{"url":"/dataset/drop","name":"DROP","full_name":"Discrete Reasoning Over Paragraphs","description_markdown":"**Discrete Reasoning Over Paragraphs** **DROP** is a crowdsourced, adversarially-created, 96k-question benchmark, in which a system must resolve references in a question, perhaps to multiple input positions, and perform discrete operations over them (such as addition, counting, or sorting). These operations require a much more comprehensive understanding of the content of paragraphs than what was necessary for prior datasets. The questions consist of passages extracted from Wikipedia articles. The dataset is split into a training set of about 77,000 questions, a development set of around 9,500 questions and a hidden test set similar in size to the development set.\r\n\r\nSource: [https://allennlp.org/drop](https://allennlp.org/drop)\r\nImage Source: [DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs](https://paperswithcode.com/paper/drop-a-reading-comprehension-benchmark/)","description_withheld":null,"homepage":"https://allennlp.org/drop","introduced_date":"2019-01-01","introduced_date_note":null,"introduced_by":{"paper":"/paper/drop-a-reading-comprehension-benchmark","title":"DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs","first_author":"Dheeru Dua","url":null},"license":{"name":"CC BY-SA 4.0","url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode"},"modalities":[{"name":"Texts","url":"/datasets/modality/texts"}],"tasks":[{"name":"Question Answering","url":"/task/question-answering","datasets_with_task":"/datasets/task/question-answering"},{"name":"Text Generation","url":"/task/text-generation","datasets_with_task":"/datasets/task/text-generation"}],"languages":[{"name":"English","url":"/datasets/language/english"}],"variants":["DROP Test","DROP","Drop (3-Shot)"],"data_loaders":[{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/drop","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/swap-uniba/drop_ita","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/ucinlp/drop","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/tensorflow/datasets","url":"https://www.tensorflow.org/datasets/catalog/drop","frameworks":["tf","jax"]},{"repo":"https://github.com/RUCAIBox/LLMBox","url":"https://github.com/RUCAIBox/LLMBox/blob/main/docs/utilization/supported-datasets.md","frameworks":["pytorch"]},{"repo":"https://github.com/allenai/allennlp-models","url":"https://docs.allennlp.org/models/main/models/rc/dataset_readers/drop/","frameworks":["pytorch"]}],"num_papers_in_archive":382,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[{"leaderboard":"/sota/question-answering-on-drop-test","task":"Question Answering","dataset_variant":"DROP Test","rows":16,"metrics":["F1"],"first_row_in_archive_order":{"model":"QDGAT (ensemble)","paper":"/paper/question-directed-graph-attention-network-for","metrics":{"F1":"88.38"},"code_links":[]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"},{"leaderboard":"/sota/question-answering-on-drop","task":"Question Answering","dataset_variant":"DROP","rows":6,"metrics":["Accuracy"],"first_row_in_archive_order":{"model":"PaLM 540B (Self Improvement, Self Consistency)","paper":"/paper/large-language-models-can-self-improve","metrics":{"Accuracy":"83"},"code_links":[]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"},{"leaderboard":"/sota/text-generation-on-drop-3-shot","task":"Text Generation","dataset_variant":"Drop (3-Shot)","rows":0,"metrics":["f1 score"],"first_row_in_archive_order":null,"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"}],"papers_with_a_benchmark_row":[{"paper":"/paper/orca-2-teaching-small-language-models-how-to","title":"Orca 2: Teaching Small Language Models How to Reason","date":"2023-11-18","rows_on_this_dataset":2,"code_links":0,"syntology":null},{"paper":"/paper/palm-2-technical-report-1","title":"PaLM 2 Technical Report","date":"2023-05-17","rows_on_this_dataset":1,"code_links":1,"syntology":null},{"paper":"/paper/gpt-4-technical-report-1","title":"GPT-4 Technical Report","date":"2023-03-15","rows_on_this_dataset":2,"code_links":11,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":5,"samples_ran":2,"samples_unverified":3,"pointer_only_for_licence":1,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/large-language-models-can-self-improve","title":"Large Language Models Can Self-Improve","date":"2022-10-20","rows_on_this_dataset":6,"code_links":0,"syntology":null},{"paper":"/paper/reasoning-like-program-executors-1","title":"Reasoning Like Program Executors","date":"2022-01-27","rows_on_this_dataset":1,"code_links":1,"syntology":null},{"paper":"/paper/question-directed-graph-attention-network-for","title":"Question Directed Graph Attention Network for Numerical Reasoning over Text","date":"2020-09-16","rows_on_this_dataset":1,"code_links":0,"syntology":null},{"paper":"/paper/language-models-are-few-shot-learners","title":"Language Models are Few-Shot Learners","date":"2020-05-28","rows_on_this_dataset":1,"code_links":67,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":65,"samples_ran":15,"samples_unverified":50,"pointer_only_for_licence":4,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/neural-symbolic-reader-scalable-integration","title":"Neural Symbolic Reader: Scalable Integration of Distributed and Symbolic Representations for Reading Comprehension","date":"2020-05-01","rows_on_this_dataset":1,"code_links":0,"syntology":null},{"paper":"/paper/injecting-numerical-reasoning-skills-into","title":"Injecting Numerical Reasoning Skills into Language Models","date":"2020-04-09","rows_on_this_dataset":1,"code_links":2,"syntology":null},{"paper":"/paper/numnet-machine-reading-comprehension-with","title":"NumNet: Machine Reading Comprehension with Numerical Reasoning","date":"2019-10-15","rows_on_this_dataset":1,"code_links":2,"syntology":null},{"paper":"/paper/tag-based-multi-span-extraction-in-reading","title":"A Simple and Effective Model for Answering Multi-span Questions","date":"2019-09-29","rows_on_this_dataset":1,"code_links":4,"syntology":null},{"paper":"/paper/giving-bert-a-calculator-finding-operations","title":"Giving BERT a Calculator: Finding Operations and Arguments with Reading Comprehension","date":"2019-08-31","rows_on_this_dataset":1,"code_links":0,"syntology":null},{"paper":"/paper/a-multi-type-multi-span-network-for-reading","title":"A Multi-Type Multi-Span Network for Reading Comprehension that Requires Discrete Reasoning","date":"2019-08-15","rows_on_this_dataset":1,"code_links":1,"syntology":null},{"paper":"/paper/drop-a-reading-comprehension-benchmark","title":"DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs","date":"2019-03-01","rows_on_this_dataset":2,"code_links":3,"syntology":null}],"syntology_totals":{"read_at":"2026-09-24T18:15:14+00:00","papers_with_samples":2,"samples_harvested":70,"samples_ran":17,"samples_unverified":53,"pointer_only_for_licence":5,"papers_with_no_sample_that_ran":0,"note":"the per-paper counts above, summed; not a rate"},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks."}