{"url":"/sota/language-modelling-on-text8","task":{"name":"Language Modelling","url":"/task/language-modelling","note":null},"dataset":{"name":"Text8","url":"/dataset/text8"},"category":"Natural Language Processing","categories":["Medical","Miscellaneous","Natural Language Processing"],"category_note":null,"description":"A language model is a model of natural language. Language models are useful for a variety of tasks, including speech recognition, machine translation, natural language generation (generating more human-like text), optical character recognition, route optimization, handwriting recognition, grammar induction, and information retrieval.\r\n\r\nLarge language models (LLMs), currently their most advanced form, are predominantly based on transformers trained on larger datasets (frequently using words scraped from the public internet). They have superseded recurrent neural network-based models, which had previously superseded the purely statistical models, such as word n-gram language model. \r\n\r\nSource: [Wikipedia](https://en.wikipedia.org/wiki/Language_model)","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":["Bit per Character (BPC)","Number of params"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Bit per Character (BPC)":null,"Number of params":"lower"}},"counts":{"rows":24,"rows_with_code":21,"rows_with_paper_page":24,"rows_dated":24,"rows_using_additional_data":1},"rows":[{"rank_in_archive_order":1,"model":"GPT-2","metrics":{"Bit per Character (BPC)":"0.98","Number of params":"1542M"},"uses_additional_data":true,"paper_date":"2019-02-14","paper":"/paper/language-models-are-unsupervised-multitask","paper_url":"https://d4mucfpksywv.cloudfront.net/better-language-models/language-models.pdf","paper_title":"Language Models are Unsupervised Multitask Learners","code":"https://github.com/huggingface/transformers","n_code_links":21,"syntology":null},{"rank_in_archive_order":2,"model":"Focus","metrics":{"Bit per Character (BPC)":"0.98","Number of params":"22M"},"uses_additional_data":false,"paper_date":"2023-05-24","paper":"/paper/2305-14952","paper_url":"https://arxiv.org/abs/2305.14952v2","paper_title":"Focus Your Attention (with Adaptive IIR Filters)","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"Transformer-XL + RMS dynamic eval + decay","metrics":{"Bit per Character (BPC)":"1.038","Number of params":"277M"},"uses_additional_data":false,"paper_date":"2019-04-17","paper":"/paper/dynamic-evaluation-of-transformer-language","paper_url":"http://arxiv.org/abs/1904.08378v1","paper_title":"Dynamic Evaluation of Transformer Language Models","code":"https://github.com/benkrause/dynamiceval-transformer","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"24L Transformer + 8K adaptive span","metrics":{"Bit per Character (BPC)":"1.07","Number of params":"209M"},"uses_additional_data":false,"paper_date":"2019-05-19","paper":"/paper/adaptive-attention-span-in-transformers","paper_url":"https://arxiv.org/abs/1905.07799v2","paper_title":"Adaptive Attention Span in Transformers","code":"https://github.com/facebookresearch/adaptive-span","n_code_links":8,"syntology":null},{"rank_in_archive_order":5,"model":"Transformer-XL - 24 layers","metrics":{"Bit per Character (BPC)":"1.08","Number of params":"277M"},"uses_additional_data":false,"paper_date":"2019-01-09","paper":"/paper/transformer-xl-attentive-language-models","paper_url":"https://arxiv.org/abs/1901.02860v3","paper_title":"Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context","code":"https://github.com/huggingface/transformers","n_code_links":37,"syntology":{"n_ran":63,"n_unverified":80,"n_samples":143,"n_pointer_only_licence":43}},{"rank_in_archive_order":6,"model":"All-attention network - 36 layers","metrics":{"Bit per Character (BPC)":"1.08","Number of params":"114M"},"uses_additional_data":false,"paper_date":"2019-07-02","paper":"/paper/augmenting-self-attention-with-persistent","paper_url":"https://arxiv.org/abs/1907.01470v1","paper_title":"Augmenting Self-attention with Persistent Memory","code":"https://github.com/lucidrains/x-transformers","n_code_links":2,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":4}},{"rank_in_archive_order":7,"model":"Transformer-LS (small)","metrics":{"Bit per Character (BPC)":"1.09"},"uses_additional_data":false,"paper_date":"2021-07-05","paper":"/paper/long-short-transformer-efficient-transformers","paper_url":"https://arxiv.org/abs/2107.02192v3","paper_title":"Long-Short Transformer: Efficient Transformers for Language and Vision","code":"https://github.com/keonlee9420/Comprehensive-Transformer-TTS","n_code_links":3,"syntology":{"n_ran":4,"n_unverified":0,"n_samples":4,"n_pointer_only_licence":2}},{"rank_in_archive_order":8,"model":"12L Transformer + 8K adaptive span","metrics":{"Bit per Character (BPC)":"1.11","Number of params":"38M"},"uses_additional_data":false,"paper_date":"2019-05-19","paper":"/paper/adaptive-attention-span-in-transformers","paper_url":"https://arxiv.org/abs/1905.07799v2","paper_title":"Adaptive Attention Span in Transformers","code":"https://github.com/facebookresearch/adaptive-span","n_code_links":8,"syntology":null},{"rank_in_archive_order":9,"model":"All-attention network - 18 layers","metrics":{"Bit per Character (BPC)":"1.11","Number of params":"38M"},"uses_additional_data":false,"paper_date":"2019-07-02","paper":"/paper/augmenting-self-attention-with-persistent","paper_url":"https://arxiv.org/abs/1907.01470v1","paper_title":"Augmenting Self-attention with Persistent Memory","code":"https://github.com/lucidrains/x-transformers","n_code_links":2,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":4}},{"rank_in_archive_order":10,"model":"BP-Transformer - 12 Layers","metrics":{"Bit per Character (BPC)":"1.11"},"uses_additional_data":false,"paper_date":"2019-11-11","paper":"/paper/bp-transformer-modelling-long-range-context","paper_url":"https://arxiv.org/abs/1911.04070v1","paper_title":"BP-Transformer: Modelling Long-Range Context via Binary Partitioning","code":"https://github.com/dmlc/dgl/tree/master/examples/pytorch/transformer","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":7,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"64-layer Character Transformer Model","metrics":{"Bit per Character (BPC)":"1.13","Number of params":"235M"},"uses_additional_data":false,"paper_date":"2018-08-09","paper":"/paper/character-level-language-modeling-with-deeper","paper_url":"http://arxiv.org/abs/1808.04444v2","paper_title":"Character-Level Language Modeling with Deeper Self-Attention","code":"https://github.com/facebookresearch/code-prediction-transformer","n_code_links":1,"syntology":null},{"rank_in_archive_order":12,"model":"GAM-RHN-10","metrics":{"Bit per Character (BPC)":"1.157","Number of params":"44.7M"},"uses_additional_data":false,"paper_date":"2019-12-13","paper":"/paper/recurrent-highway-networks-with-grouped","paper_url":"https://ieeexplore.ieee.org/document/8932404","paper_title":"Recurrent Highway Networks with Grouped Auxiliary Memory","code":"https://github.com/WilliamRo/gam_rhn","n_code_links":4,"syntology":null},{"rank_in_archive_order":13,"model":"12-layer Character Transformer Model","metrics":{"Bit per Character (BPC)":"1.18","Number of params":"44M"},"uses_additional_data":false,"paper_date":"2018-08-09","paper":"/paper/character-level-language-modeling-with-deeper","paper_url":"http://arxiv.org/abs/1808.04444v2","paper_title":"Character-Level Language Modeling with Deeper Self-Attention","code":"https://github.com/facebookresearch/code-prediction-transformer","n_code_links":1,"syntology":null},{"rank_in_archive_order":14,"model":"PAR Transformer 24B","metrics":{"Bit per Character (BPC)":"1.18"},"uses_additional_data":false,"paper_date":"2020-09-09","paper":"/paper/pay-attention-when-required","paper_url":"https://arxiv.org/abs/2009.04534v3","paper_title":"Pay Attention when Required","code":"https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/LanguageModeling","n_code_links":2,"syntology":null},{"rank_in_archive_order":15,"model":"mLSTM + dynamic eval","metrics":{"Bit per Character (BPC)":"1.19","Number of params":"45M"},"uses_additional_data":false,"paper_date":"2017-09-21","paper":"/paper/dynamic-evaluation-of-neural-sequence-models","paper_url":"http://arxiv.org/abs/1709.07432v2","paper_title":"Dynamic Evaluation of Neural Sequence Models","code":"https://github.com/benkrause/dynamic-evaluation","n_code_links":3,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"Bipartite flows (8 flows)","metrics":{"Bit per Character (BPC)":"1.23"},"uses_additional_data":false,"paper_date":"2019-05-24","paper":"/paper/discrete-flows-invertible-generative-models","paper_url":"https://arxiv.org/abs/1905.10347v1","paper_title":"Discrete Flows: Invertible Generative Models of Discrete Data","code":"https://github.com/google/edward2","n_code_links":2,"syntology":null},{"rank_in_archive_order":17,"model":"Large RHN","metrics":{"Bit per Character (BPC)":"1.27","Number of params":"46M"},"uses_additional_data":false,"paper_date":"2016-07-12","paper":"/paper/recurrent-highway-networks","paper_url":"http://arxiv.org/abs/1607.03474v5","paper_title":"Recurrent Highway Networks","code":"https://github.com/labmlai/annotated_deep_learning_paper_implementations","n_code_links":6,"syntology":{"n_ran":1,"n_unverified":4,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":18,"model":"Large mLSTM +emb +WN +VD","metrics":{"Bit per Character (BPC)":"1.27","Number of params":"45M"},"uses_additional_data":false,"paper_date":"2016-09-26","paper":"/paper/multiplicative-lstm-for-sequence-modelling","paper_url":"http://arxiv.org/abs/1609.07959v3","paper_title":"Multiplicative LSTM for sequence modelling","code":"https://github.com/astakara48/python_project","n_code_links":1,"syntology":null},{"rank_in_archive_order":19,"model":"LayerNorm HM-LSTM","metrics":{"Bit per Character (BPC)":"1.29","Number of params":"35M"},"uses_additional_data":false,"paper_date":"2016-09-06","paper":"/paper/hierarchical-multiscale-recurrent-neural","paper_url":"http://arxiv.org/abs/1609.01704v7","paper_title":"Hierarchical Multiscale Recurrent Neural Networks","code":"https://github.com/bolducp/hierarchical-rnn","n_code_links":3,"syntology":null},{"rank_in_archive_order":20,"model":"BN LSTM","metrics":{"Bit per Character (BPC)":"1.36","Number of params":"16M"},"uses_additional_data":false,"paper_date":"2016-03-30","paper":"/paper/recurrent-batch-normalization","paper_url":"http://arxiv.org/abs/1603.09025v5","paper_title":"Recurrent Batch Normalization","code":"https://github.com/cooijmanstim/recurrent-batch-normalization","n_code_links":3,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":21,"model":"Unregularised mLSTM","metrics":{"Bit per Character (BPC)":"1.40","Number of params":"45M"},"uses_additional_data":false,"paper_date":"2016-09-26","paper":"/paper/multiplicative-lstm-for-sequence-modelling","paper_url":"http://arxiv.org/abs/1609.07959v3","paper_title":"Multiplicative LSTM for sequence modelling","code":"https://github.com/astakara48/python_project","n_code_links":1,"syntology":null},{"rank_in_archive_order":22,"model":"BFN","metrics":{"Bit per Character (BPC)":"1.41"},"uses_additional_data":false,"paper_date":"2023-08-14","paper":"/paper/bayesian-flow-networks","paper_url":"https://arxiv.org/abs/2308.07037v5","paper_title":"Bayesian Flow Networks","code":"https://github.com/nnaisense/bayesian-flow-networks","n_code_links":1,"syntology":{"n_ran":11,"n_unverified":3,"n_samples":14,"n_pointer_only_licence":0}},{"rank_in_archive_order":23,"model":"td-LSTM-large","metrics":{"Bit per Character (BPC)":"1.49"},"uses_additional_data":false,"paper_date":"2016-02-26","paper":"/paper/architectural-complexity-measures-of","paper_url":"http://arxiv.org/abs/1602.08210v3","paper_title":"Architectural Complexity Measures of Recurrent Neural Networks","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":24,"model":"td-LSTM (Zhang et al., 2016)","metrics":{"Bit per Character (BPC)":"1.63"},"uses_additional_data":false,"paper_date":"2016-02-26","paper":"/paper/architectural-complexity-measures-of","paper_url":"http://arxiv.org/abs/1602.08210v3","paper_title":"Architectural Complexity Measures of Recurrent Neural Networks","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 6,264 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":6264,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":3316},"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":9,"rows_with_any_sample_ran":6,"distinct_papers_with_graph_line":8,"distinct_papers_with_any_sample_ran":5,"samples_over_distinct_papers":{"n_ran":84,"n_unverified":97,"n_samples":181,"n_pointer_only_licence":49,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":89,"n_unverified":97,"n_samples":186,"n_pointer_only_licence":53,"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"}}}