{"url":"/sota/machine-translation-on-wmt2014-english-german","task":{"name":"Machine Translation","url":"/task/machine-translation","note":null},"dataset":{"name":"WMT2014 English-German","url":"/dataset/wmt-2014"},"category":"Natural Language Processing","categories":["Natural Language Processing"],"category_note":null,"description":"**Machine translation** is the task of translating a sentence in a source language to a different target language.\r\n\r\nApproaches for machine translation can range from rule-based to statistical to neural-based. More recently, encoder-decoder attention-based architectures like BERT have attained major improvements in machine translation. \r\n\r\nOne of the most popular datasets used to benchmark machine translation systems is the WMT family of datasets. Some of the most commonly used evaluation metrics for machine translation systems include BLEU, METEOR, NIST, and others. \r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [Google seq2seq](https://github.com/google/seq2seq) )</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":["BLEU score","SacreBLEU","Number of Params","Hardware Burden","Operations per network pass"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"BLEU score":"higher","SacreBLEU":null,"Number of Params":"lower","Hardware Burden":null,"Operations per network pass":null}},"counts":{"rows":91,"rows_with_code":81,"rows_with_paper_page":90,"rows_dated":89,"rows_using_additional_data":1},"rows":[{"rank_in_archive_order":1,"model":"Transformer Cycle (Rev)","metrics":{"BLEU score":"35.14","SacreBLEU":"33.54"},"uses_additional_data":false,"paper_date":"2021-04-13","paper":"/paper/lessons-on-parameter-sharing-across-layers-in","paper_url":"https://arxiv.org/abs/2104.06022v4","paper_title":"Lessons on Parameter Sharing across Layers in Transformers","code":"https://github.com/takase/share_layer_params","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":1}},{"rank_in_archive_order":2,"model":"Noisy back-translation","metrics":{"BLEU score":"35.0","Hardware Burden":"146G","SacreBLEU":"33.8"},"uses_additional_data":true,"paper_date":"2018-08-28","paper":"/paper/understanding-back-translation-at-scale","paper_url":"http://arxiv.org/abs/1808.09381v2","paper_title":"Understanding Back-Translation at Scale","code":"https://github.com/facebookresearch/fairseq","n_code_links":3,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":3,"model":"Transformer+Rep(Uni)","metrics":{"BLEU score":"33.89","SacreBLEU":"32.35"},"uses_additional_data":false,"paper_date":"2021-04-05","paper":"/paper/rethinking-perturbations-in-encoder-decoders","paper_url":"https://arxiv.org/abs/2104.01853v1","paper_title":"Rethinking Perturbations in Encoder-Decoders for Fast Training","code":"https://github.com/takase/rethink_perturbations","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":4,"model":"T5-11B","metrics":{"BLEU score":"32.1","Number of Params":"11110M"},"uses_additional_data":false,"paper_date":"2019-10-23","paper":"/paper/exploring-the-limits-of-transfer-learning","paper_url":"https://arxiv.org/abs/1910.10683v4","paper_title":"Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":21,"n_unverified":10,"n_samples":31,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"BiBERT","metrics":{"BLEU score":"31.26"},"uses_additional_data":false,"paper_date":"2021-09-09","paper":"/paper/bert-mbert-or-bibert-a-study-on","paper_url":"https://arxiv.org/abs/2109.04588v1","paper_title":"BERT, mBERT, or BiBERT? A Study on Contextualized Embeddings for Neural Machine Translation","code":"https://github.com/fe1ixxu/BiBERT","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":1}},{"rank_in_archive_order":6,"model":"Transformer + R-Drop","metrics":{"BLEU score":"30.91","Hardware Burden":"49G"},"uses_additional_data":false,"paper_date":"2021-06-28","paper":"/paper/r-drop-regularized-dropout-for-neural","paper_url":"https://arxiv.org/abs/2106.14448v2","paper_title":"R-Drop: Regularized Dropout for Neural Networks","code":"https://github.com/dropreg/R-Drop","n_code_links":8,"syntology":{"n_ran":4,"n_unverified":2,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":7,"model":"Bi-SimCut","metrics":{"BLEU score":"30.78"},"uses_additional_data":false,"paper_date":"2022-06-06","paper":"/paper/bi-simcut-a-simple-strategy-for-boosting-1","paper_url":"https://arxiv.org/abs/2206.02368v2","paper_title":"Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation","code":"https://github.com/gpengzhi/Bi-SimCut","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"BERT-fused NMT","metrics":{"BLEU score":"30.75"},"uses_additional_data":false,"paper_date":"2020-02-17","paper":"/paper/incorporating-bert-into-neural-machine-1","paper_url":"https://arxiv.org/abs/2002.06823v1","paper_title":"Incorporating BERT into Neural Machine Translation","code":"https://github.com/bert-nmt/bert-nmt","n_code_links":3,"syntology":null},{"rank_in_archive_order":9,"model":"Data Diversification - Transformer","metrics":{"BLEU score":"30.7"},"uses_additional_data":false,"paper_date":"2019-11-05","paper":"/paper/data-diversification-an-elegant-strategy-for","paper_url":"https://arxiv.org/abs/1911.01986v4","paper_title":"Data Diversification: A Simple Strategy For Neural Machine Translation","code":"https://github.com/nxphi47/data_diversification","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":1,"n_samples":4,"n_pointer_only_licence":4}},{"rank_in_archive_order":10,"model":"SimCut","metrics":{"BLEU score":"30.56"},"uses_additional_data":false,"paper_date":"2022-06-06","paper":"/paper/bi-simcut-a-simple-strategy-for-boosting-1","paper_url":"https://arxiv.org/abs/2206.02368v2","paper_title":"Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation","code":"https://github.com/gpengzhi/Bi-SimCut","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"Mask Attention Network (big)","metrics":{"BLEU score":"30.4","Number of Params":"215M"},"uses_additional_data":false,"paper_date":"2021-03-25","paper":"/paper/mask-attention-networks-rethinking-and","paper_url":"https://arxiv.org/abs/2103.13597v1","paper_title":"Mask Attention Networks: Rethinking and Strengthen Transformer","code":"https://github.com/libertfan/man","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":3,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":12,"model":"Transformer (ADMIN init)","metrics":{"BLEU score":"30.1","Number of Params":"256M","SacreBLEU":"29.5"},"uses_additional_data":false,"paper_date":"2020-08-18","paper":"/paper/very-deep-transformers-for-neural-machine","paper_url":"https://arxiv.org/abs/2008.07772v2","paper_title":"Very Deep Transformers for Neural Machine Translation","code":"https://github.com/LiyuanLucasLiu/Transforemr-Clinic","n_code_links":4,"syntology":{"n_ran":8,"n_unverified":1,"n_samples":9,"n_pointer_only_licence":0}},{"rank_in_archive_order":13,"model":"PowerNorm (Transformer)","metrics":{"BLEU score":"30.1"},"uses_additional_data":false,"paper_date":"2020-03-17","paper":"/paper/rethinking-batch-normalization-in","paper_url":"https://arxiv.org/abs/2003.07845v2","paper_title":"PowerNorm: Rethinking Batch Normalization in Transformers","code":"https://github.com/sIncerass/powernorm","n_code_links":1,"syntology":null},{"rank_in_archive_order":14,"model":"Depth Growing","metrics":{"BLEU score":"30.07","Hardware Burden":"24G"},"uses_additional_data":false,"paper_date":"2019-07-03","paper":"/paper/depth-growing-for-neural-machine-translation","paper_url":"https://arxiv.org/abs/1907.01968v1","paper_title":"Depth Growing for Neural Machine Translation","code":"https://github.com/apeterswu/Depth_Growing_NMT","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"MUSE(Parallel Multi-scale Attention)","metrics":{"BLEU score":"29.9"},"uses_additional_data":false,"paper_date":"2019-11-17","paper":"/paper/muse-parallel-multi-scale-attention-for","paper_url":"https://arxiv.org/abs/1911.09483v1","paper_title":"MUSE: Parallel Multi-Scale Attention for Sequence to Sequence Learning","code":"https://github.com/lancopku/MUSE","n_code_links":3,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":16,"model":"Evolved Transformer Big","metrics":{"BLEU score":"29.8","Number of Params":"218M","SacreBLEU":"29.2"},"uses_additional_data":false,"paper_date":"2019-01-30","paper":"/paper/the-evolved-transformer","paper_url":"https://arxiv.org/abs/1901.11117v4","paper_title":"The Evolved Transformer","code":"https://github.com/tensorflow/tensor2tensor","n_code_links":3,"syntology":null},{"rank_in_archive_order":17,"model":"OmniNetP","metrics":{"BLEU score":"29.8"},"uses_additional_data":false,"paper_date":"2021-03-01","paper":"/paper/omninet-omnidirectional-representations-from","paper_url":"https://arxiv.org/abs/2103.01075v1","paper_title":"OmniNet: Omnidirectional Representations from Transformers","code":"https://github.com/lucidrains/omninet-pytorch","n_code_links":1,"syntology":null},{"rank_in_archive_order":18,"model":"DynamicConv","metrics":{"BLEU score":"29.7","Number of Params":"213M"},"uses_additional_data":false,"paper_date":"2019-01-29","paper":"/paper/pay-less-attention-with-lightweight-and","paper_url":"http://arxiv.org/abs/1901.10430v2","paper_title":"Pay Less Attention with Lightweight and Dynamic Convolutions","code":"https://github.com/pytorch/fairseq","n_code_links":3,"syntology":null},{"rank_in_archive_order":19,"model":"Local Joint Self-attention","metrics":{"BLEU score":"29.7"},"uses_additional_data":false,"paper_date":"2019-05-16","paper":"/paper/190506596","paper_url":"https://arxiv.org/abs/1905.06596v1","paper_title":"Joint Source-Target Self Attention with Locality Constraints","code":"https://github.com/jarfo/joint","n_code_links":2,"syntology":null},{"rank_in_archive_order":20,"model":"TaLK Convolutions","metrics":{"BLEU score":"29.6","Number of Params":"209M"},"uses_additional_data":false,"paper_date":"2020-02-08","paper":"/paper/time-aware-large-kernel-convolutions","paper_url":"https://arxiv.org/abs/2002.03184v2","paper_title":"Time-aware Large Kernel Convolutions","code":"https://github.com/lioutasb/TaLKConvolutions","n_code_links":1,"syntology":{"n_ran":6,"n_unverified":1,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":21,"model":"Transformer Big + MoS","metrics":{"BLEU score":"29.6"},"uses_additional_data":false,"paper_date":"2018-09-25","paper":"/paper/fast-and-simple-mixture-of-softmaxes-with-bpe","paper_url":"https://arxiv.org/abs/1809.09296v2","paper_title":"Fast and Simple Mixture of Softmaxes with BPE and Hybrid-LightRNN for Language Generation","code":"https://github.com/shawnkx/Fast-MoS","n_code_links":1,"syntology":null},{"rank_in_archive_order":22,"model":"AdvAug (aut+adv)","metrics":{"BLEU score":"29.57"},"uses_additional_data":false,"paper_date":"2020-06-21","paper":"/paper/advaug-robust-adversarial-augmentation-for-1","paper_url":"https://arxiv.org/abs/2006.11834v3","paper_title":"AdvAug: Robust Adversarial Augmentation for Neural Machine Translation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":23,"model":"PartialFormer","metrics":{"BLEU score":"29.56","Number of Params":"68M"},"uses_additional_data":false,"paper_date":"2023-10-23","paper":"/paper/partialformer-modeling-part-instead-of-whole","paper_url":"https://arxiv.org/abs/2310.14921v2","paper_title":"PartialFormer: Modeling Part Instead of Whole for Machine Translation","code":"https://github.com/zhengkid/partialformer","n_code_links":1,"syntology":null},{"rank_in_archive_order":24,"model":"Transformer Big + adversarial MLE","metrics":{"BLEU score":"29.52"},"uses_additional_data":false,"paper_date":"2019-06-10","paper":"/paper/improving-neural-language-modeling-via","paper_url":"https://arxiv.org/abs/1906.03805v2","paper_title":"Improving Neural Language Modeling via Adversarial Training","code":"https://github.com/ChengyueGongR/advsoft","n_code_links":1,"syntology":null},{"rank_in_archive_order":25,"model":"Transformer Big","metrics":{"BLEU score":"29.3","Hardware Burden":"9G","Number of Params":"210M"},"uses_additional_data":false,"paper_date":"2018-06-01","paper":"/paper/scaling-neural-machine-translation","paper_url":"http://arxiv.org/abs/1806.00187v3","paper_title":"Scaling Neural Machine Translation","code":"https://github.com/facebookresearch/fairseq","n_code_links":5,"syntology":null},{"rank_in_archive_order":26,"model":"Subformer-xlarge","metrics":{"BLEU score":"29.3"},"uses_additional_data":false,"paper_date":"2021-01-01","paper":"/paper/subformer-a-parameter-reduced-transformer","paper_url":"https://openreview.net/forum?id=6UurSaf08jx","paper_title":"Subformer: A Parameter Reduced Transformer","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":27,"model":"SB-NMT","metrics":{"BLEU score":"29.21"},"uses_additional_data":false,"paper_date":"2019-05-13","paper":"/paper/synchronous-bidirectional-neural-machine","paper_url":"https://arxiv.org/abs/1905.04847v1","paper_title":"Synchronous Bidirectional Neural Machine Translation","code":"https://github.com/wszlong/sb-nmt","n_code_links":2,"syntology":null},{"rank_in_archive_order":28,"model":"Transformer (big) + Relative Position Representations","metrics":{"BLEU score":"29.2"},"uses_additional_data":false,"paper_date":"2018-03-06","paper":"/paper/self-attention-with-relative-position","paper_url":"http://arxiv.org/abs/1803.02155v2","paper_title":"Self-Attention with Relative Position Representations","code":"https://github.com/tensorflow/tensor2tensor","n_code_links":13,"syntology":{"n_ran":13,"n_unverified":10,"n_samples":23,"n_pointer_only_licence":3}},{"rank_in_archive_order":29,"model":"FLOATER-large","metrics":{"BLEU score":"29.2"},"uses_additional_data":false,"paper_date":"2020-03-13","paper":"/paper/learning-to-encode-position-for-transformer","paper_url":"https://arxiv.org/abs/2003.09229v1","paper_title":"Learning to Encode Position for Transformer with Continuous Dynamical Model","code":"https://github.com/xuanqing94/FLOATER","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":3,"n_samples":6,"n_pointer_only_licence":6}},{"rank_in_archive_order":30,"model":"Local Transformer","metrics":{"BLEU score":"29.2"},"uses_additional_data":false,"paper_date":"2018-10-24","paper":"/paper/modeling-localness-for-self-attention","paper_url":"http://arxiv.org/abs/1810.10182v1","paper_title":"Modeling Localness for Self-Attention Networks","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":31,"model":"Transformer Big with FRAGE","metrics":{"BLEU score":"29.11"},"uses_additional_data":false,"paper_date":"2018-09-18","paper":"/paper/frage-frequency-agnostic-word-representation","paper_url":"https://arxiv.org/abs/1809.06858v2","paper_title":"FRAGE: Frequency-Agnostic Word Representation","code":"https://github.com/ChengyueGongR/FrequencyAgnostic","n_code_links":2,"syntology":null},{"rank_in_archive_order":32,"model":"Mask Attention Network (base)","metrics":{"BLEU score":"29.1","Number of Params":"63M"},"uses_additional_data":false,"paper_date":"2021-03-25","paper":"/paper/mask-attention-networks-rethinking-and","paper_url":"https://arxiv.org/abs/2103.13597v1","paper_title":"Mask Attention Networks: Rethinking and Strengthen Transformer","code":"https://github.com/libertfan/man","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":3,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":33,"model":"Mega","metrics":{"BLEU score":"29.01","Number of Params":"67M","SacreBLEU":"27.96"},"uses_additional_data":false,"paper_date":"2022-09-21","paper":"/paper/mega-moving-average-equipped-gated-attention","paper_url":"https://arxiv.org/abs/2209.10655v3","paper_title":"Mega: Moving Average Equipped Gated Attention","code":"https://github.com/huggingface/transformers","n_code_links":7,"syntology":{"n_ran":12,"n_unverified":1,"n_samples":13,"n_pointer_only_licence":8}},{"rank_in_archive_order":34,"model":"adequacy-oriented NMT","metrics":{"BLEU score":"28.99"},"uses_additional_data":false,"paper_date":"2018-11-21","paper":"/paper/neural-machine-translation-with-adequacy","paper_url":"http://arxiv.org/abs/1811.08541v1","paper_title":"Neural Machine Translation with Adequacy-Oriented Learning","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":35,"model":"LightConv","metrics":{"BLEU score":"28.9","Number of Params":"202M"},"uses_additional_data":false,"paper_date":"2019-01-29","paper":"/paper/pay-less-attention-with-lightweight-and","paper_url":"http://arxiv.org/abs/1901.10430v2","paper_title":"Pay Less Attention with Lightweight and Dynamic Convolutions","code":"https://github.com/pytorch/fairseq","n_code_links":3,"syntology":null},{"rank_in_archive_order":36,"model":"Weighted Transformer (large)","metrics":{"BLEU score":"28.9"},"uses_additional_data":false,"paper_date":"2017-11-06","paper":"/paper/weighted-transformer-network-for-machine","paper_url":"http://arxiv.org/abs/1711.02132v1","paper_title":"Weighted Transformer Network for Machine Translation","code":"https://github.com/JayParks/transformer","n_code_links":5,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":37,"model":"universal transformer base","metrics":{"BLEU score":"28.9"},"uses_additional_data":false,"paper_date":"2018-07-10","paper":"/paper/universal-transformers","paper_url":"http://arxiv.org/abs/1807.03819v3","paper_title":"Universal Transformers","code":"https://github.com/tensorflow/tensor2tensor","n_code_links":8,"syntology":{"n_ran":16,"n_unverified":9,"n_samples":25,"n_pointer_only_licence":24}},{"rank_in_archive_order":38,"model":"KERMIT","metrics":{"BLEU score":"28.7"},"uses_additional_data":false,"paper_date":"2019-06-04","paper":"/paper/kermit-generative-insertion-based-modeling","paper_url":"https://arxiv.org/abs/1906.01604v1","paper_title":"KERMIT: Generative Insertion-Based Modeling for Sequences","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":39,"model":"T2R + Pretrain","metrics":{"BLEU score":"28.7"},"uses_additional_data":false,"paper_date":"2021-03-24","paper":"/paper/finetuning-pretrained-transformers-into-rnns","paper_url":"https://arxiv.org/abs/2103.13076v2","paper_title":"Finetuning Pretrained Transformers into RNNs","code":"https://github.com/hazyresearch/lolcats","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":5,"n_samples":8,"n_pointer_only_licence":4}},{"rank_in_archive_order":40,"model":"AdvAug (aut)","metrics":{"BLEU score":"28.58"},"uses_additional_data":false,"paper_date":"2020-06-21","paper":"/paper/advaug-robust-adversarial-augmentation-for-1","paper_url":"https://arxiv.org/abs/2006.11834v3","paper_title":"AdvAug: Robust Adversarial Augmentation for Neural Machine Translation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":41,"model":"RNMT+","metrics":{"BLEU score":"28.5","Hardware Burden":"44G","Operations per network pass":"2.81G"},"uses_additional_data":false,"paper_date":"2018-04-26","paper":"/paper/the-best-of-both-worlds-combining-recent","paper_url":"http://arxiv.org/abs/1804.09849v2","paper_title":"The Best of Both Worlds: Combining Recent Advances in Neural Machine Translation","code":"https://github.com/tensorflow/lingvo","n_code_links":3,"syntology":null},{"rank_in_archive_order":42,"model":"Synthesizer (Random + Vanilla)","metrics":{"BLEU score":"28.47"},"uses_additional_data":false,"paper_date":"2020-05-02","paper":"/paper/synthesizer-rethinking-self-attention-in","paper_url":"https://arxiv.org/abs/2005.00743v3","paper_title":"Synthesizer: Rethinking Self-Attention in Transformer Models","code":"https://github.com/10-zin/Synthesizer","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":43,"model":"Hardware Aware Transformer","metrics":{"BLEU score":"28.4","Number of Params":"48M"},"uses_additional_data":false,"paper_date":"2020-05-28","paper":"/paper/hat-hardware-aware-transformers-for-efficient","paper_url":"https://arxiv.org/abs/2005.14187v1","paper_title":"HAT: Hardware-Aware Transformers for Efficient Natural Language Processing","code":"https://github.com/mit-han-lab/hardware-aware-transformers","n_code_links":4,"syntology":null},{"rank_in_archive_order":44,"model":"Transformer Big","metrics":{"BLEU score":"28.4","Hardware Burden":"871G","Operations per network pass":"2300000000.0G"},"uses_additional_data":false,"paper_date":"2017-06-12","paper":"/paper/attention-is-all-you-need","paper_url":"https://arxiv.org/abs/1706.03762v7","paper_title":"Attention Is All You Need","code":"https://github.com/huggingface/transformers","n_code_links":595,"syntology":{"n_ran":610,"n_unverified":336,"n_samples":946,"n_pointer_only_licence":451}},{"rank_in_archive_order":45,"model":"Transformer + SRU","metrics":{"BLEU score":"28.4","Hardware Burden":"34G"},"uses_additional_data":false,"paper_date":"2017-09-08","paper":"/paper/simple-recurrent-units-for-highly","paper_url":"http://arxiv.org/abs/1709.02755v5","paper_title":"Simple Recurrent Units for Highly Parallelizable Recurrence","code":"https://github.com/aymericdamien/TopDeepLearning","n_code_links":11,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":46,"model":"Evolved Transformer Base","metrics":{"BLEU score":"28.4","Hardware Burden":"2488G"},"uses_additional_data":false,"paper_date":"2019-01-30","paper":"/paper/the-evolved-transformer","paper_url":"https://arxiv.org/abs/1901.11117v4","paper_title":"The Evolved Transformer","code":"https://github.com/tensorflow/tensor2tensor","n_code_links":3,"syntology":null},{"rank_in_archive_order":47,"model":"Rfa-Gate-arccos","metrics":{"BLEU score":"28.2"},"uses_additional_data":false,"paper_date":"2021-03-03","paper":"/paper/random-feature-attention-1","paper_url":"https://arxiv.org/abs/2103.02143v2","paper_title":"Random Feature Attention","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":48,"model":"Transformer-DRILL Base","metrics":{"BLEU score":"28.1"},"uses_additional_data":false,"paper_date":"2019-05-14","paper":"/paper/deep-residual-output-layers-for-neural","paper_url":"https://arxiv.org/abs/1905.05513v2","paper_title":"Deep Residual Output Layers for Neural Language Generation","code":"https://github.com/idiap/drill","n_code_links":1,"syntology":null},{"rank_in_archive_order":49,"model":"AdvAug (mixup)","metrics":{"BLEU score":"28.08"},"uses_additional_data":false,"paper_date":"2020-06-21","paper":"/paper/advaug-robust-adversarial-augmentation-for-1","paper_url":"https://arxiv.org/abs/2006.11834v3","paper_title":"AdvAug: Robust Adversarial Augmentation for Neural Machine Translation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":50,"model":"CMLM+LAT+4 iterations","metrics":{"BLEU score":"27.35"},"uses_additional_data":false,"paper_date":"2020-11-12","paper":"/paper/incorporating-a-local-translation-mechanism","paper_url":"https://arxiv.org/abs/2011.06132v1","paper_title":"Incorporating a Local Translation Mechanism into Non-autoregressive Translation","code":"https://github.com/shawnkx/NAT-with-Local-AT","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":2,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":51,"model":"Transformer Base","metrics":{"BLEU score":"27.3","Operations per network pass":"330000000.0G"},"uses_additional_data":false,"paper_date":"2017-06-12","paper":"/paper/attention-is-all-you-need","paper_url":"https://arxiv.org/abs/1706.03762v7","paper_title":"Attention Is All You Need","code":"https://github.com/huggingface/transformers","n_code_links":595,"syntology":{"n_ran":610,"n_unverified":336,"n_samples":946,"n_pointer_only_licence":451}},{"rank_in_archive_order":52,"model":"Levenshtein Transformer (distillation)","metrics":{"BLEU score":"27.27"},"uses_additional_data":false,"paper_date":"2019-05-27","paper":"/paper/levenshtein-transformer","paper_url":"https://arxiv.org/abs/1905.11006v2","paper_title":"Levenshtein Transformer","code":"https://github.com/pytorch/fairseq","n_code_links":3,"syntology":null},{"rank_in_archive_order":53,"model":"DisCo + Mask-Predict (non-autoregressive)","metrics":{"BLEU score":"27.06"},"uses_additional_data":false,"paper_date":null,"paper":"/paper/non-autoregressive-translation-with","paper_url":"https://proceedings.icml.cc/static/paper_files/icml/2020/477-Paper.pdf","paper_title":"Non-autoregressive Translation with Disentangled Context Transformer","code":"https://github.com/facebookresearch/DisCo","n_code_links":1,"syntology":null},{"rank_in_archive_order":54,"model":"Adaptively Sparse Transformer (alpha-entmax)","metrics":{"BLEU score":"26.93"},"uses_additional_data":false,"paper_date":"2019-08-30","paper":"/paper/adaptively-sparse-transformers","paper_url":"https://arxiv.org/abs/1909.00015v2","paper_title":"Adaptively Sparse Transformers","code":"https://github.com/deep-spin/entmax","n_code_links":3,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":55,"model":"ResMLP-12","metrics":{"BLEU score":"26.8"},"uses_additional_data":false,"paper_date":"2021-05-07","paper":"/paper/resmlp-feedforward-networks-for-image","paper_url":"https://arxiv.org/abs/2105.03404v2","paper_title":"ResMLP: Feedforward networks for image classification with data-efficient training","code":"https://github.com/rwightman/pytorch-image-models","n_code_links":19,"syntology":{"n_ran":5,"n_unverified":2,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":56,"model":"CNAT","metrics":{"BLEU score":"26.6"},"uses_additional_data":false,"paper_date":"2021-03-21","paper":"/paper/non-autoregressive-translation-by-learning","paper_url":"https://arxiv.org/abs/2103.11405v1","paper_title":"Non-Autoregressive Translation by Learning Target Categorical Codes","code":"https://github.com/baoy-nlp/CNAT","n_code_links":1,"syntology":null},{"rank_in_archive_order":57,"model":"Lite Transformer","metrics":{"BLEU score":"26.5","Number of Params":"17.3M"},"uses_additional_data":false,"paper_date":"2020-04-24","paper":"/paper/lite-transformer-with-long-short-range","paper_url":"https://arxiv.org/abs/2004.11886v1","paper_title":"Lite Transformer with Long-Short Range Attention","code":"https://github.com/mit-han-lab/lite-transformer","n_code_links":2,"syntology":null},{"rank_in_archive_order":58,"model":"ConvS2S (ensemble)","metrics":{"BLEU score":"26.4","Hardware Burden":"54G"},"uses_additional_data":false,"paper_date":"2017-05-08","paper":"/paper/convolutional-sequence-to-sequence-learning","paper_url":"http://arxiv.org/abs/1705.03122v3","paper_title":"Convolutional Sequence to Sequence Learning","code":"https://github.com/facebookresearch/fairseq","n_code_links":37,"syntology":null},{"rank_in_archive_order":59,"model":"ResMLP-6","metrics":{"BLEU score":"26.4"},"uses_additional_data":false,"paper_date":"2021-05-07","paper":"/paper/resmlp-feedforward-networks-for-image","paper_url":"https://arxiv.org/abs/2105.03404v2","paper_title":"ResMLP: Feedforward networks for image classification with data-efficient training","code":"https://github.com/rwightman/pytorch-image-models","n_code_links":19,"syntology":{"n_ran":5,"n_unverified":2,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":60,"model":"Average Attention Network","metrics":{"BLEU score":"26.31"},"uses_additional_data":false,"paper_date":"2018-05-02","paper":"/paper/accelerating-neural-transformer-via-an","paper_url":"http://arxiv.org/abs/1805.00631v3","paper_title":"Accelerating Neural Transformer via an Average Attention Network","code":"https://github.com/bzhangXMU/transformer-aan","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":19,"n_samples":19,"n_pointer_only_licence":0}},{"rank_in_archive_order":61,"model":"GNMT+RL","metrics":{"BLEU score":"26.3"},"uses_additional_data":false,"paper_date":"2016-09-26","paper":"/paper/googles-neural-machine-translation-system","paper_url":"http://arxiv.org/abs/1609.08144v2","paper_title":"Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation","code":"https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/Translation/GNMT","n_code_links":28,"syntology":{"n_ran":23,"n_unverified":23,"n_samples":46,"n_pointer_only_licence":12}},{"rank_in_archive_order":62,"model":"SliceNet","metrics":{"BLEU score":"26.1"},"uses_additional_data":false,"paper_date":"2017-06-09","paper":"/paper/depthwise-separable-convolutions-for-neural","paper_url":"http://arxiv.org/abs/1706.03059v2","paper_title":"Depthwise Separable Convolutions for Neural Machine Translation","code":"https://github.com/tensorflow/tensor2tensor","n_code_links":2,"syntology":null},{"rank_in_archive_order":63,"model":"Average Attention Network (w/o FFN)","metrics":{"BLEU score":"26.05"},"uses_additional_data":false,"paper_date":"2018-05-02","paper":"/paper/accelerating-neural-transformer-via-an","paper_url":"http://arxiv.org/abs/1805.00631v3","paper_title":"Accelerating Neural Transformer via an Average Attention Network","code":"https://github.com/bzhangXMU/transformer-aan","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":19,"n_samples":19,"n_pointer_only_licence":0}},{"rank_in_archive_order":64,"model":"MoE","metrics":{"BLEU score":"26.03","Hardware Burden":"24G"},"uses_additional_data":false,"paper_date":"2017-01-23","paper":"/paper/outrageously-large-neural-networks-the","paper_url":"http://arxiv.org/abs/1701.06538v1","paper_title":"Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer","code":"https://github.com/davidmrau/mixture-of-experts","n_code_links":4,"syntology":{"n_ran":4,"n_unverified":2,"n_samples":6,"n_pointer_only_licence":6}},{"rank_in_archive_order":65,"model":"Average Attention Network (w/o gate)","metrics":{"BLEU score":"25.91"},"uses_additional_data":false,"paper_date":"2018-05-02","paper":"/paper/accelerating-neural-transformer-via-an","paper_url":"http://arxiv.org/abs/1805.00631v3","paper_title":"Accelerating Neural Transformer via an Average Attention Network","code":"https://github.com/bzhangXMU/transformer-aan","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":19,"n_samples":19,"n_pointer_only_licence":0}},{"rank_in_archive_order":66,"model":"Adaptively Sparse Transformer (1.5-entmax)","metrics":{"BLEU score":"25.89"},"uses_additional_data":false,"paper_date":"2019-08-30","paper":"/paper/adaptively-sparse-transformers","paper_url":"https://arxiv.org/abs/1909.00015v2","paper_title":"Adaptively Sparse Transformers","code":"https://github.com/deep-spin/entmax","n_code_links":3,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":67,"model":"DenseNMT","metrics":{"BLEU score":"25.52"},"uses_additional_data":false,"paper_date":"2018-06-03","paper":"/paper/dense-information-flow-for-neural-machine","paper_url":"http://arxiv.org/abs/1806.00722v2","paper_title":"Dense Information Flow for Neural Machine Translation","code":"https://github.com/yanyao-shen/fairseq","n_code_links":1,"syntology":null},{"rank_in_archive_order":68,"model":"GLAT","metrics":{"BLEU score":"25.21"},"uses_additional_data":false,"paper_date":"2020-08-18","paper":"/paper/glancing-transformer-for-non-autoregressive","paper_url":"https://arxiv.org/abs/2008.07905v3","paper_title":"Glancing Transformer for Non-Autoregressive Neural Machine Translation","code":"https://github.com/flc777/glat","n_code_links":2,"syntology":null},{"rank_in_archive_order":69,"model":"CMLM+LAT+1 iterations","metrics":{"BLEU score":"25.20"},"uses_additional_data":false,"paper_date":"2020-11-12","paper":"/paper/incorporating-a-local-translation-mechanism","paper_url":"https://arxiv.org/abs/2011.06132v1","paper_title":"Incorporating a Local Translation Mechanism into Non-autoregressive Translation","code":"https://github.com/shawnkx/NAT-with-Local-AT","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":2,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":70,"model":"ConvS2S","metrics":{"BLEU score":"25.16","Hardware Burden":"72G"},"uses_additional_data":false,"paper_date":"2017-05-08","paper":"/paper/convolutional-sequence-to-sequence-learning","paper_url":"http://arxiv.org/abs/1705.03122v3","paper_title":"Convolutional Sequence to Sequence Learning","code":"https://github.com/facebookresearch/fairseq","n_code_links":37,"syntology":null},{"rank_in_archive_order":71,"model":"ByteNet","metrics":{"BLEU score":"23.75"},"uses_additional_data":false,"paper_date":"2016-10-31","paper":"/paper/neural-machine-translation-in-linear-time","paper_url":"http://arxiv.org/abs/1610.10099v2","paper_title":"Neural Machine Translation in Linear Time","code":"https://github.com/paarthneekhara/byteNet-tensorflow","n_code_links":11,"syntology":null},{"rank_in_archive_order":72,"model":"FlowSeq-large (NPD n = 30)","metrics":{"BLEU score":"23.64"},"uses_additional_data":false,"paper_date":"2019-09-05","paper":"/paper/flowseq-non-autoregressive-conditional","paper_url":"https://arxiv.org/abs/1909.02480v3","paper_title":"FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow","code":"https://github.com/XuezheMax/flowseq","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":73,"model":"FlowSeq-large (NPD n = 15)","metrics":{"BLEU score":"23.14"},"uses_additional_data":false,"paper_date":"2019-09-05","paper":"/paper/flowseq-non-autoregressive-conditional","paper_url":"https://arxiv.org/abs/1909.02480v3","paper_title":"FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow","code":"https://github.com/XuezheMax/flowseq","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":74,"model":"FlowSeq-large (IWD n = 15)","metrics":{"BLEU score":"22.94"},"uses_additional_data":false,"paper_date":"2019-09-05","paper":"/paper/flowseq-non-autoregressive-conditional","paper_url":"https://arxiv.org/abs/1909.02480v3","paper_title":"FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow","code":"https://github.com/XuezheMax/flowseq","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":75,"model":"Denoising autoencoders (non-autoregressive)","metrics":{"BLEU score":"21.54"},"uses_additional_data":false,"paper_date":"2018-02-19","paper":"/paper/deterministic-non-autoregressive-neural","paper_url":"http://arxiv.org/abs/1802.06901v3","paper_title":"Deterministic Non-Autoregressive Neural Sequence Modeling by Iterative Refinement","code":"https://github.com/nyu-dl/dl4mt-nonauto","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":1,"n_samples":4,"n_pointer_only_licence":1}},{"rank_in_archive_order":76,"model":"RNN Enc-Dec Att","metrics":{"BLEU score":"20.9"},"uses_additional_data":false,"paper_date":"2015-08-17","paper":"/paper/effective-approaches-to-attention-based","paper_url":"http://arxiv.org/abs/1508.04025v5","paper_title":"Effective Approaches to Attention-based Neural Machine Translation","code":"https://github.com/philipperemy/keras-attention-mechanism","n_code_links":44,"syntology":{"n_ran":2,"n_unverified":6,"n_samples":8,"n_pointer_only_licence":2}},{"rank_in_archive_order":77,"model":"FlowSeq-large","metrics":{"BLEU score":"20.85"},"uses_additional_data":false,"paper_date":"2019-09-05","paper":"/paper/flowseq-non-autoregressive-conditional","paper_url":"https://arxiv.org/abs/1909.02480v3","paper_title":"FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow","code":"https://github.com/XuezheMax/flowseq","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":78,"model":"PBMT","metrics":{"BLEU score":"20.7"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":79,"model":"Deep-Att","metrics":{"BLEU score":"20.7","Hardware Burden":"119G"},"uses_additional_data":false,"paper_date":"2016-06-14","paper":"/paper/deep-recurrent-models-with-fast-forward","paper_url":"http://arxiv.org/abs/1606.04199v3","paper_title":"Deep Recurrent Models with Fast-Forward Connections for Neural Machine Translation","code":"https://github.com/fstahlberg/tensor2tensor-usr","n_code_links":1,"syntology":null},{"rank_in_archive_order":80,"model":"Phrase Based MT","metrics":{"BLEU score":"20.7"},"uses_additional_data":false,"paper_date":"2015-09-01","paper":"/paper/edinburghs-syntax-based-systems-at-wmt-2015","paper_url":"https://aclanthology.org/W15-3024","paper_title":"Edinburgh's Syntax-Based Systems at WMT 2015","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":81,"model":"PBSMT + NMT","metrics":{"BLEU score":"20.23"},"uses_additional_data":false,"paper_date":"2018-04-20","paper":"/paper/phrase-based-neural-unsupervised-machine","paper_url":"http://arxiv.org/abs/1804.07755v2","paper_title":"Phrase-Based & Neural Unsupervised Machine Translation","code":"https://github.com/huggingface/transformers","n_code_links":14,"syntology":null},{"rank_in_archive_order":82,"model":"NAT +FT + NPD","metrics":{"BLEU score":"19.17"},"uses_additional_data":false,"paper_date":"2017-11-07","paper":"/paper/non-autoregressive-neural-machine-translation-1","paper_url":"http://arxiv.org/abs/1711.02281v2","paper_title":"Non-Autoregressive Neural Machine Translation","code":"https://github.com/salesforce/nonauto-nmt","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":1}},{"rank_in_archive_order":83,"model":"FlowSeq-base","metrics":{"BLEU score":"18.55"},"uses_additional_data":false,"paper_date":"2019-09-05","paper":"/paper/flowseq-non-autoregressive-conditional","paper_url":"https://arxiv.org/abs/1909.02480v3","paper_title":"FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow","code":"https://github.com/XuezheMax/flowseq","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":84,"model":"Seq-KD + Seq-Inter + Word-KD","metrics":{"BLEU score":"18.5"},"uses_additional_data":false,"paper_date":"2016-06-25","paper":"/paper/sequence-level-knowledge-distillation","paper_url":"http://arxiv.org/abs/1606.07947v4","paper_title":"Sequence-Level Knowledge Distillation","code":"https://github.com/harvardnlp/seq2seq-attn","n_code_links":6,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":85,"model":"Unsupervised PBSMT","metrics":{"BLEU score":"17.94"},"uses_additional_data":false,"paper_date":"2018-04-20","paper":"/paper/phrase-based-neural-unsupervised-machine","paper_url":"http://arxiv.org/abs/1804.07755v2","paper_title":"Phrase-Based & Neural Unsupervised Machine Translation","code":"https://github.com/huggingface/transformers","n_code_links":14,"syntology":null},{"rank_in_archive_order":86,"model":"NSE-NSE","metrics":{"BLEU score":"17.9"},"uses_additional_data":false,"paper_date":"2016-07-14","paper":"/paper/neural-semantic-encoders","paper_url":"http://arxiv.org/abs/1607.04315v3","paper_title":"Neural Semantic Encoders","code":"https://github.com/Smerity/keras_snli","n_code_links":3,"syntology":null},{"rank_in_archive_order":87,"model":"Unsupervised NMT + Transformer","metrics":{"BLEU score":"17.16"},"uses_additional_data":false,"paper_date":"2018-04-20","paper":"/paper/phrase-based-neural-unsupervised-machine","paper_url":"http://arxiv.org/abs/1804.07755v2","paper_title":"Phrase-Based & Neural Unsupervised Machine Translation","code":"https://github.com/huggingface/transformers","n_code_links":14,"syntology":null},{"rank_in_archive_order":88,"model":"SMT + iterative backtranslation (unsupervised)","metrics":{"BLEU score":"14.08"},"uses_additional_data":false,"paper_date":"2018-09-04","paper":"/paper/unsupervised-statistical-machine-translation","paper_url":"http://arxiv.org/abs/1809.01272v1","paper_title":"Unsupervised Statistical Machine Translation","code":"https://github.com/artetxem/vecmap","n_code_links":3,"syntology":null},{"rank_in_archive_order":89,"model":"Reverse RNN Enc-Dec","metrics":{"BLEU score":"14.0"},"uses_additional_data":false,"paper_date":"2015-08-17","paper":"/paper/effective-approaches-to-attention-based","paper_url":"http://arxiv.org/abs/1508.04025v5","paper_title":"Effective Approaches to Attention-based Neural Machine Translation","code":"https://github.com/philipperemy/keras-attention-mechanism","n_code_links":44,"syntology":{"n_ran":2,"n_unverified":6,"n_samples":8,"n_pointer_only_licence":2}},{"rank_in_archive_order":90,"model":"RNN Enc-Dec","metrics":{"BLEU score":"11.3"},"uses_additional_data":false,"paper_date":"2015-08-17","paper":"/paper/effective-approaches-to-attention-based","paper_url":"http://arxiv.org/abs/1508.04025v5","paper_title":"Effective Approaches to Attention-based Neural Machine Translation","code":"https://github.com/philipperemy/keras-attention-mechanism","n_code_links":44,"syntology":{"n_ran":2,"n_unverified":6,"n_samples":8,"n_pointer_only_licence":2}},{"rank_in_archive_order":91,"model":"MAT","metrics":{"SacreBLEU":"29.9"},"uses_additional_data":false,"paper_date":"2020-06-18","paper":"/paper/multi-branch-attentive-transformer","paper_url":"https://arxiv.org/abs/2006.10270v2","paper_title":"Multi-branch Attentive Transformer","code":"https://github.com/HA-Transformer/HA-Transformer","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}}],"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 7,081 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":7081,"papers_extracted_not_yet_verified":217,"boards_without_verdict":27,"papers_not_yet_extracted":2325},"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":45,"rows_with_any_sample_ran":38,"distinct_papers_with_graph_line":32,"distinct_papers_with_any_sample_ran":28,"samples_over_distinct_papers":{"n_ran":758,"n_unverified":440,"n_samples":1198,"n_pointer_only_licence":531,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":1387,"n_unverified":833,"n_samples":2220,"n_pointer_only_licence":986,"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"}}}