{"url":"/sota/knowledge-distillation-on-cifar-100","task":{"name":"Knowledge Distillation","url":"/task/knowledge-distillation","note":null},"dataset":{"name":"CIFAR-100","url":"/dataset/cifar-100"},"category":"Computer Vision","categories":["Computer Vision","Natural Language Processing"],"category_note":null,"description":"Knowledge distillation is the process of transferring knowledge from a large model to a smaller one. While large models (such as very deep neural networks or ensembles of many models) have higher knowledge capacity than small models, this capacity might not be fully utilized.","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":["Top-1 Accuracy (%)"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Top-1 Accuracy (%)":"higher"}},"counts":{"rows":27,"rows_with_code":24,"rows_with_paper_page":27,"rows_dated":27,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"SRD (T:resnet-32x4, S:shufflenet-v2)","metrics":{"Top-1 Accuracy (%)":"79.86"},"uses_additional_data":false,"paper_date":"2023-03-20","paper":"/paper/a-closer-look-at-the-training-dynamics-of","paper_url":"https://arxiv.org/abs/2303.11098v5","paper_title":"Understanding the Role of the Projector in Knowledge Distillation","code":"https://github.com/yoshitomo-matsubara/torchdistill","n_code_links":4,"syntology":null},{"rank_in_archive_order":2,"model":"shufflenet-v2(T:resnet-32x4, S:shufflenet-v2)","metrics":{"Top-1 Accuracy (%)":"78.76"},"uses_additional_data":false,"paper_date":"2024-03-03","paper":"/paper/logit-standardization-in-knowledge","paper_url":"https://arxiv.org/abs/2403.01427v1","paper_title":"Logit Standardization in Knowledge Distillation","code":"https://github.com/sunshangquan/logit-standardardization-kd","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":3,"model":"MV-MR (T: CLIP/ViT-B-16 S: resnet50)","metrics":{"Top-1 Accuracy (%)":"78.6"},"uses_additional_data":false,"paper_date":"2023-03-21","paper":"/paper/mv-mr-multi-views-and-multi-representations","paper_url":"https://arxiv.org/abs/2303.12130v2","paper_title":"MV-MR: multi-views and multi-representations for self-supervised learning and knowledge distillation","code":"https://github.com/vkinakh/mv-mr","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"resnet8x4\n(T: resnet32x4 S: resnet8x4)","metrics":{"Top-1 Accuracy (%)":"78.28"},"uses_additional_data":false,"paper_date":"2024-03-03","paper":"/paper/logit-standardization-in-knowledge","paper_url":"https://arxiv.org/abs/2403.01427v1","paper_title":"Logit Standardization in Knowledge Distillation","code":"https://github.com/sunshangquan/logit-standardardization-kd","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":5,"model":"resnet8x4 (T: resnet32x4 S: resnet8x4 [modified])","metrics":{"Top-1 Accuracy (%)":"78.08"},"uses_additional_data":false,"paper_date":"2022-03-26","paper":"/paper/knowledge-distillation-with-the-reused","paper_url":"https://arxiv.org/abs/2203.14001v1","paper_title":"Knowledge Distillation with the Reused Teacher Classifier","code":"https://github.com/DefangChen/SimKD","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":6,"model":"ReviewKD++(T:resnet-32x4, S:shufflenet-v2)","metrics":{"Top-1 Accuracy (%)":"77.93"},"uses_additional_data":false,"paper_date":"2023-05-26","paper":"/paper/improving-knowledge-distillation-via-1","paper_url":"https://arxiv.org/abs/2305.17007v1","paper_title":"Improving Knowledge Distillation via Regularizing Feature Norm and Direction","code":"https://github.com/wangyz1608/knowledge-distillation-via-nd","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":7,"model":"ReviewKD++(T:resnet-32x4, S:shufflenet-v1)","metrics":{"Top-1 Accuracy (%)":"77.68"},"uses_additional_data":false,"paper_date":"2023-05-26","paper":"/paper/improving-knowledge-distillation-via-1","paper_url":"https://arxiv.org/abs/2305.17007v1","paper_title":"Improving Knowledge Distillation via Regularizing Feature Norm and Direction","code":"https://github.com/wangyz1608/knowledge-distillation-via-nd","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":8,"model":"resnet8x4 (T: resnet32x4 S: resnet8x4)","metrics":{"Top-1 Accuracy (%)":"77.50"},"uses_additional_data":false,"paper_date":"2023-10-05","paper":"/paper/luminet-the-bright-side-of-perceptual","paper_url":"https://arxiv.org/abs/2310.03669v2","paper_title":"LumiNet: The Bright Side of Perceptual Knowledge Distillation","code":"https://github.com/ismail31416/luminet","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"resnet8x4 (T: resnet32x4 S: resnet8x4)","metrics":{"Top-1 Accuracy (%)":"76.68"},"uses_additional_data":false,"paper_date":"2021-12-01","paper":"/paper/information-theoretic-representation","paper_url":"https://arxiv.org/abs/2112.00459v3","paper_title":"Information Theoretic Representation Distillation","code":"https://github.com/roymiles/ITRD","n_code_links":1,"syntology":null},{"rank_in_archive_order":10,"model":"resnet8x4 (T: resnet32x4 S: resnet8x4)","metrics":{"Top-1 Accuracy (%)":"76.31"},"uses_additional_data":false,"paper_date":"2022-05-21","paper":"/paper/knowledge-distillation-from-a-stronger","paper_url":"https://arxiv.org/abs/2205.10536v3","paper_title":"Knowledge Distillation from A Stronger Teacher","code":"https://github.com/yoshitomo-matsubara/torchdistill","n_code_links":3,"syntology":{"n_ran":10,"n_unverified":1,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"DKD++(T:resnet-32x4, S:resnet-8x4)","metrics":{"Top-1 Accuracy (%)":"76.28"},"uses_additional_data":false,"paper_date":"2023-05-26","paper":"/paper/improving-knowledge-distillation-via-1","paper_url":"https://arxiv.org/abs/2305.17007v1","paper_title":"Improving Knowledge Distillation via Regularizing Feature Norm and Direction","code":"https://github.com/wangyz1608/knowledge-distillation-via-nd","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":12,"model":"resnet8x4 (T: resnet32x4 S: resnet8x4)","metrics":{"Top-1 Accuracy (%)":"76.15"},"uses_additional_data":false,"paper_date":"2020-12-15","paper":"/paper/wasserstein-contrastive-representation","paper_url":"https://arxiv.org/abs/2012.08674v2","paper_title":"Wasserstein Contrastive Representation Distillation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":13,"model":"ReviewKD++(T:WRN-40-2, S:WRN-40-1)","metrics":{"Top-1 Accuracy (%)":"75.66"},"uses_additional_data":false,"paper_date":"2023-05-26","paper":"/paper/improving-knowledge-distillation-via-1","paper_url":"https://arxiv.org/abs/2305.17007v1","paper_title":"Improving Knowledge Distillation via Regularizing Feature Norm and Direction","code":"https://github.com/wangyz1608/knowledge-distillation-via-nd","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":14,"model":"resnet8x4 (T: resnet32x4 S: resnet8x4)","metrics":{"Top-1 Accuracy (%)":"75.63"},"uses_additional_data":false,"paper_date":"2021-04-19","paper":"/paper/distilling-knowledge-via-knowledge-review","paper_url":"https://arxiv.org/abs/2104.09044v1","paper_title":"Distilling Knowledge via Knowledge Review","code":"https://github.com/yoshitomo-matsubara/torchdistill","n_code_links":7,"syntology":{"n_ran":6,"n_unverified":2,"n_samples":8,"n_pointer_only_licence":5}},{"rank_in_archive_order":15,"model":"resnet8x4 (T: resnet32x4 S: resnet8x4)","metrics":{"Top-1 Accuracy (%)":"75.51"},"uses_additional_data":false,"paper_date":"2019-10-23","paper":"/paper/contrastive-representation-distillation-1","paper_url":"https://arxiv.org/abs/1910.10699v3","paper_title":"Contrastive Representation Distillation","code":"https://github.com/HobbitLong/RepDistiller","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":16,"model":"vgg8 (T:vgg13 S:vgg8)","metrics":{"Top-1 Accuracy (%)":"74.93"},"uses_additional_data":false,"paper_date":"2021-12-01","paper":"/paper/information-theoretic-representation","paper_url":"https://arxiv.org/abs/2112.00459v3","paper_title":"Information Theoretic Representation Distillation","code":"https://github.com/roymiles/ITRD","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"vgg8 (T:vgg13 S:vgg8)","metrics":{"Top-1 Accuracy (%)":"74.84"},"uses_additional_data":false,"paper_date":"2021-04-19","paper":"/paper/distilling-knowledge-via-knowledge-review","paper_url":"https://arxiv.org/abs/2104.09044v1","paper_title":"Distilling Knowledge via Knowledge Review","code":"https://github.com/yoshitomo-matsubara/torchdistill","n_code_links":7,"syntology":{"n_ran":6,"n_unverified":2,"n_samples":8,"n_pointer_only_licence":5}},{"rank_in_archive_order":18,"model":"vgg8 (T:vgg13 S:vgg8)","metrics":{"Top-1 Accuracy (%)":"74.72"},"uses_additional_data":false,"paper_date":"2020-12-15","paper":"/paper/wasserstein-contrastive-representation","paper_url":"https://arxiv.org/abs/2012.08674v2","paper_title":"Wasserstein Contrastive Representation Distillation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":19,"model":"vgg8 (T:vgg13 S:vgg8)","metrics":{"Top-1 Accuracy (%)":"74.29"},"uses_additional_data":false,"paper_date":"2019-10-23","paper":"/paper/contrastive-representation-distillation-1","paper_url":"https://arxiv.org/abs/1910.10699v3","paper_title":"Contrastive Representation Distillation","code":"https://github.com/HobbitLong/RepDistiller","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":20,"model":"resnet8x4 (T: resnet32x4 S: resnet8x4)","metrics":{"Top-1 Accuracy (%)":"73.33"},"uses_additional_data":false,"paper_date":"2015-03-09","paper":"/paper/distilling-the-knowledge-in-a-neural-network","paper_url":"http://arxiv.org/abs/1503.02531v1","paper_title":"Distilling the Knowledge in a Neural Network","code":"https://github.com/labmlai/annotated_deep_learning_paper_implementations","n_code_links":64,"syntology":{"n_ran":16,"n_unverified":21,"n_samples":37,"n_pointer_only_licence":10}},{"rank_in_archive_order":21,"model":"vgg8 (T:vgg13 S:vgg8)","metrics":{"Top-1 Accuracy (%)":"72.98"},"uses_additional_data":false,"paper_date":"2015-03-09","paper":"/paper/distilling-the-knowledge-in-a-neural-network","paper_url":"http://arxiv.org/abs/1503.02531v1","paper_title":"Distilling the Knowledge in a Neural Network","code":"https://github.com/labmlai/annotated_deep_learning_paper_implementations","n_code_links":64,"syntology":{"n_ran":16,"n_unverified":21,"n_samples":37,"n_pointer_only_licence":10}},{"rank_in_archive_order":22,"model":"KD++(T:resnet56, S:resnet20)","metrics":{"Top-1 Accuracy (%)":"72.53"},"uses_additional_data":false,"paper_date":"2023-05-26","paper":"/paper/improving-knowledge-distillation-via-1","paper_url":"https://arxiv.org/abs/2305.17007v1","paper_title":"Improving Knowledge Distillation via Regularizing Feature Norm and Direction","code":"https://github.com/wangyz1608/knowledge-distillation-via-nd","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":23,"model":"resnet110 (T:resnet110 S:resnet20)","metrics":{"Top-1 Accuracy (%)":"71.99"},"uses_additional_data":false,"paper_date":"2021-12-01","paper":"/paper/information-theoretic-representation","paper_url":"https://arxiv.org/abs/2112.00459v3","paper_title":"Information Theoretic Representation Distillation","code":"https://github.com/roymiles/ITRD","n_code_links":1,"syntology":null},{"rank_in_archive_order":24,"model":"resnet110 (T:resnet110 S:resnet20)","metrics":{"Top-1 Accuracy (%)":"71.88"},"uses_additional_data":false,"paper_date":"2020-12-15","paper":"/paper/wasserstein-contrastive-representation","paper_url":"https://arxiv.org/abs/2012.08674v2","paper_title":"Wasserstein Contrastive Representation Distillation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":25,"model":"resnet110 (T:resnet110 S:resnet20)","metrics":{"Top-1 Accuracy (%)":"71.56"},"uses_additional_data":false,"paper_date":"2019-10-23","paper":"/paper/contrastive-representation-distillation-1","paper_url":"https://arxiv.org/abs/1910.10699v3","paper_title":"Contrastive Representation Distillation","code":"https://github.com/HobbitLong/RepDistiller","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":26,"model":"DKD++(T:resnet50, S:mobilenetv2)","metrics":{"Top-1 Accuracy (%)":"70.82"},"uses_additional_data":false,"paper_date":"2023-05-26","paper":"/paper/improving-knowledge-distillation-via-1","paper_url":"https://arxiv.org/abs/2305.17007v1","paper_title":"Improving Knowledge Distillation via Regularizing Feature Norm and Direction","code":"https://github.com/wangyz1608/knowledge-distillation-via-nd","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":27,"model":"resnet110 (T:resnet110 S:resnet20)","metrics":{"Top-1 Accuracy (%)":"70.67"},"uses_additional_data":false,"paper_date":"2015-03-09","paper":"/paper/distilling-the-knowledge-in-a-neural-network","paper_url":"http://arxiv.org/abs/1503.02531v1","paper_title":"Distilling the Knowledge in a Neural Network","code":"https://github.com/labmlai/annotated_deep_learning_paper_implementations","n_code_links":64,"syntology":{"n_ran":16,"n_unverified":21,"n_samples":37,"n_pointer_only_licence":10}}],"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":18,"rows_with_any_sample_ran":18,"distinct_papers_with_graph_line":7,"distinct_papers_with_any_sample_ran":7,"samples_over_distinct_papers":{"n_ran":39,"n_unverified":24,"n_samples":63,"n_pointer_only_licence":22,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":96,"n_unverified":68,"n_samples":164,"n_pointer_only_licence":66,"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"}}}