com/tencentarc/open-magvit2","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":1}},{"rank_in_archive_order":9,"model":"ViT-VQGAN (16x16)","metrics":{"FID":"1.28"},"uses_additional_data":false,"paper_date":"2021-10-09","paper":"/paper/vector-quantized-image-modeling-with-improved-1","paper_url":"https://arxiv.org/abs/2110.04627v3","paper_title":"Vector-quantized Image Modeling with Improved VQGAN","code":"https://github.com/lucidrains/DALLE2-pytorch","n_code_links":5,"syntology":{"n_ran":6,"n_unverified":3,"n_samples":9,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"MaskBit (16x16)","metrics":{"FID":"1.66"},"uses_additional_data":false,"paper_date":"2024-09-24","paper":"/paper/maskbit-embedding-free-image-generation-via","paper_url":"https://arxiv.org/abs/2409.16211v2","paper_title":"MaskBit: Embedding-free Image Generation via Bit Tokens","code":"https://github.com/markweberdev/maskbit","n_code_links":1,"syntology":{"n_ran":8,"n_unverified":1,"n_samples":9,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"TiTok-S-128","metrics":{"FID":"1.71"},"uses_additional_data":false,"paper_date":"2024-06-11","paper":"/paper/an-image-is-worth-32-tokens-for","paper_url":"https://arxiv.org/abs/2406.07550v1","paper_title":"An Image is Worth 32 Tokens for Reconstruction and Generation","code":"https://github.com/bytedance/1d-tokenizer","n_code_links":2,"syntology":null},{"rank_in_archive_order":12,"model":"RQ-VAE (8x8x16)","metrics":{"FID":"1.83"},"uses_additional_data":false,"paper_date":"2022-03-03","paper":"/paper/autoregressive-image-generation-using","paper_url":"https://arxiv.org/abs/2203.01941v2","paper_title":"Autoregressive Image Generation using Residual Quantization","code":"https://github.com/kakaobrain/rq-vae-transformer","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":13,"model":"MaskGIT-VQGAN (16x16)","metrics":{"FID":"2.28"},"uses_additional_data":false,"paper_date":"2022-02-08","paper":"/paper/maskgit-masked-generative-image-transformer","paper_url":"https://arxiv.org/abs/2202.04200v1","paper_title":"MaskGIT: Masked Generative Image Transformer","code":"https://github.com/lucidrains/soundstorm-pytorch","n_code_links":9,"syntology":{"n_ran":14,"n_unverified":7,"n_samples":21,"n_pointer_only_licence":4}},{"rank_in_archive_order":14,"model":"VQGAN-LC (16x16)","metrics":{"FID":"2.62","LPIPS":"0.120","PSNR":"23.80","SSIM":"0.589"},"uses_additional_data":false,"paper_date":"2024-06-17","paper":"/paper/scaling-the-codebook-size-of-vqgan-to-100000","paper_url":"https://arxiv.org/abs/2406.11837v1","paper_title":"Scaling the Codebook Size of VQGAN to 100,000 with a Utilization Rate of 99%","code":"https://github.com/zh460045050/vqgan-lc","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":15,"model":"Taming-VQGAN (16x16)","metrics":{"FID":"3.64","LPIPS":"0.177","PSNR":"19.93","SSIM":"0.542"},"uses_additional_data":false,"paper_date":"2020-12-17","paper":"/paper/taming-transformers-for-high-resolution-image","paper_url":"https://arxiv.org/abs/2012.09841v3","paper_title":"Taming Transformers for High-Resolution Image Synthesis","code":"https://github.com/CompVis/taming-transformers","n_code_links":13,"syntology":{"n_ran":6,"n_unverified":0,"n_samples":6,"n_pointer_only_licence":4}}],"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 6,885 of the 9,623 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":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-25T09:33:49+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":9,"rows_with_any_sample_ran":9,"distinct_papers_with_graph_line":9,"distinct_papers_with_any_sample_ran":9,"samples_over_distinct_papers":{"n_ran":63,"n_unverified":16,"n_samples":79,"n_pointer_only_licence":30,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":63,"n_unverified":16,"n_samples":79,"n_pointer_only_licence":30,"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"}}}