{"url":"/sota/language-modelling-on-the-pile","task":{"name":"Language Modelling","url":"/task/language-modelling","note":null},"dataset":{"name":"The Pile","url":"/dataset/the-pile"},"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 They have superseded recurrent neural network-based models, which had previously superseded the purely statistical models, such as word n-gram language model.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":["Bits per byte","Test perplexity"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Bits per byte":null,"Test perplexity":"lower"}},"counts":{"rows":39,"rows_with_code":35,"rows_with_paper_page":39,"rows_dated":39,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"Test-Time Fine-Tuning with SIFT + Llama-3.2 (3B)","metrics":{"Bits per byte":"0.557"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":2,"model":"Test-Time Fine-Tuning with SIFT + Phi-3 (3.8B)","metrics":{"Bits per byte":"0.595"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":3,"model":"Test-Time Fine-Tuning with SIFT + Llama-3.2 (1B)","metrics":{"Bits per byte":"0.606"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":4,"model":"Gemma-2 27B","metrics":{"Bits per byte":"0.629"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"GLM-130B","metrics":{"Bits per byte":"0.634"},"uses_additional_data":false,"paper_date":"2022-10-05","paper":"/paper/glm-130b-an-open-bilingual-pre-trained-model","paper_url":"https://arxiv.org/abs/2210.02414v2","paper_title":"GLM-130B: An Open Bilingual Pre-trained Model","code":"https://github.com/thudm/chatglm2-6b","n_code_links":9,"syntology":{"n_ran":13,"n_unverified":8,"n_samples":21,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"Llama-3.2 3B","metrics":{"Bits per byte":"0.640"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":7,"model":"Jurassic-1","metrics":{"Bits per byte":"0.65"},"uses_additional_data":false,"paper_date":"2022-10-05","paper":"/paper/glm-130b-an-open-bilingual-pre-trained-model","paper_url":"https://arxiv.org/abs/2210.02414v2","paper_title":"GLM-130B: An Open Bilingual Pre-trained Model","code":"https://github.com/thudm/chatglm2-6b","n_code_links":9,"syntology":{"n_ran":13,"n_unverified":8,"n_samples":21,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"Phi-3 14B","metrics":{"Bits per byte":"0.651"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":9,"model":"Gemma-2 9B","metrics":{"Bits per byte":"0.670"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"Phi-3 7B","metrics":{"Bits per byte":"0.678"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"Phi-3 3.8B","metrics":{"Bits per byte":"0.679"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":12,"model":"Llama-3.2 1B","metrics":{"Bits per byte":"0.697"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":13,"model":"GPT-3 Davinci 175B (pre-trained)","metrics":{"Bits per byte":"0.7177"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/the-pile-an-800gb-dataset-of-diverse-text-for","paper_url":"https://arxiv.org/abs/2101.00027v1","paper_title":"The Pile: An 800GB Dataset of Diverse Text for Language Modeling","code":"https://github.com/EleutherAI/gpt-neo","n_code_links":22,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"Gemma-2 2B","metrics":{"Bits per byte":"0.721"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":15,"model":"Llama-3.2-Instruct 3B","metrics":{"Bits per byte":"0.737"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"GPT-3","metrics":{"Bits per byte":"0.742"},"uses_additional_data":false,"paper_date":"2022-10-05","paper":"/paper/glm-130b-an-open-bilingual-pre-trained-model","paper_url":"https://arxiv.org/abs/2210.02414v2","paper_title":"GLM-130B: An Open Bilingual Pre-trained Model","code":"https://github.com/thudm/chatglm2-6b","n_code_links":9,"syntology":{"n_ran":13,"n_unverified":8,"n_samples":21,"n_pointer_only_licence":0}},{"rank_in_archive_order":17,"model":"Test-Time Fine-Tuning with SIFT + GPT-2 (774M)","metrics":{"Bits per byte":"0.762"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":18,"model":"GPT-3 Curie 6.7B (pre-trained)","metrics":{"Bits per byte":"0.7980"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/the-pile-an-800gb-dataset-of-diverse-text-for","paper_url":"https://arxiv.org/abs/2101.00027v1","paper_title":"The Pile: An 800GB Dataset of Diverse Text for Language Modeling","code":"https://github.com/EleutherAI/gpt-neo","n_code_links":22,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":19,"model":"Llama-3.2-Instruct 1B","metrics":{"Bits per byte":"0.807"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":20,"model":"GPT-2 Large 774M (test-time training on nearest neighbors)","metrics":{"Bits per byte":"0.85"},"uses_additional_data":false,"paper_date":"2023-05-29","paper":"/paper/test-time-training-on-nearest-neighbors-for","paper_url":"https://arxiv.org/abs/2305.18466v3","paper_title":"Test-Time Training on Nearest Neighbors for Large Language Models","code":"https://github.com/socialfoundations/tttlm","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":5,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":21,"model":"Test-Time Fine-Tuning with SIFT + GPT-2 (124M)","metrics":{"Bits per byte":"0.862"},"uses_additional_data":false,"paper_date":"2024-10-10","paper":"/paper/efficiently-learning-at-test-time-active-fine","paper_url":"https://arxiv.org/abs/2410.08020v2","paper_title":"Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs","code":"https://github.com/jonhue/activeft","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":22,"model":"GPT-3 Babbage 1.3B (pre-trained)","metrics":{"Bits per byte":"0.8718"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/the-pile-an-800gb-dataset-of-diverse-text-for","paper_url":"https://arxiv.org/abs/2101.00027v1","paper_title":"The Pile: An 800GB Dataset of Diverse Text for Language Modeling","code":"https://github.com/EleutherAI/gpt-neo","n_code_links":22,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":23,"model":"GPT-3 Ada 350M (pre-trained)","metrics":{"Bits per byte":"0.9631"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/the-pile-an-800gb-dataset-of-diverse-text-for","paper_url":"https://arxiv.org/abs/2101.00027v1","paper_title":"The Pile: An 800GB Dataset of Diverse Text for Language Modeling","code":"https://github.com/EleutherAI/gpt-neo","n_code_links":22,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":24,"model":"GPT-2 XL 1.5B (pre-trained)","metrics":{"Bits per byte":"1.0468"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/the-pile-an-800gb-dataset-of-diverse-text-for","paper_url":"https://arxiv.org/abs/2101.00027v1","paper_title":"The Pile: An 800GB Dataset of Diverse Text for Language Modeling","code":"https://github.com/EleutherAI/gpt-neo","n_code_links":22,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":25,"model":"GPT-2 Large 774M (pre-trained)","metrics":{"Bits per byte":"1.0828"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/the-pile-an-800gb-dataset-of-diverse-text-for","paper_url":"https://arxiv.org/abs/2101.00027v1","paper_title":"The Pile: An 800GB Dataset of Diverse Text for Language Modeling","code":"https://github.com/EleutherAI/gpt-neo","n_code_links":22,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":26,"model":"GPT-2 Medium 355M (pre-trained)","metrics":{"Bits per byte":"1.0928"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/the-pile-an-800gb-dataset-of-diverse-text-for","paper_url":"https://arxiv.org/abs/2101.00027v1","paper_title":"The Pile: An 800GB Dataset of Diverse Text for Language Modeling","code":"https://github.com/EleutherAI/gpt-neo","n_code_links":22,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":27,"model":"GPT-2 Small 124M (pre-trained)","metrics":{"Bits per byte":"1.2253"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/the-pile-an-800gb-dataset-of-diverse-text-for","paper_url":"https://arxiv.org/abs/2101.00027v1","paper_title":"The Pile: An 800GB Dataset of Diverse Text for Language Modeling","code":"https://github.com/EleutherAI/gpt-neo","n_code_links":22,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":28,"model":"Larger Transformer 771M (fine-tuned)","metrics":{"Test perplexity":"10"},"uses_additional_data":false,"paper_date":"2024-02-02","paper":"/paper/specialized-language-models-with-cheap","paper_url":"https://arxiv.org/abs/2402.01093v2","paper_title":"Need a Small Specialized Language Model? Plan Early!","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":29,"model":"Hybrid H3 125M","metrics":{"Test perplexity":"10.2"},"uses_additional_data":false,"paper_date":"2022-12-28","paper":"/paper/hungry-hungry-hippos-towards-language","paper_url":"https://arxiv.org/abs/2212.14052v3","paper_title":"Hungry Hungry Hippos: Towards Language Modeling with State Space Models","code":"https://github.com/hazyresearch/safari","n_code_links":3,"syntology":{"n_ran":7,"n_unverified":8,"n_samples":15,"n_pointer_only_licence":0}},{"rank_in_archive_order":30,"model":"GPT-Neo 2.7B","metrics":{"Test perplexity":"10.44"},"uses_additional_data":false,"paper_date":"2022-10-04","paper":"/paper/knowledge-unlearning-for-mitigating-privacy","paper_url":"https://arxiv.org/abs/2210.01504v2","paper_title":"Knowledge Unlearning for Mitigating Privacy Risks in Language Models","code":"https://github.com/joeljang/knowledge-unlearning","n_code_links":2,"syntology":null},{"rank_in_archive_order":31,"model":"Transformer 125M","metrics":{"Test perplexity":"10.7"},"uses_additional_data":false,"paper_date":"2022-12-28","paper":"/paper/hungry-hungry-hippos-towards-language","paper_url":"https://arxiv.org/abs/2212.14052v3","paper_title":"Hungry Hungry Hippos: Towards Language Modeling with State Space Models","code":"https://github.com/hazyresearch/safari","n_code_links":3,"syntology":{"n_ran":7,"n_unverified":8,"n_samples":15,"n_pointer_only_licence":0}},{"rank_in_archive_order":32,"model":"GPT-Neo 1.3B","metrics":{"Test perplexity":"11.46"},"uses_additional_data":false,"paper_date":"2022-10-04","paper":"/paper/knowledge-unlearning-for-mitigating-privacy","paper_url":"https://arxiv.org/abs/2210.01504v2","paper_title":"Knowledge Unlearning for Mitigating Privacy Risks in Language Models","code":"https://github.com/joeljang/knowledge-unlearning","n_code_links":2,"syntology":null},{"rank_in_archive_order":33,"model":"Smaller Transformer 126M (fine-tuned)","metrics":{"Test perplexity":"12"},"uses_additional_data":false,"paper_date":"2024-02-02","paper":"/paper/specialized-language-models-with-cheap","paper_url":"https://arxiv.org/abs/2402.01093v2","paper_title":"Need a Small Specialized Language Model? Plan Early!","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":34,"model":"OPT 2.7B","metrics":{"Test perplexity":"17.81"},"uses_additional_data":false,"paper_date":"2022-10-04","paper":"/paper/knowledge-unlearning-for-mitigating-privacy","paper_url":"https://arxiv.org/abs/2210.01504v2","paper_title":"Knowledge Unlearning for Mitigating Privacy Risks in Language Models","code":"https://github.com/joeljang/knowledge-unlearning","n_code_links":2,"syntology":null},{"rank_in_archive_order":35,"model":"GPT-Neo 125M","metrics":{"Test perplexity":"17.83"},"uses_additional_data":false,"paper_date":"2022-10-04","paper":"/paper/knowledge-unlearning-for-mitigating-privacy","paper_url":"https://arxiv.org/abs/2210.01504v2","paper_title":"Knowledge Unlearning for Mitigating Privacy Risks in Language Models","code":"https://github.com/joeljang/knowledge-unlearning","n_code_links":2,"syntology":null},{"rank_in_archive_order":36,"model":"OPT 1.3B","metrics":{"Test perplexity":"19.55"},"uses_additional_data":false,"paper_date":"2022-10-04","paper":"/paper/knowledge-unlearning-for-mitigating-privacy","paper_url":"https://arxiv.org/abs/2210.01504v2","paper_title":"Knowledge Unlearning for Mitigating Privacy Risks in Language Models","code":"https://github.com/joeljang/knowledge-unlearning","n_code_links":2,"syntology":null},{"rank_in_archive_order":37,"model":"Larger Transformer 771M (pre-trained)","metrics":{"Test perplexity":"28.1"},"uses_additional_data":false,"paper_date":"2024-02-02","paper":"/paper/specialized-language-models-with-cheap","paper_url":"https://arxiv.org/abs/2402.01093v2","paper_title":"Need a Small Specialized Language Model? Plan Early!","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":38,"model":"OPT 125M","metrics":{"Test perplexity":"32.26"},"uses_additional_data":false,"paper_date":"2022-10-04","paper":"/paper/knowledge-unlearning-for-mitigating-privacy","paper_url":"https://arxiv.org/abs/2210.01504v2","paper_title":"Knowledge Unlearning for Mitigating Privacy Risks in Language Models","code":"https://github.com/joeljang/knowledge-unlearning","n_code_links":2,"syntology":null},{"rank_in_archive_order":39,"model":"Smaller Transformer 126M (pre-trained)","metrics":{"Test perplexity":"33"},"uses_additional_data":false,"paper_date":"2024-02-02","paper":"/paper/specialized-language-models-with-cheap","paper_url":"https://arxiv.org/abs/2402.01093v2","paper_title":"Need a Small Specialized Language Model? Plan Early!","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 6,885 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":29,"rows_with_any_sample_ran":28,"distinct_papers_with_graph_line":5,"distinct_papers_with_any_sample_ran":4,"samples_over_distinct_papers":{"n_ran":23,"n_unverified":21,"n_samples":44,"n_pointer_only_licence":0,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":91,"n_unverified":45,"n_samples":136,"n_pointer_only_licence":0,"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"}}}