{"url":"/sota/long-range-modeling-on-scrolls","task":{"name":"Long-range modeling","url":"/task/long-range-modeling","note":null},"dataset":{"name":"SCROLLS","url":"/dataset/scrolls"},"category":"Natural Language Processing","categories":["Medical","Miscellaneous","Natural Language Processing"],"category_note":null,"description":"A new task for testing the long-sequence modeling capabilities and efficiency of language models.\r\n\r\nImage credit: [SCROLLS: Standardized CompaRison Over Long Language Sequences](https://arxiv.org/pdf/2201.03533v1.pdf)","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":["Avg.","GovRep","SumScr","QMSum","Qspr","Nrtv","QALT EM-T/H","CNLI"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Avg.":null,"GovRep":null,"SumScr":null,"QMSum":null,"Qspr":null,"Nrtv":null,"QALT EM-T/H":null,"CNLI":null}},"counts":{"rows":13,"rows_with_code":12,"rows_with_paper_page":13,"rows_dated":13,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"CoLT5 XL","metrics":{"Avg.":"43.51","CNLI":"88.4","GovRep":"61.3/32.2/33.8","Nrtv":"31.1","QALT EM-T/H":"48.1/43.8","QMSum":"36.2/12.9/24.3","Qspr":"53.9","SumScr":"36.4/10.2/21.7"},"uses_additional_data":false,"paper_date":"2023-03-17","paper":"/paper/colt5-faster-long-range-transformers-with","paper_url":"https://arxiv.org/abs/2303.09752v3","paper_title":"CoLT5: Faster Long-Range Transformers with Conditional Computation","code":null,"n_code_links":0,"syntology":{"n_ran":0,"n_unverified":9,"n_samples":9,"n_pointer_only_licence":0}},{"rank_in_archive_order":2,"model":"LongT5 XL","metrics":{"Avg.":"42.53","CNLI":"88.2","GovRep":"61.1 / 32.3 / 33.7","Nrtv":"29.3","QALT EM-T/H":"46.0 / 42.1","QMSum":"34.9 / 11.8 / 23.5","Qspr":"53.1","SumScr":"35.8 / 9.6 / 21.1"},"uses_additional_data":false,"paper_date":"2021-12-15","paper":"/paper/longt5-efficient-text-to-text-transformer-for","paper_url":"https://arxiv.org/abs/2112.07916v2","paper_title":"LongT5: Efficient Text-To-Text Transformer for Long Sequences","code":"https://github.com/google-research/longt5","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":3,"model":"LongT5 Large","metrics":{"Avg.":"41.03","CNLI":"87.3","GovRep":"61.3/32.2/33.8","Nrtv":"27.2","QALT EM-T/H":"40.6 / 38.6","QMSum":"35.1 / 12.0 / 23.3","Qspr":"52.3","SumScr":"60.3 / 31.1 / 32.8"},"uses_additional_data":false,"paper_date":"2021-12-15","paper":"/paper/longt5-efficient-text-to-text-transformer-for","paper_url":"https://arxiv.org/abs/2112.07916v2","paper_title":"LongT5: Efficient Text-To-Text Transformer for Long Sequences","code":"https://github.com/google-research/longt5","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":4,"model":"BART-LS","metrics":{"Avg.":"39.76","CNLI":"87.1","GovRep":"59.4 / 29.8 / 30.8","Nrtv":"26.2","QALT EM-T/H":"37.8 / 34.0","QMSum":"35.1 / 11.0 / 22.0","Qspr":"48.7","SumScr":"37.7 / 10.2 / 21.5"},"uses_additional_data":false,"paper_date":"2022-09-21","paper":"/paper/adapting-pretrained-text-to-text-models-for","paper_url":"https://arxiv.org/abs/2209.10052v2","paper_title":"Adapting Pretrained Text-to-Text Models for Long Text Sequences","code":"https://github.com/facebookresearch/bart_ls","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"LongT5 Base","metrics":{"Avg.":"38.6","CNLI":"85.6","GovRep":"57.7 / 30.0 / 31.4","Nrtv":"23.0","QALT EM-T/H":"37.9 / 36.6","QMSum":"33.9 / 11.0 / 22.8","Qspr":"46.6","SumScr":"34.8 / 9.6 / 21.1"},"uses_additional_data":false,"paper_date":"2021-12-15","paper":"/paper/longt5-efficient-text-to-text-transformer-for","paper_url":"https://arxiv.org/abs/2112.07916v2","paper_title":"LongT5: Efficient Text-To-Text Transformer for Long Sequences","code":"https://github.com/google-research/longt5","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":6,"model":"BART-large SLED","metrics":{"Avg.":"37.99","CNLI":"87.3","GovRep":"57.5 / 26.3 / 27.4","Nrtv":"24.1","QALT EM-T/H":"34.8 / 34.8","QMSum":"34.2 / 11.0 / 22.0","Qspr":"46.9","SumScr":"35.2 / 8.7 / 19.4"},"uses_additional_data":false,"paper_date":"2022-08-01","paper":"/paper/efficient-long-text-understanding-with-short","paper_url":"https://arxiv.org/abs/2208.00748v3","paper_title":"Efficient Long-Text Understanding with Short-Text Models","code":"https://github.com/mivg/sled","n_code_links":1,"syntology":null},{"rank_in_archive_order":7,"model":"UL2","metrics":{"Avg.":"37.87","GovRep":"53.6 / 26.1 / 28.8","Nrtv":"24.2","QALT EM-T/H":"45.8 / 40.7","QMSum":"31.1 / 8.5 / 20.4","Qspr":"37.6","SumScr":"32.9 / 7.8 / 19.4"},"uses_additional_data":false,"paper_date":"2022-05-10","paper":"/paper/unifying-language-learning-paradigms","paper_url":"https://arxiv.org/abs/2205.05131v3","paper_title":"UL2: Unifying Language Learning Paradigms","code":"https://github.com/google-research/google-research","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":16,"n_samples":16,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"LED Base","metrics":{"Avg.":"29.16","CNLI":"71.5","GovRep":"56.2 / 26.6 / 28.8","Nrtv":"18.5","QALT EM-T/H":"25.8 / 25.4\t","QMSum":"25.1 / 6.7 / 18.8","Qspr":"26.6","SumScr":"24.2 / 4.5 / 15.4"},"uses_additional_data":false,"paper_date":"2022-01-10","paper":"/paper/scrolls-standardized-comparison-over-long","paper_url":"https://arxiv.org/abs/2201.03533v2","paper_title":"SCROLLS: Standardized CompaRison Over Long Language Sequences","code":"https://github.com/tau-nlp/scrolls","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":6,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":9,"model":"BART Base","metrics":{"Avg.":"29.01","CNLI":"77.4","GovRep":"47.9 / 18.6 / 22.7","Nrtv":"15.4","QALT EM-T/H":"26.0 / 25.9","QMSum":"30.2 / 8.7 / 20.7","Qspr":"26.3","SumScr":"27.2 / 4.9 / 16.7"},"uses_additional_data":false,"paper_date":"2022-01-10","paper":"/paper/scrolls-standardized-comparison-over-long","paper_url":"https://arxiv.org/abs/2201.03533v2","paper_title":"SCROLLS: Standardized CompaRison Over Long Language Sequences","code":"https://github.com/tau-nlp/scrolls","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":6,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"Naive","metrics":{"Avg.":"19.35","CNLI":"66","GovRep":"45.3 / 17.9 / 20.8","Nrtv":"1.5","QALT EM-T/H":"25.2 / 26.1","QMSum":"14.2 / 2.0 / 9.3","Qspr":"3.4","SumScr":"19.6 / 1.8 / 11.0"},"uses_additional_data":false,"paper_date":"2022-01-10","paper":"/paper/scrolls-standardized-comparison-over-long","paper_url":"https://arxiv.org/abs/2201.03533v2","paper_title":"SCROLLS: Standardized CompaRison Over Long Language Sequences","code":"https://github.com/tau-nlp/scrolls","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":6,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"PEGASUS-X","metrics":{"GovRep":"60.3 / 30.0 / 31.5","QMSum":"33.2 / 9.6 / 21.6 ","SumScr":"35.7 / 9.1 / 20.6"},"uses_additional_data":false,"paper_date":"2022-08-08","paper":"/paper/investigating-efficiently-extending","paper_url":"https://arxiv.org/abs/2208.04347v1","paper_title":"Investigating Efficiently Extending Transformers for Long Input Summarization","code":"https://github.com/google-research/pegasus","n_code_links":2,"syntology":null},{"rank_in_archive_order":12,"model":"PEGASUS-X-Base","metrics":{"GovRep":"59.3 / 29.3 / 30.9","QMSum":"32.9 / 9.8 / 21.4","SumScr":" 35.0 / 8.9 / 20.4"},"uses_additional_data":false,"paper_date":"2022-08-08","paper":"/paper/investigating-efficiently-extending","paper_url":"https://arxiv.org/abs/2208.04347v1","paper_title":"Investigating Efficiently Extending Transformers for Long Input Summarization","code":"https://github.com/google-research/pegasus","n_code_links":2,"syntology":null},{"rank_in_archive_order":13,"model":"UL2 20B","metrics":{"CNLI":"88.7"},"uses_additional_data":false,"paper_date":"2022-05-10","paper":"/paper/unifying-language-learning-paradigms","paper_url":"https://arxiv.org/abs/2205.05131v3","paper_title":"UL2: Unifying Language Learning Paradigms","code":"https://github.com/google-research/google-research","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":16,"n_samples":16,"n_pointer_only_licence":0}}],"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":4,"distinct_papers_with_any_sample_ran":2,"samples_over_distinct_papers":{"n_ran":3,"n_unverified":31,"n_samples":34,"n_pointer_only_licence":1,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":9,"n_unverified":59,"n_samples":68,"n_pointer_only_licence":3,"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"}}}