five

sentence-transformers/msmarco-co-condenser-margin-mse-sym-mnrl-mean-v1

收藏
Hugging Face2024-05-15 更新2024-06-12 收录
下载链接：
https://hf-mirror.com/datasets/sentence-transformers/msmarco-co-condenser-margin-mse-sym-mnrl-mean-v1
下载链接
链接失效反馈
官方服务：
资源简介：
--- language: - en multilinguality: - monolingual size_categories: - 10M<n<100M task_categories: - feature-extraction - sentence-similarity pretty_name: MS MARCO with hard negatives from co-condenser-margin-mse-sym-mnrl-mean-v1 tags: - sentence-transformers dataset_info: - config_name: triplet features: - name: query dtype: string - name: positive dtype: string - name: negative dtype: string splits: - name: train num_bytes: 362903736 num_examples: 502939 download_size: 237820181 dataset_size: 362903736 - config_name: triplet-50 features: - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string - name: negative_9 dtype: string - name: negative_10 dtype: string - name: negative_11 dtype: string - name: negative_12 dtype: string - name: negative_13 dtype: string - name: negative_14 dtype: string - name: negative_15 dtype: string - name: negative_16 dtype: string - name: negative_17 dtype: string - name: negative_18 dtype: string - name: negative_19 dtype: string - name: negative_20 dtype: string - name: negative_21 dtype: string - name: negative_22 dtype: string - name: negative_23 dtype: string - name: negative_24 dtype: string - name: negative_25 dtype: string - name: negative_26 dtype: string - name: negative_27 dtype: string - name: negative_28 dtype: string - name: negative_29 dtype: string - name: negative_30 dtype: string - name: negative_31 dtype: string - name: negative_32 dtype: string - name: negative_33 dtype: string - name: negative_34 dtype: string - name: negative_35 dtype: string - name: negative_36 dtype: string - name: negative_37 dtype: string - name: negative_38 dtype: string - name: negative_39 dtype: string - name: negative_40 dtype: string - name: negative_41 dtype: string - name: negative_42 dtype: string - name: negative_43 dtype: string - name: negative_44 dtype: string - name: negative_45 dtype: string - name: negative_46 dtype: string - name: negative_47 dtype: string - name: negative_48 dtype: string - name: negative_49 dtype: string - name: negative_50 dtype: string - name: query dtype: string - name: positive dtype: string splits: - name: train num_bytes: 9055260814 num_examples: 502939 download_size: 5914557509 dataset_size: 9055260814 - config_name: triplet-50-ids features: - name: negative_1 dtype: int64 - name: negative_2 dtype: int64 - name: negative_3 dtype: int64 - name: negative_4 dtype: int64 - name: negative_5 dtype: int64 - name: negative_6 dtype: int64 - name: negative_7 dtype: int64 - name: negative_8 dtype: int64 - name: negative_9 dtype: int64 - name: negative_10 dtype: int64 - name: negative_11 dtype: int64 - name: negative_12 dtype: int64 - name: negative_13 dtype: int64 - name: negative_14 dtype: int64 - name: negative_15 dtype: int64 - name: negative_16 dtype: int64 - name: negative_17 dtype: int64 - name: negative_18 dtype: int64 - name: negative_19 dtype: int64 - name: negative_20 dtype: int64 - name: negative_21 dtype: int64 - name: negative_22 dtype: int64 - name: negative_23 dtype: int64 - name: negative_24 dtype: int64 - name: negative_25 dtype: int64 - name: negative_26 dtype: int64 - name: negative_27 dtype: int64 - name: negative_28 dtype: int64 - name: negative_29 dtype: int64 - name: negative_30 dtype: int64 - name: negative_31 dtype: int64 - name: negative_32 dtype: int64 - name: negative_33 dtype: int64 - name: negative_34 dtype: int64 - name: negative_35 dtype: int64 - name: negative_36 dtype: int64 - name: negative_37 dtype: int64 - name: negative_38 dtype: int64 - name: negative_39 dtype: int64 - name: negative_40 dtype: int64 - name: negative_41 dtype: int64 - name: negative_42 dtype: int64 - name: negative_43 dtype: int64 - name: negative_44 dtype: int64 - name: negative_45 dtype: int64 - name: negative_46 dtype: int64 - name: negative_47 dtype: int64 - name: negative_48 dtype: int64 - name: negative_49 dtype: int64 - name: negative_50 dtype: int64 - name: query dtype: int64 - name: positive dtype: int64 splits: - name: train num_bytes: 209222624 num_examples: 502939 download_size: 178195399 dataset_size: 209222624 - config_name: triplet-all features: - name: query dtype: string - name: positive dtype: string - name: negative dtype: string splits: - name: train num_bytes: 19858927973 num_examples: 26637550 download_size: 4244209962 dataset_size: 19858927973 - config_name: triplet-all-ids features: - name: query dtype: int64 - name: positive dtype: int64 - name: negative dtype: int64 splits: - name: train num_bytes: 639301200 num_examples: 26637550 download_size: 190130784 dataset_size: 639301200 - config_name: triplet-hard features: - name: query dtype: string - name: positive dtype: string - name: negative dtype: string splits: - name: train num_bytes: 8477376947 num_examples: 11662655 download_size: 2152733400 dataset_size: 8477376947 - config_name: triplet-hard-ids features: - name: query dtype: int64 - name: positive dtype: int64 - name: negative dtype: int64 splits: - name: train num_bytes: 279903720 num_examples: 11662655 download_size: 89399265 dataset_size: 279903720 - config_name: triplet-ids features: - name: query dtype: int64 - name: positive dtype: int64 - name: negative dtype: int64 splits: - name: train num_bytes: 12070536 num_examples: 502939 download_size: 10132130 dataset_size: 12070536 configs: - config_name: triplet data_files: - split: train path: triplet/train-* - config_name: triplet-50 data_files: - split: train path: triplet-50/train-* - config_name: triplet-50-ids data_files: - split: train path: triplet-50-ids/train-* - config_name: triplet-all data_files: - split: train path: triplet-all/train-* - config_name: triplet-all-ids data_files: - split: train path: triplet-all-ids/train-* - config_name: triplet-hard data_files: - split: train path: triplet-hard/train-* - config_name: triplet-hard-ids data_files: - split: train path: triplet-hard-ids/train-* - config_name: triplet-ids data_files: - split: train path: triplet-ids/train-* --- # MS MARCO with hard negatives from co-condenser-margin-mse-sym-mnrl-mean-v1 [MS MARCO](https://microsoft.github.io/msmarco/) is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine. For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train [Sentence Transformer models](https://www.sbert.net). ## Related Datasets These are the datasets generated using the 13 different models: * [msmarco-bm25](https://huggingface.co/datasets/sentence-transformers/msmarco-bm25) * [msmarco-msmarco-distilbert-base-tas-b](https://huggingface.co/datasets/sentence-transformers/msmarco-msmarco-distilbert-base-tas-b) * [msmarco-msmarco-distilbert-base-v3](https://huggingface.co/datasets/sentence-transformers/msmarco-msmarco-distilbert-base-v3) * [msmarco-msmarco-MiniLM-L-6-v3](https://huggingface.co/datasets/sentence-transformers/msmarco-msmarco-MiniLM-L-6-v3) * [msmarco-distilbert-margin-mse-cls-dot-v2](https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-cls-dot-v2) * [msmarco-distilbert-margin-mse-cls-dot-v1](https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-cls-dot-v1) * [msmarco-distilbert-margin-mse-mean-dot-v1](https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-mean-dot-v1) * [msmarco-mpnet-margin-mse-mean-v1](https://huggingface.co/datasets/sentence-transformers/msmarco-mpnet-margin-mse-mean-v1) * [msmarco-co-condenser-margin-mse-cls-v1](https://huggingface.co/datasets/sentence-transformers/msmarco-co-condenser-margin-mse-cls-v1) * [msmarco-distilbert-margin-mse-mnrl-mean-v1](https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-mnrl-mean-v1) * [msmarco-distilbert-margin-mse-sym-mnrl-mean-v1](https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-sym-mnrl-mean-v1) * [msmarco-distilbert-margin-mse-sym-mnrl-mean-v2](https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-sym-mnrl-mean-v2) * [msmarco-co-condenser-margin-mse-sym-mnrl-mean-v1](https://huggingface.co/datasets/sentence-transformers/msmarco-co-condenser-margin-mse-sym-mnrl-mean-v1) ## Dataset Subsets ### Unique Triplets For each query-positive pair, mine the passage most similar to the query and consider it as a negative. We release two subsets, one with strings (`triplet`) and one with IDs (`triplet-ids`) to be used with [sentence-transformers/msmarco-corpus](https://huggingface.co/datasets/sentence-transformers/msmarco-corpus). #### `triplet` subset * Columns: "query", "positive", "negative" * Column types: `str`, `str`, `str` * Examples: ```python { "query": "what are the liberal arts?", "positive": 'liberal arts. 1. the academic course of instruction at a college intended to provide general knowledge and comprising the arts, humanities, natural sciences, and social sciences, as opposed to professional or technical subjects.', "negative": 'The New York State Education Department requires 60 Liberal Arts credits in a Bachelor of Science program and 90 Liberal Arts credits in a Bachelor of Arts program. In the list of course descriptions, courses which are liberal arts for all students are identified by (Liberal Arts) after the course number.' } ``` * Deduplified: No #### `triplet-ids` subset * Columns: "query", "positive", "negative" * Column types: `int`, `int`, `int` * Examples: ```python { "query": 571018, "positive": 7349777, "negative": 6948601 } ``` * Deduplified: No ### All Triplets For each query-positive pair, mine the 50 most similar passages to the query and consider them as negatives, resulting in 50 triplets for each query-positive pair. We release two subsets, one with strings (`triplet-all`) and one with IDs (`triplet-all-ids`) to be used with [sentence-transformers/msmarco-corpus](https://huggingface.co/datasets/sentence-transformers/msmarco-corpus). #### `triplet-all` subset * Columns: "query", "positive", "negative" * Column types: `str`, `str`, `str` * Examples: ```python { "query": "what are the liberal arts?", "positive": 'liberal arts. 1. the academic course of instruction at a college intended to provide general knowledge and comprising the arts, humanities, natural sciences, and social sciences, as opposed to professional or technical subjects.', "negative": 'The New York State Education Department requires 60 Liberal Arts credits in a Bachelor of Science program and 90 Liberal Arts credits in a Bachelor of Arts program. In the list of course descriptions, courses which are liberal arts for all students are identified by (Liberal Arts) after the course number.' } ``` * Deduplified: No #### `triplet-all-ids` subset * Columns: "query", "positive", "negative" * Column types: `int`, `int`, `int` * Examples: ```python { "query": 571018, "positive": 7349777, "negative": 6948601 } ``` * Deduplified: No ### Hard Triplets For each query-positive pair, mine the 50 most similar passages to the query and consider them as negatives. Filter these 50 negatives such that `similarity(query, positive) > similarity(query, negative) + margin`, with [cross-encoder/ms-marco-MiniLM-L-6-v2](https://huggingface.co/cross-encoder/ms-marco-MiniLM-L-6-v2) and `margin = 3.0`. In short, we rely on a CrossEncoder to try and make sure that the negatives are indeed dissimilar to the query. We release two subsets, one with strings (`triplet-hard`) and one with IDs (`triplet-hard-ids`) to be used with [sentence-transformers/msmarco-corpus](https://huggingface.co/datasets/sentence-transformers/msmarco-corpus). #### `triplet-hard` subset * Columns: "query", "positive", "negative" * Column types: `str`, `str`, `str` * Examples: ```python { "query": "what are the liberal arts?", "positive": 'liberal arts. 1. the academic course of instruction at a college intended to provide general knowledge and comprising the arts, humanities, natural sciences, and social sciences, as opposed to professional or technical subjects.', "negative": 'The New York State Education Department requires 60 Liberal Arts credits in a Bachelor of Science program and 90 Liberal Arts credits in a Bachelor of Arts program. In the list of course descriptions, courses which are liberal arts for all students are identified by (Liberal Arts) after the course number.' } ``` * Deduplified: No #### `triplet-hard-ids` subset * Columns: "query", "positive", "negative" * Column types: `int`, `int`, `int` * Examples: ```python { "query": 571018, "positive": 7349777, "negative": 6948601 } ``` * Deduplified: No ### 50 "Triplets" For each query-positive pair, mine the 50 most similar passages to the query and consider them as negatives. Rather than storing this data as 50 triplets, we store it all as one sample with 50 negative columns. We release two subsets, one with strings (`triplet-50`) and one with IDs (`triplet-50-ids`) to be used with [sentence-transformers/msmarco-corpus](https://huggingface.co/datasets/sentence-transformers/msmarco-corpus). #### `triplet-50` subset * Columns: "query", "positive", 'negative_1', 'negative_2', 'negative_3', 'negative_4', 'negative_5', 'negative_6', 'negative_7', 'negative_8', 'negative_9', 'negative_10', 'negative_11', 'negative_12', 'negative_13', 'negative_14', 'negative_15', 'negative_16', 'negative_17', 'negative_18', 'negative_19', 'negative_20', 'negative_21', 'negative_22', 'negative_23', 'negative_24', 'negative_25', 'negative_26', 'negative_27', 'negative_28', 'negative_29', 'negative_30', 'negative_31', 'negative_32', 'negative_33', 'negative_34', 'negative_35', 'negative_36', 'negative_37', 'negative_38', 'negative_39', 'negative_40', 'negative_41', 'negative_42', 'negative_43', 'negative_44', 'negative_45', 'negative_46', 'negative_47', 'negative_48', 'negative_49', 'negative_50' * Column types: `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str`, `str` * Examples: ```python { "query": "what are the liberal arts?", "positive": "liberal arts. 1. the academic course of instruction at a college intended to provide general knowledge and comprising the arts, humanities, natural sciences, and social sciences, as opposed to professional or technical subjects.", "negative_1": "The New York State Education Department requires 60 Liberal Arts credits in a Bachelor of Science program and 90 Liberal Arts credits in a Bachelor of Arts program. In the list of course descriptions, courses which are liberal arts for all students are identified by (Liberal Arts) after the course number.", "negative_2": "What Does it Mean to Study Liberal Arts? A liberal arts major offers a broad overview of the arts, sciences, and humanities. Within the context of a liberal arts degree, you can study modern languages, music, English, anthropology, history, women's studies, psychology, math, political science or many other disciplines.", "negative_3": "What Is Liberal Studies? Liberal studies, also known as liberal arts, comprises a broad exploration of social sciences, natural sciences, humanities, and the arts. If you are interested in a wide-ranging education in humanities, communication, and thinking, read on to find out about the educational and career possibilities in liberal studies.", "negative_4": "You can choose from an array of liberal arts majors. Most of these are offered in the liberal arts departments of colleges that belong to universities and at smaller colleges that are designated as liberal arts institutions.", "negative_5": "Majors. You can choose from an array of liberal arts majors. Most of these are offered in the liberal arts departments of colleges that belong to universities and at smaller colleges that are designated as liberal arts institutions.", "negative_6": "liberal arts. plural noun. Definition of liberal arts for English Language Learners. : areas of study (such as history, language, and literature) that are intended to give you general knowledge rather than to develop specific skills needed for a profession. Nglish: Translation of liberal arts for Spanish speakers Britannica.com: Encyclopedia article about liberal arts.", "negative_7": "Because they award less than 50% of their degrees in engineering, and the rest in liberal arts (sciences). Baccalaureate colleges are a type of Liberal Arts colleges, But offering lesser number of degrees compared to LAC. It's the other way round. A liberal arts college focuses on liberal arts, e.g. sciences, literature, history, sociology, etc. They might offer a few professional degrees (most frequently engineering) as well, but typically the professional majors are well integrated into the liberal arts framework as well.", "negative_8": "A liberal arts college is a four-year institution that focuses on the study of liberal arts. Liberal arts colleges are geared more toward the acquisition of knowledge and less toward specific professions. [MORE: The Path to Higher Education] Graduate school.", "negative_9": "1 BA = Bachelor of Arts degree BS = Bachelor of Science degree. 2 I think the question requires more of an explanation than what the terms BA and BS translate to. 3 B.A. (Bachelor of Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree. I think the question requires more of an explanation than what the terms BA and BS translate to. 2 B.A. (Bachelor of Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree.", "negative_10": "West Hills College LemooreAssociate of Arts (A.A.), Liberal Arts and Sciences/Liberal StudiesAssociate of Arts (A.A.), Liberal Arts and Sciences/Liberal Studies. -Student Government President for two years. -Valedictorian. -Alpha Gamma Sigma (Alpha Chi chapter) President/College Relations Liaison.", "negative_11": "You can pursue associate degree in academic area such as business administration, law, arts, engineering, paralegal studies, liberal arts, computer science, and more. Q: What are online associate programs?", "negative_12": "liberal arts definition The areas of learning that cultivate general intellectual ability rather than technical or professional skills. Liberal arts is often used as a synonym for humanities, because literature, languages, history, and philosophy are often considered the primary subjects of the liberal arts.", "negative_13": "liberal arts definition. The areas of learning that cultivate general intellectual ability rather than technical or professional skills. Liberal arts is often used as a synonym for humanities, because literature, languages, history, and philosophy are often considered the primary subjects of the liberal arts.", "negative_14": "College Rankings. Best Liberal Arts Colleges-Narrow your search with the U.S. News rankings of Liberal Arts Colleges, schools that emphasize undergrad liberal arts education. More College Rankings & Lists.", "negative_15": "Liberal arts college. A liberal arts college is a college with an emphasis on undergraduate study in the liberal arts and sciences. A liberal arts college aims to impart a broad general knowledge and develop general intellectual capacities, in contrast to a professional, vocational, or technical curriculum.", "negative_16": "Associate in Liberal Arts Degree. Some subjects that are emphasized in a liberal arts associate's degree program include literature, sciences, history, foreign languages, mathematics and philosophy.", "negative_17": "Gonzaga University \u00e2\u0080\u0093 A Catholic Liberal Arts Education. Gonzaga University is a private liberal arts college located in Spokane, Washington. Providing a Catholic liberal arts education, we are dedicated to the Jesuit, Catholic, humanistic ideals of educating the mind, body and spirit to create men and women for others.", "negative_18": "Communications majors had average starting salaries of $43,700 last year, a bit higher than liberal arts and sciences/general studies grads. Another major, education, which is also arguably a liberal arts degree, logged an average 2012 starting salary of $40,700, in the mid-range of the liberal arts degrees. Here are NACE\u00e2\u0080\u0099s tallies on 2012 average starting salaries for those with liberal arts degrees, broken out as a table:", "negative_19": "In a 3-2 program, you end up with two bachelor's degrees: a liberal arts degree and an engineering degree. Examples of 3-2 programs include Colby College (liberal arts) with Dartmouth College, Mount Holyoke (liberal arts) with Caltech, Reed College (liberal arts) with Columbia, Rensselaer or Caltech.", "negative_20": "The two most common types of transfer associate degrees are the Associate of Arts (AA), a liberal arts degree, and the Associate of Science (AS), a liberal arts degree with a greater focus on math and sciences.", "negative_21": "Class of 2014 First-Destination Survey: Salaries for Liberal Arts/Humanities Majors Liberal arts/general studies majors earned the top average starting salary among Class of 2014 liberal arts graduates at the bachelor\u00e2\u0080\u0099s degree level, according to NACE\u00e2\u0080\u0099s Spring 2015 Salary Survey report.lass of 2014 First-Destination Survey: Salaries for Liberal Arts/Humanities Majors Liberal arts/general studies majors earned the top average starting salary among Class of 2014 liberal arts graduates at the bachelor\u00e2\u0080\u0099s degree level, according to NACE\u00e2\u0080\u0099s Spring 2015 Salary Survey report.", "negative_22": "1 I think the question requires more of an explanation than what the terms BA and BS translate to. B.A. (Bachelor of Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree.", "negative_23": "What Does LCSW Stand For? / Human and Social... / Liberal Arts and... / Education and Career FAQs", "negative_24": "Shale boom: Pipeline welders make $150,000 in Ohio, while liberal arts majors flounder. The economy is tough, especially if you have a liberal arts degree, writes Ohio Treasurer Josh Mandel. While liberal arts majors are forced to take low-paying jobs, pipeline welders are making six figures thanks to the country\u00e2\u0080\u0099s oil and gas boom.", "negative_25": "1 I think the question requires more of an explanation than what the terms BA and BS translate to. 2 B.A. (Bachelor of Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree.", "negative_26": "Madison, Wisconsin is known not only as a college town with an incredibly liberal arts scene, it is also a hub of left-wing political ideology. So what are the most liberal, forward-thinking college towns in America?", "negative_27": "What is a Bachelor of Arts (B.A.)? A Bachelor of the Arts degree program provides students with a more expansive education, requiring fewer credits that are directly linked to a particular major. Instead, students are expected to earn credits in a variety of liberal arts subjects.", "negative_28": "liberal arts definition The areas of learning that cultivate general intellectual ability rather than technical or professional skills. The term liberal arts is often used as a synonym for humanities, although the liberal arts also include the sciences.", "negative_29": "liberal arts definition. The areas of learning that cultivate general intellectual ability rather than technical or professional skills. The term liberal arts is often used as a synonym for humanities, although the liberal arts also include the sciences.", "negative_30": "Liberal arts college. A liberal arts college is a college with an emphasis on undergraduate study in the liberal arts and sciences. A liberal arts college aims to impart a broad general knowledge and develop general intellectual capacities, in contrast to a professional, vocational, or technical curriculum. Students in a liberal arts college generally major in a particular discipline while receiving exposure to a wide range of academic subjects, including sciences as well as the traditional humanities subjects taught", "negative_31": "BA = Bachelor of Arts degree BS = Bachelor of Science degreeI think the question requires more of an explanation than what the terms BA and BS translate to. B.A. (Bachelor of \u00e2\u0080\u00a6 Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree.", "negative_32": "About Liberal Arts and Sciences. Liberal studies in the arts and sciences give you the opportunity to obtain a well-rounded education. These degrees can be used as a springboard to specific graduate studies. There are also a number of individual studies that you might take in concert with a liberal arts or science curriculum.", "negative_33": "Liberal Arts Degrees. A liberal arts education is defined as the general knowledge that develops the rational thought and intellectual capabilities of individuals, communities and societies. Primarily, the following subjects fall under the domain of liberal arts studies: Literature. languages. Philosophy.", "negative_34": "To this Mannoia warns, \u00e2\u0080\u009cLike faith without works is dead, an education that remains. only theoretical is of little value.\u00e2\u0080\u009d79 Third, the integration of faith and learning, the very motto of our university, is what. thrusts Christian liberal arts education beyond the liberal arts.", "negative_35": "The liberal arts education at the secondary school level prepares the student for higher education at a university. They are thus meant for the more academically minded students. In addition to the usual curriculum, students of a liberal arts education often study Latin and Ancient Greek. Some liberal arts education provide general education, others have a specific focus.", "negative_36": "Liberal Arts Defined. The liberal arts are a set of academic disciplines that include the sciences and the humanities. When you study a liberal arts curriculum, you don't have to have one specific career goal, although you might. Instead, you'll assemble a broad foundation of knowledge that can be used in a wide spectrum of careers.", "negative_37": "What Kind of Classes Are In An AA Degree Program? Similar to a Bachelor of Arts (BA), an Associate of Arts provides students with a foundational education in liberal arts. Studies may include coursework in humanities, social sciences, history, and mathematics, among other subjects.", "negative_38": "1 BA = Bachelor of Arts degree BS = Bachelor of Science degree. 2 I think the question requires more of an explanation than what the terms BA and BS translate to. B.A. (Bachelor of Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree.", "negative_39": "ba bachelor of arts degree bs bachelor of science degreei think the question requires more of an explanation than what the terms ba and bs translate to b a bachelor of arts a bachelor of arts b a degree is what is generally called a liberal arts degree", "negative_40": "BA = Bachelor of Arts degree BS = Bachelor of Science degree . I think the question requires more of an explanation than what the terms BA and BS translate to. . B.A. (Bac\u00e2\u0080\u00a6helor of Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree.", "negative_41": "Class of 2014 First-Destination Survey: Salaries for Liberal Arts/Humanities Majors Liberal arts/general studies majors earned the top average starting salary among Class of 2014 liberal arts graduates at the bachelor\u00e2\u0080\u0099s degree level, according to NACE\u00e2\u0080\u0099s Spring 2015 Salary Survey report.lass of 2014: Top-Paid Liberal Arts Majors Majors in foreign languages and literatures were the top-paid among Class of 2014 liberal arts graduates at the bachelor\u00e2\u0080\u0099s degree level, according to results of NACE\u00e2\u0080\u0099s September 2014 Salary Survey.", "negative_42": "The University of Puget Sound is a beautiful liberal arts campus where students and faculty engage in intellectual and exciti... What is your overall opinion of this school? The University of Puget Sound is a beautiful liberal arts campus where students and faculty engage in intellectual and exciting discussions.", "negative_43": "Baccalaureate degrees: Most degrees awarded from a liberal arts college are four-year bachelor's degrees such as a B.A. (bachelor of arts) or B.S. (bachelor of science). Small size: Nearly all liberal arts colleges have fewer than 5,000 students, and most are in the 1,000 to 2,500 student range.", "negative_44": "1 BA = Bachelor of Arts degree BS = Bachelor of Science degree. 2 I think the question requires more of an explanation than what the terms BA and BS translate to. 3 B.A. (Bachelor of Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree.", "negative_45": "Occidental College is the oldest liberal arts college in Los Angeles and one of the few liberal arts colleges located in a major city. In 2014, U.S. News and World Report ranked Occidental as No. 44 on the list of National Liberal Arts Colleges.", "negative_46": "Class of 2014 First-Destination Survey: Salaries for Liberal Arts/Humanities Majors Liberal arts/general studies majors earned the top average starting salary among Class of 2014 liberal arts graduates at the bachelor\u00e2\u0080\u0099s degree level, according to NACE\u00e2\u0080\u0099s Spring 2015 Salary Survey report.", "negative_47": "The Trivium are the first three of the seven liberal arts and sciences and the Quadrivium are the remaining four. The three subjects which make up the Trivium are p The Liberal Arts of Logic, Grammar, and Rhetoric by Sister Miriam Joseph.", "negative_48": "liberal arts definition. The areas of learning that cultivate general intellectual ability rather than technical or professional skills. The term liberal arts is often used as a synonym for humanities, although the liberal arts also include the sciences. The word liberal comes from the Latin liberalis, meaning suitable for a free man, as opposed to a slave.", "negative_49": "An interdisciplinary liberal arts background, spanning both social and biological sciences, is the ideal preparation for the MSW program, but you are encouraged to apply even if your bachelor\u00e2\u0080\u0099s degree was not in the liberal arts.", "negative_50": "Confidence votes 5. 1 BA = Bachelor of Arts degree BS = Bachelor of Science degree. 2 I think the question requires more of an explanation than what the terms BA and BS translate to. 3 B.A. (Bachelor of Arts) A bachelor of arts (B.A.) degree is what is generally called a liberal arts degree." } ``` * Deduplified: No #### `triplet-50-ids` subset * Columns: "query", "positive", 'negative_1', 'negative_2', 'negative_3', 'negative_4', 'negative_5', 'negative_6', 'negative_7', 'negative_8', 'negative_9', 'negative_10', 'negative_11', 'negative_12', 'negative_13', 'negative_14', 'negative_15', 'negative_16', 'negative_17', 'negative_18', 'negative_19', 'negative_20', 'negative_21', 'negative_22', 'negative_23', 'negative_24', 'negative_25', 'negative_26', 'negative_27', 'negative_28', 'negative_29', 'negative_30', 'negative_31', 'negative_32', 'negative_33', 'negative_34', 'negative_35', 'negative_36', 'negative_37', 'negative_38', 'negative_39', 'negative_40', 'negative_41', 'negative_42', 'negative_43', 'negative_44', 'negative_45', 'negative_46', 'negative_47', 'negative_48', 'negative_49', 'negative_50' * Column types: `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int`, `int` * Examples: ```python { "query": 571018, "positive": 7349777, "negative_1": 6948601, "negative_2": 5129919, "negative_3": 6717931, "negative_4": 1065943, "negative_5": 1626276, "negative_6": 981824, "negative_7": 6449111, "negative_8": 1028927, "negative_9": 2524942, "negative_10": 5810175, "negative_11": 6236527, "negative_12": 7179545, "negative_13": 168979, "negative_14": 150383, "negative_15": 168983, "negative_16": 7027047, "negative_17": 3559703, "negative_18": 8768336, "negative_19": 5476579, "negative_20": 915244, "negative_21": 2202253, "negative_22": 1743842, "negative_23": 7727041, "negative_24": 1036624, "negative_25": 8432142, "negative_26": 2236979, "negative_27": 724018, "negative_28": 7179544, "negative_29": 7349780, "negative_30": 7179539, "negative_31": 6072080, "negative_32": 7790852, "negative_33": 4873670, "negative_34": 4389296, "negative_35": 2305477, "negative_36": 1626275, "negative_37": 291845, "negative_38": 1743847, "negative_39": 1508485, "negative_40": 4298457, "negative_41": 1831337, "negative_42": 1760417, "negative_43": 8768340, "negative_44": 8432143, "negative_45": 1971355, "negative_46": 1133925, "negative_47": 2105819, "negative_48": 168975, "negative_49": 5132446, "negative_50": 1316646, } ``` * Deduplified: No
提供机构：
sentence-transformers
原始信息汇总

数据集概述

基本信息

  • 名称: MS MARCO with hard negatives from co-condenser-margin-mse-sym-mnrl-mean-v1
  • 语言: 英语
  • 多语言性: 单语
  • 大小: 10M<n<100M
  • 任务类别: 特征提取, 句子相似度
  • 标签: sentence-transformers

数据集配置与特征

  1. triplet

    • 特征:
      • query: 字符串
      • positive: 字符串
      • negative: 字符串
    • 训练集:
      • 字节数: 362903736
      • 示例数: 502939
      • 下载大小: 237820181
      • 数据集大小: 362903736

  2. triplet-50

    • 特征:
      • query: 字符串
      • positive: 字符串
      • negative_1 至 negative_50: 字符串
    • 训练集:
      • 字节数: 9055260814
      • 示例数: 502939
      • 下载大小: 5914557509
      • 数据集大小: 9055260814

  3. triplet-50-ids

    • 特征:
      • query: 整数
      • positive: 整数
      • negative_1 至 negative_50: 整数
    • 训练集:
      • 字节数: 209222624
      • 示例数: 502939
      • 下载大小: 178195399
      • 数据集大小: 209222624

  4. triplet-all

    • 特征:
      • query: 字符串
      • positive: 字符串
      • negative: 字符串
    • 训练集:
      • 字节数: 19858927973
      • 示例数: 26637550
      • 下载大小: 4244209962
      • 数据集大小: 19858927973

  5. triplet-all-ids

    • 特征:
      • query: 整数
      • positive: 整数
      • negative: 整数
    • 训练集:
      • 字节数: 639301200
      • 示例数: 26637550
      • 下载大小: 190130784
      • 数据集大小: 639301200

  6. triplet-hard

    • 特征:
      • query: 字符串
      • positive: 字符串
      • negative: 字符串
    • 训练集:
      • 字节数: 8477376947
      • 示例数: 11662655
      • 下载大小: 2152733400
      • 数据集大小: 8477376947

  7. triplet-hard-ids

    • 特征:
      • query: 整数
      • positive: 整数
      • negative: 整数
    • 训练集:
      • 字节数: 279903720
      • 示例数: 11662655
      • 下载大小: 89399265
      • 数据集大小: 279903720

  8. triplet-ids

    • 特征:
      • query: 整数
      • positive: 整数
      • negative: 整数
    • 训练集:
      • 字节数: 12070536
      • 示例数: 502939
      • 下载大小: 10132130
      • 数据集大小: 12070536

数据集子集

  • Unique Triplets: 每个query-positive对，提取最相似的段落作为negative。

    • triplet: 字符串格式
    • triplet-ids: 整数格式

  • All Triplets: 每个query-positive对，提取50个最相似的段落作为negatives。

    • triplet-all: 字符串格式
    • triplet-all-ids: 整数格式

  • Hard Triplets: 每个query-positive对，提取50个最相似的段落作为negatives，并使用CrossEncoder确保negatives与query不相似。

    • triplet-hard: 字符串格式
    • triplet-hard-ids: 整数格式

  • 50 "Triplets": 每个query-positive对，提取50个最相似的段落作为negatives，存储为单一样本。

    • triplet-50: 字符串格式
    • triplet-50-ids: 整数格式
搜集汇总
数据集介绍
main_image_url
构建方式
该数据集源自微软构建的大规模信息检索语料库MS MARCO，其核心构建思路在于为每个查询与相关正例段落挖掘高质量的难负例。具体而言，研究人员利用co-condenser-margin-mse-sym-mnrl-mean-v1模型，针对每个查询-正例对，从语料库中检索出与查询最为相似的50个段落作为候选负例。在此基础上，数据集提供了多种配置：其一为保留所有50个候选负例的完整三元组集合；其二通过交叉编码器ms-marco-MiniLM-L-6-v2进行过滤，仅保留那些满足相似度条件（即正例与查询的相似度大于负例与查询的相似度加上一个预设的边际值）的样本，从而形成“硬”三元组；其三则直接选取相似度最高的一个段落作为唯一负例。所有配置均同时提供文本字符串与整数标识符两种数据格式，以适配不同的使用场景。
特点
该数据集最显著的特点在于其丰富的层次化结构与高度的灵活性。它提供了从单一负例到50个负例的多种三元组配置，包括基础三元组、全部三元组以及经过严格筛选的硬三元组，满足了从简单对比学习到复杂难例挖掘等不同训练策略的需求。每个配置均包含查询、正例与负例字段，且负例均来自多个先进模型的检索结果，确保了负例的多样性与挑战性。此外，数据集同时存储了原始文本与对应的语料库标识符，使得用户既可以直接加载文本进行训练，也可以结合外部语料库进行更精细化的处理。这种设计不仅提升了数据集的易用性，也为其在句子嵌入、语义相似度计算等领域的应用提供了坚实的数据基础。
使用方法
该数据集专为训练Sentence Transformer模型而设计，可通过HuggingFace Datasets库便捷加载。用户可根据训练目标选择不同的配置：若需进行基础的三元组损失训练，可选用'triplet'或'triplet-ids'配置；若希望利用多个负例提升模型判别力，则可使用'triplet-50'或'triplet-all'配置；而追求更高训练难度的场景下，'triplet-hard'系列配置尤为合适。加载后，数据以字典形式呈现，包含'query'、'positive'及'negative'（或'negative_1'至'negative_50'）等字段。用户可直接将这些字段输入至Sentence Transformer的训练管线中，通过对比学习范式优化模型。对于使用标识符的配置，需配合'sentence-transformers/msmarco-corpus'语料库进行文本映射。
背景与挑战
背景概述
该数据集由Sentence Transformers团队基于微软MS MARCO语料库构建，核心研究问题聚焦于如何通过大规模、高质量的难负样本挖掘来提升句向量模型的检索与语义匹配能力。MS MARCO作为源自Bing搜索引擎真实用户查询的大规模信息检索基准，自2016年发布以来便成为评估稠密检索模型性能的标杆。该数据集利用co-condenser-margin-mse-sym-mnrl-mean-v1等13种不同模型，为每个查询-正例对挖掘50个最相似的段落作为负样本，从而构建出包含三元组、全部负样本及硬负样本等多种配置的子集。其影响力体现在为Sentence Transformer等稠密检索框架提供了标准化的训练数据，推动了对比学习与难负样本策略在信息检索领域的深度融合。
当前挑战
该数据集所解决的领域问题在于克服传统信息检索中负样本选取的随机性与低效性——简单负样本无法有效训练模型区分语义相近的文本对，导致检索精度受限。构建过程中面临的核心挑战包括：1）如何从海量语料中高效检索与查询语义高度相似的候选段落，这要求整合多种嵌入模型的输出并平衡计算开销；2）硬负样本的筛选需依赖CrossEncoder进行二次验证，确保负样本与查询的相似度显著低于正例，但边际阈值的设定（如margin=3.0）缺乏通用标准，易引入噪声或丢失有效样本；3）数据规模庞大（如triplet-all子集包含超2600万条样本），对存储、加载及训练时的采样策略提出严苛要求。
常用场景
经典使用场景
在信息检索与自然语言处理领域，该数据集最经典的使用场景是训练和评估句嵌入模型（Sentence Transformer），尤其是在稠密检索任务中通过三元组损失（triplet loss）进行对比学习。基于MS MARCO真实用户查询与Bing搜索引擎的交互数据，数据集中每个查询对应一个正例段落和多个通过多种模型挖掘的难负例（hard negatives），使模型能够学习到更精细的语义区分能力。研究者常利用其提供的不同子集（如triplet-hard、triplet-all）来优化模型对查询与相关文档之间相似度的排序性能，从而提升检索系统的召回率与精确度。
实际应用
在实际应用中，该数据集主要赋能搜索引擎、智能问答系统和对话式AI产品的语义理解模块。例如，企业级搜索平台可基于此数据集训练的模型实现文档与用户意图的精准匹配，提升内部知识库的检索效率。在电商领域，其可用于商品标题与用户查询的语义对齐，优化推荐系统的相关性。此外，该数据集还支持构建多轮对话中的上下文检索器，使虚拟助手能够从海量文档中快速定位答案片段，显著改善用户体验。其提供的ID版本子集便于与MS MARCO语料库直接对接，降低了工业级部署的工程成本。
衍生相关工作
该数据集衍生了一系列经典工作，包括基于co-condenser架构的对比学习框架，其通过margin-MSE损失和对称多负样本排序（sym-MNRL）策略，进一步提升了句嵌入的判别力。研究者还提出了多种难负例采样策略，如利用交叉编码器动态过滤噪音负样本，以及结合知识蒸馏的轻量级模型训练方法。此外，该数据集催生了多个高性能Sentence Transformer模型（如all-MiniLM-L6-v2系列），这些模型在语义文本相似度（STS）和检索基准任务上取得了突破，并推动了后续如SimCSE、Contriever等无监督对比学习范式的演进。
以上内容由遇见数据集搜集并总结生成
关于我们
遇见数据集——让每个数据集都被发现，让每一次遇见都有价值。
数发科官网www.sfktec.com
联系我们
youjian
selectdataset@iotsh.com.cn
商业合作
数据驱动未来，携手共赢发展
二维码
社区交流群
二维码
科研交流群
商业服务