diff --git a/datasets/russian_super_glue/README.md b/datasets/russian_super_glue/README.md
new file mode 100644
index 00000000000..b574ee1ffee
--- /dev/null
+++ b/datasets/russian_super_glue/README.md
@@ -0,0 +1,683 @@
+---
+pretty_name: Russian SuperGLUE
+annotations_creators:
+- crowdsourced
+- expert-generated
+language_creators:
+- crowdsourced
+- expert-generated
+languages:
+- ru-RU
+licenses:
+- mit
+multilinguality:
+- monolingual
+size_categories:
+- 100K<n<1M
+- 1M<n<10M
+- 10M<n<100M
+- 100M<n<1B
+source_datasets:
+- original
+task_categories:
+- text-classification
+task_ids:
+- natural-language-inference
+- multi-class-classification
+---
+
+# Dataset Card for [Russian SuperGLUE]
+
+## Table of Contents
+- [Table of Contents](#table-of-contents)
+- [Dataset Description](#dataset-description)
+  - [Dataset Summary](#dataset-summary)
+  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
+  - [Languages](#languages)
+- [Dataset Structure](#dataset-structure)
+  - [Data Instances](#data-instances)
+  - [Data Fields](#data-fields)
+  - [Data Splits](#data-splits)
+- [Dataset Creation](#dataset-creation)
+  - [Curation Rationale](#curation-rationale)
+  - [Source Data](#source-data)
+  - [Annotations](#annotations)
+  - [Personal and Sensitive Information](#personal-and-sensitive-information)
+- [Considerations for Using the Data](#considerations-for-using-the-data)
+  - [Social Impact of Dataset](#social-impact-of-dataset)
+  - [Discussion of Biases](#discussion-of-biases)
+  - [Other Known Limitations](#other-known-limitations)
+- [Additional Information](#additional-information)
+  - [Dataset Curators](#dataset-curators)
+  - [Licensing Information](#licensing-information)
+  - [Citation Information](#citation-information)
+  - [Contributions](#contributions)
+
+## Dataset Description
+
+- **Homepage:** https://russiansuperglue.com/
+- **Repository:** https://github.com/RussianNLP/RussianSuperGLUE
+- **Paper:** https://russiansuperglue.com/download/main_article
+- **Leaderboard:** https://russiansuperglue.com/leaderboard/2
+- **Point of Contact:** [More Information Needed]
+
+### Dataset Summary
+
+Modern universal language models and transformers such as BERT, ELMo, XLNet, RoBERTa and others need to be properly
+compared and evaluated. In the last year, new models and methods for pretraining and transfer learning have driven 
+striking performance improvements across a range of language understanding tasks.
+
+
+We offer testing methodology based on tasks, typically proposed for “strong AI” — logic, commonsense, reasoning.
+Adhering to the GLUE and SuperGLUE methodology, we present a set of test tasks for general language understanding
+and leaderboard models.
+
+
+For the first time a complete test for Russian language was developed, which is similar to its English analog.
+Many datasets were composed for the first time, and a leaderboard of models for the Russian language with comparable
+results is also presented.
+
+### Supported Tasks and Leaderboards
+
+Supported tasks, barring a few additions, are equivalent to the original SuperGLUE tasks.
+
+|Task Name|Equiv. to|
+|----|---:|
+|Linguistic Diagnostic for Russian|Broadcoverage Diagnostics (AX-b)|
+|Russian Commitment Bank (RCB)|CommitmentBank (CB)|
+|Choice of Plausible Alternatives for Russian language (PARus)|Choice of Plausible Alternatives (COPA)|
+|Russian Multi-Sentence Reading Comprehension (MuSeRC)|Multi-Sentence Reading Comprehension (MultiRC)|
+|Textual Entailment Recognition for Russian (TERRa)|Recognizing Textual Entailment (RTE)|
+|Russian Words in Context (based on RUSSE)|Words in Context (WiC)|
+|The Winograd Schema Challenge (Russian)|The Winograd Schema Challenge (WSC)|
+|Yes/no Question Answering Dataset for the Russian (DaNetQA)|BoolQ|
+|Russian Reading Comprehension with Commonsense Reasoning (RuCoS)|Reading Comprehension with Commonsense Reasoning (ReCoRD)|
+
+### Languages
+
+All tasks are in Russian.
+
+## Dataset Structure
+
+### Data Instances
+
+Note that there are no labels in the `test` splits. This is signified by the `-1` value.
+
+#### LiDiRus
+
+- **Size of downloaded dataset files:** 0.047 MB
+- **Size of the generated dataset:** 0.47 MB
+- **Total amount of disk used:** 0.517 MB
+
+An example of 'test' looks as follows
+```
+{
+    "sentence1": "Новая игровая консоль доступна по цене.",
+    "sentence2": "Новая игровая консоль недоступна по цене.",
+    "knowledge": "",
+    "lexical-semantics": "Morphological negation",
+    "logic": "Negation",
+    "predicate-argument-structure": "",
+    "idx": 10,
+    "label": 1
+}
+```
+
+#### RCB
+
+- **Size of downloaded dataset files:** 0.134 MB
+- **Size of the generated dataset:** 0.504 MB
+- **Total amount of disk used:** 0.641 MB
+
+An example of 'train'/'dev' looks as follows
+```
+{
+    "premise": "— Пойдём пообедаем. Я с утра ничего не ел. Отель, как видишь, весьма посредственный, но мне сказали,
+     что в здешнем ресторане отлично готовят.",
+    "hypothesis": "В здешнем ресторане отлично готовят.",
+    "verb": "сказать",
+    "negation": "no_negation",
+    "idx": 10,
+    "label": 2
+}
+```
+
+An example of 'test' looks as follows
+```
+{
+    "premise": "Я уверен, что вместе мы победим. Да, парламентское большинство думает иначе.",
+    "hypothesis": "Вместе мы проиграем.",
+    "verb": "думать",
+    "negation": "no_negation",
+    "idx": 10,
+    "label": -1
+}
+```
+
+#### PARus
+
+- **Size of downloaded dataset files:** 0.057 MB
+- **Size of the generated dataset:** 0.187 MB
+- **Total amount of disk used:**  0.245 MB
+
+An example of 'train'/'dev' looks as follows
+```
+{
+    "premise": "Женщина чинила кран.",
+    "choice1": "Кран подтекал.",
+    "choice2": "Кран был выключен.",
+    "question": "cause",
+    "idx": 10,
+    "label": 0
+}
+```
+
+An example of 'test' looks as follows
+```
+{
+    "premise": "Ребятам было страшно.",
+    "choice1": "Их вожатый рассказал им историю про призрака.",
+    "choice2": "Они жарили маршмеллоу на костре.",
+    "question": "cause",
+    "idx": 10,
+    "label": -1
+}
+```
+#### MuSeRC
+
+- **Size of downloaded dataset files:** 1.2 MB
+- **Size of the generated dataset:** 57 MB
+- **Total amount of disk used:** 59 MB
+
+An example of 'train'/'dev' looks as follows
+```
+{
+    "paragraph": "(1) Но люди не могут существовать без природы, поэтому в парке стояли железобетонные скамейки —
+     деревянные моментально ломали. (2) В парке бегали ребятишки, водилась шпана, которая развлекалась игрой в карты,
+     пьянкой, драками, «иногда насмерть». (3) «Имали они тут и девок...» (4) Верховодил шпаной Артемка-мыло, с
+     вспененной белой головой. (5) Людочка сколько ни пыталась усмирить лохмотья на буйной голове Артемки, ничего у
+     неё не получалось. (6) Его «кудри, издали напоминавшие мыльную пену, изблизя оказались что липкие рожки из
+     вокзальной столовой — сварили их, бросили комком в пустую тарелку, так они, слипшиеся, неподъёмно и лежали.
+     (7) Да и не ради причёски приходил парень к Людочке. (8) Как только её руки становились занятыми ножницами
+     и расчёской, Артемка начинал хватать её за разные места. (9) Людочка сначала увёртывалась от хватких рук Артемки,
+     а когда не помогло, стукнула его машинкой по голове и пробила до крови, пришлось лить йод на голову «ухажористого
+     человека». (10) Артемка заулюлюкал и со свистом стал ловить воздух. (11) С тех пор «домогания свои хулиганские
+     прекратил», более того, шпане повелел Людочку не трогать.",
+    "question": "Как развлекались в парке ребята?",
+    "answer": "Развлекались игрой в карты, пьянкой, драками, снимали они тут и девок.",
+    "idx":
+    {
+        "paragraph": 0,
+        "question": 2,
+        "answer": 10
+    },
+    "label": 1
+}
+```
+
+An example of 'test' looks as follows
+```
+
+{
+    "paragraph": "\"(1) Издательство Viking Press совместно с компанией TradeMobile выпустят мобильное приложение,
+     посвященное Анне Франк, передает The Daily Telegraph. (2) Программа будет включать в себя фрагменты из дневника
+     Анны, озвученные британской актрисой Хеленой Бонэм Картер. (3) Помимо этого, в приложение войдут фотографии
+     и видеозаписи, документы из архива Фонда Анны Франк, план здания в Амстердаме, где Анна с семьей скрывались от
+     нацистов, и факсимильные копии страниц дневника. (4) Приложение, которое получит название Anne Frank App, выйдет
+     18 октября. (5) Интерфейс программы будет англоязычным. (6) На каких платформах будет доступно Anne Frank App,
+     не уточняется. Анна Франк родилась в Германии в 1929 году. (7) Когда в стране начались гонения на евреев, Анна с
+     семьей перебрались в Нидерланды. (8) С 1942 года члены семьи Франк и еще несколько человек скрывались от нацистов
+     в потайных комнатах дома в Амстердаме, который занимала компания отца Анны. (9) В 1944 году группу по доносу
+     обнаружили гестаповцы. (10) Обитатели \"Убежища\" (так Анна называла дом в дневнике) были отправлены в концлагеря;
+     выжить удалось только отцу девочки Отто Франку. (11) Находясь в \"Убежище\", Анна вела дневник, в котором описывала
+     свою жизнь и жизнь своих близких. (12) После ареста книгу с записями сохранила подруга семьи Франк и впоследствии
+     передала ее отцу Анны. (13) Дневник был впервые опубликован в 1947 году. (14) Сейчас он переведен более
+     чем на 60 языков.\"",
+    "question": "Какая информация войдет в новой мобильное приложение?",
+    "answer": "Видеозаписи Анны Франк.",
+    "idx":
+    {
+        "paragraph": 0,
+        "question": 2,
+        "answer": 10
+    },
+    "label": -1
+}
+```
+#### TERRa
+
+- **Size of downloaded dataset files:** 0.887 MB
+- **Size of the generated dataset:** 3.28 MB
+- **Total amount of disk used:** 4.19 MB
+
+An example of 'train'/'dev' looks as follows
+```
+{
+    "premise": "Музей, расположенный в Королевских воротах, меняет экспозицию. На смену выставке, рассказывающей об
+     истории ворот и их реставрации, придет «Аптека трех королей». Как рассказали в музее, посетители попадут в 
+     традиционный интерьер аптеки.",
+    "hypothesis": "Музей закроется навсегда.",
+    "idx": 10,
+    "label": 1
+}
+```
+
+An example of 'test' looks as follows
+```
+{
+    "premise": "Маршрутка полыхала несколько минут. Свидетели утверждают, что приезду пожарных салон «Газели» выгорел полностью. К счастью, пассажиров внутри не было, а водитель успел выскочить из кабины.",
+    "hypothesis": "Маршрутка выгорела.",
+    "idx": 10,
+    "label": -1
+}
+```
+
+#### RUSSE
+
+- **Size of downloaded dataset files:** 3.7 MB
+- **Size of the generated dataset:** 20 MB
+- **Total amount of disk used:** 24 MB
+
+An example of 'train'/'dev' looks as follows
+```
+{
+    "word": "дух",
+    "sentence1": "Завертелась в доме веселая коловерть: праздничный стол, праздничный дух, шумные разговоры",
+    "sentence2": "Вижу: духи собралися / Средь белеющих равнин. // Бесконечны, безобразны, / В мутной месяца игре / Закружились бесы разны, / Будто листья в ноябре",
+    "start1": 68,
+    "start2": 6,
+    "end1": 72,
+    "end2": 11,
+    "gold_sense1": 3,
+    "gold_sense2": 4,
+    "idx": 10,
+    "label": 0
+}
+```
+
+An example of 'test' looks as follows
+```
+{
+    "word": "доска",
+    "sentence1": "На 40-й день после трагедии в переходе была установлена мемориальная доска, надпись на которой гласит: «В память о погибших и пострадавших от террористического акта 8 августа 2000 года».",
+    "sentence2": "Фото с 36-летним миллиардером привлекло сеть его необычной фигурой при стойке на доске и кремом на лице.",
+    "start1": 69,
+    "start2": 81,
+    "end1": 73,
+    "end2": 85,
+    "gold_sense1": -1,
+    "gold_sense2": -1,
+    "idx": 10,
+    "label": -1
+}
+```
+
+#### RWSD
+
+- **Size of downloaded dataset files:** 0.04 MB
+- **Size of the generated dataset:** 0.279 MB
+- **Total amount of disk used:**  0.320 MB
+
+An example of 'train'/'dev' looks as follows
+```
+{
+    "text": "Женя поблагодарила Сашу за помощь, которую она оказала.",
+    "span1_index": 0,
+    "span2_index": 6,
+    "span1_text": "Женя",
+    "span2_text": "она оказала",
+    "idx": 10,
+    "label": 0
+}
+```
+
+An example of 'test' looks as follows
+```
+{
+    "text": "Мод и Дора видели, как через прерию несутся поезда, из двигателей тянулись клубы черного дыма. Ревущие 
+    звуки их моторов и дикие, яростные свистки можно было услышать издалека. Лошади убежали, когда они увидели 
+    приближающийся поезд.",
+    "span1_index": 22,
+    "span2_index": 30,
+    "span1_text": "свистки",
+    "span2_text": "они увидели",
+    "idx": 10,
+    "label": -1
+}
+```
+
+#### DaNetQA
+
+- **Size of downloaded dataset files:** 1.3 MB
+- **Size of the generated dataset:** 4.6 MB
+- **Total amount of disk used:**  5.9 MB
+
+An example of 'train'/'dev' looks as follows
+```
+{
+    "question": "Вреден ли алкоголь на первых неделях беременности?",
+    "passage": "А Бакингем-Хоуз и её коллеги суммировали последствия, найденные в обзорных статьях ранее. Частые случаи
+     задержки роста плода, результатом чего является укороченный средний срок беременности и сниженный вес при рождении.
+     По сравнению с нормальными детьми, дети 3-4-недельного возраста демонстрируют «менее оптимальную»  двигательную 
+     активность, рефлексы, и ориентацию в пространстве, а дети 4-6 лет показывают низкий уровень работы 
+     нейроповеденческих функций, внимания, эмоциональной экспрессии, и развития речи и языка. Величина этих влияний 
+     часто небольшая, частично в связи с независимыми переменными: включая употребление во время беременности 
+     алкоголя/табака, а также факторы среды . У детей школьного возраста проблемы с устойчивым вниманием и контролем 
+     своего поведения, а также незначительные с ростом, познавательными и языковыми способностями.",
+    "idx": 10,
+    "label": 1
+}
+```
+
+An example of 'test' looks as follows
+```
+{
+    "question": "Вредна ли жесткая вода?",
+    "passage": "Различают временную  жёсткость, обусловленную гидрокарбонатами кальция и магния Са2; Mg2, и постоянную
+     жёсткость, вызванную присутствием других солей, не выделяющихся при кипячении воды: в основном, сульфатов и 
+     хлоридов Са и Mg. Жёсткая вода при умывании сушит кожу, в ней плохо образуется пена при использовании мыла.
+     Использование жёсткой воды вызывает появление осадка  на стенках котлов, в трубах и т. п. В то же время, 
+     использование слишком мягкой воды может приводить к коррозии труб, так как, в этом случае отсутствует 
+     кислотно-щелочная буферность, которую обеспечивает гидрокарбонатная  жёсткость. Потребление жёсткой или мягкой 
+     воды обычно не является опасным для здоровья, однако есть данные о том, что высокая жёсткость способствует 
+     образованию мочевых камней, а низкая — незначительно увеличивает риск сердечно-сосудистых заболеваний. Вкус 
+     природной питьевой воды, например, воды родников, обусловлен именно присутствием солей жёсткости.",
+    "idx": 100,
+    "label": -1
+}
+```
+
+#### RuCoS
+
+- **Size of downloaded dataset files:** 54 MB
+- **Size of the generated dataset:** 193 MB
+- **Total amount of disk used:** 249 MB
+
+An example of 'train'/'dev' looks as follows
+```
+{
+    "passage": "В Абхазии 24 августа на досрочных выборах выбирают нового президента. Кто бы ни стал победителем,
+     возможности его будут ограничены, говорят эксперты, опрошенные DW. В Абхазии 24 августа проходят досрочные выборы
+     президента не признанной международным сообществом республики. Толчком к их проведению стали массовые протесты в 
+     конце мая 2014 года, в результате которых со своего поста был вынужден уйти действующий президент Абхазии Александр
+     Анкваб. Эксперты называют среди наиболее перспективных кандидатов находящегося в оппозиции политика Рауля Хаджимбу,
+     экс-главу службы безопасности Аслана Бжанию и генерала Мираба Кишмарию, исполняющего обязанности министра обороны.
+     У кого больше шансов\n\"Ставки делаются на победу Хаджимбы.\n@highlight\nВ Швеции задержаны двое граждан РФ в связи
+     с нападением на чеченского блогера\n@highlight\nТуризм в эпоху коронавируса: куда поехать? И ехать ли 
+     вообще?\n@highlight\nКомментарий: Россия накануне эпидемии - виноватые назначены заранее",
+    "query": "Несмотря на то, что Кремль вложил много денег как в @placeholder, так и в Южную Осетию, об экономическом
+     восстановлении данных регионов говорить не приходится, считает Хальбах: \"Многие по-прежнему живут в 
+     полуразрушенных домах и временных жилищах\".",
+    "entities":
+    [
+        "DW.",
+        "Абхазии ",
+        "Александр Анкваб.",
+        "Аслана Бжанию ",
+        "Мираба Кишмарию,",
+        "РФ ",
+        "Рауля Хаджимбу,",
+        "Россия ",
+        "Хаджимбы.",
+        "Швеции "
+    ],
+    "answers":
+    [
+        "Абхазии"
+    ],
+    "idx":
+    {
+        "passage": 500,
+        "query": 500
+    }
+}
+```
+
+An example of 'test' looks as follows
+```
+{
+    "passage": "Почему и как изменится курс белорусского рубля? Какие инструменты следует предпочесть населению, чтобы 
+    сохранить сбережения, DW рассказали финансовые аналитики Беларуси. На последних валютных торгах БВФБ 2015 года в 
+    среду, 30 декабря, курс белорусского рубля к доллару - 18569, к евро - 20300, к российскому рублю - 255. В 2016 
+    году белорусскому рублю пророчат падение как минимум на 12 процентов к корзине валют, к которой привязан его курс. 
+    А чтобы избежать потерь, белорусам советуют диверсифицировать инвестиционные портфели. Чем обусловлены прогнозные 
+    изменения котировок белорусского рубля, и какие финансовые инструменты стоит предпочесть, чтобы минимизировать риск
+    потерь?\n@highlight\nВ Германии за сутки выявлено более 100 новых заражений коронавирусом\n@highlight\nРыночные цены
+    на нефть рухнули из-за провала переговоров ОПЕК+\n@highlight\nВ Италии за сутки произошел резкий скачок смертей от
+    COVID-19",
+    "query": "Последнее, убежден аналитик, инструмент для узкого круга профессиональных инвесторов, культуры следить за
+     финансовым состоянием предприятий - такой, чтобы играть на рынке корпоративных облигаций, - в @placeholder пока нет.",
+    "entities":
+    [
+        "DW ",
+        "Беларуси.",
+        "Германии ",
+        "Италии ",
+        "ОПЕК+"
+    ],
+    "answers": [],
+    "idx":
+    {
+        "passage": 500,
+        "query": 500
+    }
+}
+```
+
+### Data Fields
+
+#### LiDiRus
+
+- `idx`: an `int32` feature
+- `label`: a classification label, with possible values `entailment` (0), `not_entailment` (1)
+- `sentence1`: a `string` feature
+- `sentence2`: a `string` feature
+- `knowledge`: a `string` feature with possible values `''`, `'World knowledge'`, `'Common sense'`
+- `lexical-semantics`: a `string` feature
+- `logic`: a `string` feature
+- `predicate-argument-structure`: a `string` feature
+
+
+#### RCB
+
+- `idx`: an `int32` feature
+- `label`: a classification label, with possible values `entailment` (0), `contradiction` (1), `neutral` (2)
+- `premise`: a `string` feature
+- `hypothesis`: a `string` feature
+- `verb`: a `string` feature
+- `negation`: a `string` feature with possible values `'no_negation'`, `'negation'`, `''`, `'double_negation'`
+
+#### PARus
+
+- `idx`: an `int32` feature
+- `label`: a classification label, with possible values `choice1` (0), `choice2` (1)
+- `premise`: a `string` feature
+- `choice1`: a `string` feature
+- `choice2`: a `string` feature
+- `question`: a `string` feature with possible values `'cause'`, `'effect'`
+
+#### MuSeRC
+- `idx`: an `int32` feature
+- `label` : a classification label, with possible values `false` (0) , `true` (1) (does the provided `answer` contain 
+  a factual response to the `question`)
+- `paragraph`: a `string` feature
+- `question`: a `string` feature
+- `answer`: a `string` feature
+
+
+#### TERRa
+- `idx`: an `int32` feature
+- `label`: a classification label, with possible values `entailment` (0), `not_entailment` (1)
+- `premise`: a `string` feature
+- `hypothesis`: a `string` feature
+
+#### RUSSE
+- `idx`: an `int32` feature
+- `label` : a classification label, with possible values `false` (0), `true` (1) (whether the given `word` used in the
+  same sense in both sentences)
+- `word`: a `string` feature
+- `sentence1`: a `string` feature
+- `sentence2`: a `string` feature
+- `gold_sense1`: an `int32` feature
+- `gold_sense2`: an `int32` feature
+- `start1`: an `int32` feature
+- `start2`: an `int32` feature
+- `end1`: an `int32` feature
+- `end2`: an `int32` feature
+
+#### RWSD
+
+- `idx`: an `int32` feature
+- `label` : a classification label, with possible values `false` (0), `true` (1) (whether the given spans are
+  coreferential)
+- `text`: a `string` feature
+- `span1_index`: an `int32` feature
+- `span2_index`: an `int32` feature
+- `span1_text`: a `string` feature
+- `span2_text`: a `string` feature
+
+
+#### DaNetQA
+- `idx`: an `int32` feature
+- `label` : a classification label, with possible values `false` (0), `true` (1) (yes/no answer to the `question` found
+  in the `passage`)
+- `question`: a `string` feature
+- `passage`: a `string` feature
+
+#### RuCoS
+
+- `idx`: an `int32` feature
+- `passage`: a `string` feature
+- `query`: a `string` feature
+- `entities`: a `list of strings` feature
+- `answers`: a `list of strings` feature
+
+
+[More Information Needed]
+
+### Data Splits
+
+#### 	LiDiRus
+|   |test|
+|---|---:|
+|LiDiRus|1104|
+
+#### RCB
+
+| |train|validation|test|
+|----|---:|----:|---:|
+|RCB|438|220|438|
+
+#### PARus
+
+| |train|validation|test|
+|----|---:|----:|---:|
+|PARus|400|100|500|
+
+#### MuSeRC
+
+| |train|validation|test|
+|----|---:|----:|---:|
+|MuSeRC|500|100|322|
+
+
+#### TERRa
+
+| |train|validation|test|
+|----|---:|----:|---:|
+|TERRa|2616|307|3198|
+
+
+#### RUSSE
+
+| |train|validation|test|
+|----|---:|----:|---:|
+|RUSSE|19845|8508|18892|
+
+
+#### RWSD
+
+| |train|validation|test|
+|----|---:|----:|---:|
+|RWSD|606|204|154|
+
+
+#### DaNetQA
+
+| |train|validation|test|
+|----|---:|----:|---:|
+|DaNetQA|1749|821|805|
+
+
+#### RuCoS
+
+| |train|validation|test|
+|----|---:|----:|---:|
+|RuCoS|72193|7577|7257|
+
+## Dataset Creation
+
+### Curation Rationale
+
+[More Information Needed]
+
+### Source Data
+
+#### Initial Data Collection and Normalization
+
+[More Information Needed]
+
+#### Who are the source language producers?
+
+[More Information Needed]
+
+### Annotations
+
+#### Annotation process
+
+[More Information Needed]
+
+#### Who are the annotators?
+
+[More Information Needed]
+
+### Personal and Sensitive Information
+
+[More Information Needed]
+
+## Considerations for Using the Data
+
+### Social Impact of Dataset
+
+[More Information Needed]
+
+### Discussion of Biases
+
+[More Information Needed]
+
+### Other Known Limitations
+
+[More Information Needed]
+
+## Additional Information
+
+### Dataset Curators
+
+[More Information Needed]
+
+### Licensing Information
+
+All our datasets are published by MIT License.
+
+### Citation Information
+```
+@article{shavrina2020russiansuperglue,
+                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},
+                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova, Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and Evlampiev, Andrey},
+                  journal={arXiv preprint arXiv:2010.15925},
+                  year={2020}
+                  }
+```
+### Contributions
+
+Thanks to [@slowwavesleep](https://github.com/slowwavesleep) for adding this dataset.
diff --git a/datasets/russian_super_glue/dataset_infos.json b/datasets/russian_super_glue/dataset_infos.json
new file mode 100644
index 00000000000..968c599ceff
--- /dev/null
+++ b/datasets/russian_super_glue/dataset_infos.json
@@ -0,0 +1 @@
+{"lidirus": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\n\"LiDiRus (Linguistic Diagnostic for Russian) is a diagnostic dataset that covers a large volume of linguistic phenomena,\nwhile allowing you to evaluate information systems on a simple test of textual entailment recognition.\nSee more details diagnostics.\n", "citation": "\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/LiDiRus", "license": "", "features": {"sentence1": {"dtype": "string", "id": null, "_type": "Value"}, "sentence2": {"dtype": "string", "id": null, "_type": "Value"}, "knowledge": {"dtype": "string", "id": null, "_type": "Value"}, "lexical-semantics": {"dtype": "string", "id": null, "_type": "Value"}, "logic": {"dtype": "string", "id": null, "_type": "Value"}, "predicate-argument-structure": {"dtype": "string", "id": null, "_type": "Value"}, "idx": {"dtype": "int32", "id": null, "_type": "Value"}, "label": {"num_classes": 2, "names": ["entailment", "not_entailment"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "lidirus", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"test": {"name": "test", "num_bytes": 470306, "num_examples": 1104, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/LiDiRus": {"num_bytes": 47118, "checksum": "e1bc2c55cadbc98b0e77a43e1560af8e2fe336ab4bc372f545f69cb48a479cdc"}}, "download_size": 47118, "post_processing_size": null, "dataset_size": 470306, "size_in_bytes": 517424}, "rcb": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\nThe Russian Commitment Bank is a corpus of naturally occurring discourses whose final sentence contains\na clause-embedding predicate under an entailment canceling operator (question, modal, negation, antecedent\nof conditional).\n", "citation": "\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/RCB", "license": "", "features": {"premise": {"dtype": "string", "id": null, "_type": "Value"}, "hypothesis": {"dtype": "string", "id": null, "_type": "Value"}, "verb": {"dtype": "string", "id": null, "_type": "Value"}, "negation": {"dtype": "string", "id": null, "_type": "Value"}, "idx": {"dtype": "int32", "id": null, "_type": "Value"}, "label": {"num_classes": 3, "names": ["entailment", "contradiction", "neutral"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "rcb", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"train": {"name": "train", "num_bytes": 199712, "num_examples": 438, "dataset_name": "russian_super_glue"}, "validation": {"name": "validation", "num_bytes": 97993, "num_examples": 220, "dataset_name": "russian_super_glue"}, "test": {"name": "test", "num_bytes": 207031, "num_examples": 438, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/RCB": {"num_bytes": 136700, "checksum": "9ecc0bd0cc04e04922349212452166404b5a75b67be9e6aa996c80c64beb6ddb"}}, "download_size": 136700, "post_processing_size": null, "dataset_size": 504736, "size_in_bytes": 641436}, "parus": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\nChoice of Plausible Alternatives for Russian language\nChoice of Plausible Alternatives for Russian language (PARus) evaluation provides researchers with a tool for assessing\nprogress in open-domain commonsense causal reasoning. Each question in PARus is composed of a premise and two\nalternatives, where the task is to select the alternative that more plausibly has a causal relation with the premise.\nThe correct alternative is randomized so that the expected performance of randomly guessing is 50%.\n", "citation": "\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/PARus", "license": "", "features": {"premise": {"dtype": "string", "id": null, "_type": "Value"}, "choice1": {"dtype": "string", "id": null, "_type": "Value"}, "choice2": {"dtype": "string", "id": null, "_type": "Value"}, "question": {"dtype": "string", "id": null, "_type": "Value"}, "idx": {"dtype": "int32", "id": null, "_type": "Value"}, "label": {"num_classes": 2, "names": ["choice1", "choice2"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "parus", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"train": {"name": "train", "num_bytes": 74467, "num_examples": 400, "dataset_name": "russian_super_glue"}, "validation": {"name": "validation", "num_bytes": 19397, "num_examples": 100, "dataset_name": "russian_super_glue"}, "test": {"name": "test", "num_bytes": 93192, "num_examples": 500, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/PARus": {"num_bytes": 57585, "checksum": "7093c859a6ab07eab54c86cdd686b7b14afffee46abcd3e2d43876b1fdb8fa59"}}, "download_size": 57585, "post_processing_size": null, "dataset_size": 187056, "size_in_bytes": 244641}, "muserc": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\nWe present a reading comprehension challenge in which questions can only be answered by taking into account information\nfrom multiple sentences. The dataset is the first to study multi-sentence inference at scale, with an open-ended set of\nquestion types that requires reasoning skills.\n", "citation": "@inproceedings{fenogenova-etal-2020-read,\n    title = \"Read and Reason with {M}u{S}e{RC} and {R}u{C}o{S}: Datasets for Machine Reading Comprehension for {R}ussian\",\n    author = \"Fenogenova, Alena  and\n      Mikhailov, Vladislav  and\n      Shevelev, Denis\",\n    booktitle = \"Proceedings of the 28th International Conference on Computational Linguistics\",\n    month = dec,\n    year = \"2020\",\n    address = \"Barcelona, Spain (Online)\",\n    publisher = \"International Committee on Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.coling-main.570\",\n    doi = \"10.18653/v1/2020.coling-main.570\",\n    pages = \"6481--6497\",\n    abstract = \"The paper introduces two Russian machine reading comprehension (MRC) datasets, called MuSeRC and RuCoS,\n    which require reasoning over multiple sentences and commonsense knowledge to infer the answer. The former follows\n    the design of MultiRC, while the latter is a counterpart of the ReCoRD dataset. The datasets are included\n    in RussianSuperGLUE, the Russian general language understanding benchmark. We provide a comparative analysis\n    and demonstrate that the proposed tasks are relatively more complex as compared to the original ones for English.\n    Besides, performance results of human solvers and BERT-based models show that MuSeRC and RuCoS represent a challenge\n    for recent advanced neural models. We thus hope to facilitate research in the field of MRC for Russian and prompt\n    the study of multi-hop reasoning in a cross-lingual scenario.\",\n}\n\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/MuSeRC", "license": "", "features": {"paragraph": {"dtype": "string", "id": null, "_type": "Value"}, "question": {"dtype": "string", "id": null, "_type": "Value"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}, "idx": {"paragraph": {"dtype": "int32", "id": null, "_type": "Value"}, "question": {"dtype": "int32", "id": null, "_type": "Value"}, "answer": {"dtype": "int32", "id": null, "_type": "Value"}}, "label": {"num_classes": 2, "names": ["False", "True"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "muserc", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"train": {"name": "train", "num_bytes": 31651155, "num_examples": 11950, "dataset_name": "russian_super_glue"}, "validation": {"name": "validation", "num_bytes": 5964157, "num_examples": 2235, "dataset_name": "russian_super_glue"}, "test": {"name": "test", "num_bytes": 19850930, "num_examples": 7614, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/MuSeRC": {"num_bytes": 1196720, "checksum": "679d69df99113d8ca6416eda5583a81ab324f5438097216a416d09e745d2f668"}}, "download_size": 1196720, "post_processing_size": null, "dataset_size": 57466242, "size_in_bytes": 58662962}, "terra": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\nTextual Entailment Recognition has been proposed recently as a generic task that captures major semantic inference\nneeds across many NLP applications, such as Question Answering, Information Retrieval, Information Extraction,\nand Text Summarization. This task requires to recognize, given two text fragments, whether the meaning of one text is\nentailed (can be inferred) from the other text.\n", "citation": "\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/TERRa", "license": "", "features": {"premise": {"dtype": "string", "id": null, "_type": "Value"}, "hypothesis": {"dtype": "string", "id": null, "_type": "Value"}, "idx": {"dtype": "int32", "id": null, "_type": "Value"}, "label": {"num_classes": 2, "names": ["entailment", "not_entailment"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "terra", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"train": {"name": "train", "num_bytes": 1409243, "num_examples": 2616, "dataset_name": "russian_super_glue"}, "validation": {"name": "validation", "num_bytes": 161485, "num_examples": 307, "dataset_name": "russian_super_glue"}, "test": {"name": "test", "num_bytes": 1713499, "num_examples": 3198, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/TERRa": {"num_bytes": 907346, "checksum": "fc7320210b5b6f7087615f13558868de55f46d5e0e365d9d82968c66e6e0dba7"}}, "download_size": 907346, "post_processing_size": null, "dataset_size": 3284227, "size_in_bytes": 4191573}, "russe": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\nWiC: The Word-in-Context Dataset A reliable benchmark for the evaluation of context-sensitive word embeddings.\nDepending on its context, an ambiguous word can refer to multiple, potentially unrelated, meanings. Mainstream static\nword embeddings, such as Word2vec and GloVe, are unable to reflect this dynamic semantic nature. Contextualised word\nembeddings are an attempt at addressing this limitation by computing dynamic representations for words which can adapt\nbased on context.\nRussian SuperGLUE task borrows original data from the Russe project, Word Sense Induction and Disambiguation\nshared task (2018)\n", "citation": "@inproceedings{RUSSE2018,\n  author    = {Panchenko, Alexander and Lopukhina, Anastasia and Ustalov, Dmitry and Lopukhin, Konstantin and Arefyev,\n               Nikolay and Leontyev, Alexey and Loukachevitch, Natalia},\n  title     = {{RUSSE'2018: A Shared Task on Word Sense Induction for the Russian Language}},\n  booktitle = {Computational Linguistics and Intellectual Technologies:\n               Papers from the Annual International Conference ``Dialogue''},\n  year      = {2018},\n  pages     = {547--564},\n  url       = {http://www.dialog-21.ru/media/4539/panchenkoaplusetal.pdf},\n  address   = {Moscow, Russia},\n  publisher = {RSUH},\n  issn      = {2221-7932},\n  language  = {english},\n}\n\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/RUSSE", "license": "", "features": {"word": {"dtype": "string", "id": null, "_type": "Value"}, "sentence1": {"dtype": "string", "id": null, "_type": "Value"}, "sentence2": {"dtype": "string", "id": null, "_type": "Value"}, "start1": {"dtype": "int32", "id": null, "_type": "Value"}, "start2": {"dtype": "int32", "id": null, "_type": "Value"}, "end1": {"dtype": "int32", "id": null, "_type": "Value"}, "end2": {"dtype": "int32", "id": null, "_type": "Value"}, "gold_sense1": {"dtype": "int32", "id": null, "_type": "Value"}, "gold_sense2": {"dtype": "int32", "id": null, "_type": "Value"}, "idx": {"dtype": "int32", "id": null, "_type": "Value"}, "label": {"num_classes": 2, "names": ["False", "True"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "russe", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"train": {"name": "train", "num_bytes": 6913280, "num_examples": 19845, "dataset_name": "russian_super_glue"}, "validation": {"name": "validation", "num_bytes": 2957491, "num_examples": 8505, "dataset_name": "russian_super_glue"}, "test": {"name": "test", "num_bytes": 10046000, "num_examples": 18892, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/RUSSE": {"num_bytes": 3806009, "checksum": "60ecf42ea0f3893e857e0a9522ab92a2ae2ec713d1ab361f2f6f594d0f5324a5"}}, "download_size": 3806009, "post_processing_size": null, "dataset_size": 19916771, "size_in_bytes": 23722780}, "rwsd": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\nA Winograd schema is a pair of sentences that differ in only one or two words and that contain an ambiguity that is\nresolved in opposite ways in the two sentences and requires the use of world knowledge and reasoning for its resolution.\nThe schema takes its name from a well-known example by Terry Winograd.\nThe set would then be presented as a challenge for AI programs, along the lines of the Turing test. The strengths of\nthe challenge are that it is clear-cut, in that the answer to each schema is a binary choice; vivid, in that it is\nobvious to non-experts that a program that fails to get the right answers clearly has serious gaps in its understanding;\nand difficult, in that it is far beyond the current state of the art.\n", "citation": "\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/RWSD", "license": "", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}, "span1_index": {"dtype": "int32", "id": null, "_type": "Value"}, "span2_index": {"dtype": "int32", "id": null, "_type": "Value"}, "span1_text": {"dtype": "string", "id": null, "_type": "Value"}, "span2_text": {"dtype": "string", "id": null, "_type": "Value"}, "idx": {"dtype": "int32", "id": null, "_type": "Value"}, "label": {"num_classes": 2, "names": ["False", "True"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "rwsd", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"train": {"name": "train", "num_bytes": 132274, "num_examples": 606, "dataset_name": "russian_super_glue"}, "validation": {"name": "validation", "num_bytes": 87959, "num_examples": 204, "dataset_name": "russian_super_glue"}, "test": {"name": "test", "num_bytes": 59051, "num_examples": 154, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/RWSD": {"num_bytes": 40508, "checksum": "65894cf114f022a0469bbd535f045d50880e03aa822cd1f3693a54b3665fa962"}}, "download_size": 40508, "post_processing_size": null, "dataset_size": 279284, "size_in_bytes": 319792}, "danetqa": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\nDaNetQA is a question answering dataset for yes/no questions. These questions are naturally occurring -- they are\ngenerated in unprompted and unconstrained settings.\n\nEach example is a triplet of (question, passage, answer), with the title of the page as optional additional context.\nThe text-pair classification setup is similar to existing natural language inference tasks.\n\nBy sampling questions from a distribution of information-seeking queries (rather than prompting annotators for\ntext pairs), we observe significantly more challenging examples compared to existing NLI datasets.\n", "citation": "@InProceedings{10.1007/978-3-030-72610-2_4,\nauthor=\"Glushkova, Taisia\nand Machnev, Alexey\nand Fenogenova, Alena\nand Shavrina, Tatiana\nand Artemova, Ekaterina\nand Ignatov, Dmitry I.\",\neditor=\"van der Aalst, Wil M. P.\nand Batagelj, Vladimir\nand Ignatov, Dmitry I.\nand Khachay, Michael\nand Koltsova, Olessia\nand Kutuzov, Andrey\nand Kuznetsov, Sergei O.\nand Lomazova, Irina A.\nand Loukachevitch, Natalia\nand Napoli, Amedeo\nand Panchenko, Alexander\nand Pardalos, Panos M.\nand Pelillo, Marcello\nand Savchenko, Andrey V.\nand Tutubalina, Elena\",\ntitle=\"DaNetQA: A Yes/No Question Answering Dataset for the Russian Language\",\nbooktitle=\"Analysis of Images, Social Networks and Texts\",\nyear=\"2021\",\npublisher=\"Springer International Publishing\",\naddress=\"Cham\",\npages=\"57--68\",\nabstract=\"DaNetQA, a new question-answering corpus, follows BoolQ\u00a0[2] design: it comprises natural yes/no questions.\nEach question is paired with a paragraph from Wikipedia and an answer, derived from the paragraph. The task is to take\nboth the question and a paragraph as input and come up with a yes/no answer, i.e. to produce a binary output. In this\npaper, we present a reproducible approach to DaNetQA creation and investigate transfer learning methods for task and\nlanguage transferring. For task transferring we leverage three similar sentence modelling tasks: 1) a corpus of\nparaphrases, Paraphraser, 2) an NLI task, for which we use the Russian part of XNLI, 3) another question answering task,\nSberQUAD. For language transferring we use English to Russian translation together\nwith multilingual language fine-tuning.\",\nisbn=\"978-3-030-72610-2\"\n}\n\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/DaNetQA", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "passage": {"dtype": "string", "id": null, "_type": "Value"}, "idx": {"dtype": "int32", "id": null, "_type": "Value"}, "label": {"num_classes": 2, "names": ["False", "True"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "danetqa", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"train": {"name": "train", "num_bytes": 2474006, "num_examples": 1749, "dataset_name": "russian_super_glue"}, "validation": {"name": "validation", "num_bytes": 1076455, "num_examples": 821, "dataset_name": "russian_super_glue"}, "test": {"name": "test", "num_bytes": 1023062, "num_examples": 805, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/DaNetQA": {"num_bytes": 1293761, "checksum": "b5b4bcfe17e1eb16aa13a7aab4ca088871e27b0851468e9a07b9b528bb42fb96"}}, "download_size": 1293761, "post_processing_size": null, "dataset_size": 4573523, "size_in_bytes": 5867284}, "rucos": {"description": "Recent advances in the field of universal language models and transformers require the development of a methodology for\ntheir broad diagnostics and testing for general intellectual skills - detection of natural language inference,\ncommonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first\ntime, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from\nscratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating\nmodels and an overall leaderboard of transformer models for the Russian language.\nRussian reading comprehension with Commonsense reasoning (RuCoS) is a large-scale reading comprehension dataset which\nrequires commonsense reasoning. RuCoS consists of queries automatically generated from CNN/Daily Mail news articles;\nthe answer to each query is a text span from a summarizing passage of the corresponding news. The goal of RuCoS is to\nevaluate a machine`s ability of commonsense reasoning in reading comprehension.\n", "citation": "@inproceedings{fenogenova-etal-2020-read,\n    title = \"Read and Reason with {M}u{S}e{RC} and {R}u{C}o{S}: Datasets for Machine Reading Comprehension for {R}ussian\",\n    author = \"Fenogenova, Alena  and\n      Mikhailov, Vladislav  and\n      Shevelev, Denis\",\n    booktitle = \"Proceedings of the 28th International Conference on Computational Linguistics\",\n    month = dec,\n    year = \"2020\",\n    address = \"Barcelona, Spain (Online)\",\n    publisher = \"International Committee on Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.coling-main.570\",\n    doi = \"10.18653/v1/2020.coling-main.570\",\n    pages = \"6481--6497\",\n    abstract = \"The paper introduces two Russian machine reading comprehension (MRC) datasets, called MuSeRC and RuCoS,\n    which require reasoning over multiple sentences and commonsense knowledge to infer the answer. The former follows\n    the design of MultiRC, while the latter is a counterpart of the ReCoRD dataset. The datasets are included\n    in RussianSuperGLUE, the Russian general language understanding benchmark. We provide a comparative analysis\n    and demonstrate that the proposed tasks are relatively more complex as compared to the original ones for English.\n    Besides, performance results of human solvers and BERT-based models show that MuSeRC and RuCoS represent a challenge\n    for recent advanced neural models. We thus hope to facilitate research in the field of MRC for Russian and prompt\n    the study of multi-hop reasoning in a cross-lingual scenario.\",\n}\n\n@article{shavrina2020russiansuperglue,\n                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},\n                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,\n                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and\n                  Evlampiev, Andrey},\n                  journal={arXiv preprint arXiv:2010.15925},\n                  year={2020}\n                  }\n", "homepage": "https://russiansuperglue.com/tasks/task_info/RuCoS", "license": "", "features": {"passage": {"dtype": "string", "id": null, "_type": "Value"}, "query": {"dtype": "string", "id": null, "_type": "Value"}, "entities": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answers": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "idx": {"passage": {"dtype": "int32", "id": null, "_type": "Value"}, "query": {"dtype": "int32", "id": null, "_type": "Value"}}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "russian_super_glue", "config_name": "rucos", "version": {"version_str": "0.0.1", "description": null, "major": 0, "minor": 0, "patch": 1}, "splits": {"train": {"name": "train", "num_bytes": 160095378, "num_examples": 72193, "dataset_name": "russian_super_glue"}, "validation": {"name": "validation", "num_bytes": 16980563, "num_examples": 7577, "dataset_name": "russian_super_glue"}, "test": {"name": "test", "num_bytes": 15535209, "num_examples": 7257, "dataset_name": "russian_super_glue"}}, "download_checksums": {"https://russiansuperglue.com/tasks/download/RuCoS": {"num_bytes": 56208297, "checksum": "e2f42700122e79cfcce792b54df792630f033eb21e14863cede5852f3aa0d078"}}, "download_size": 56208297, "post_processing_size": null, "dataset_size": 192611150, "size_in_bytes": 248819447}}
\ No newline at end of file
diff --git a/datasets/russian_super_glue/dummy/danetqa/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/danetqa/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..0e11ca4c16b
Binary files /dev/null and b/datasets/russian_super_glue/dummy/danetqa/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/dummy/lidirus/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/lidirus/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..e790abd4f76
Binary files /dev/null and b/datasets/russian_super_glue/dummy/lidirus/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/dummy/muserc/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/muserc/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..7bd7c6077ce
Binary files /dev/null and b/datasets/russian_super_glue/dummy/muserc/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/dummy/parus/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/parus/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..ad69d61c070
Binary files /dev/null and b/datasets/russian_super_glue/dummy/parus/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/dummy/rcb/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/rcb/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..51d9aeffb9e
Binary files /dev/null and b/datasets/russian_super_glue/dummy/rcb/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/dummy/rucos/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/rucos/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..688ab8fd65f
Binary files /dev/null and b/datasets/russian_super_glue/dummy/rucos/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/dummy/russe/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/russe/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..0001a0f8575
Binary files /dev/null and b/datasets/russian_super_glue/dummy/russe/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/dummy/rwsd/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/rwsd/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..33852ff684e
Binary files /dev/null and b/datasets/russian_super_glue/dummy/rwsd/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/dummy/terra/0.0.1/dummy_data.zip b/datasets/russian_super_glue/dummy/terra/0.0.1/dummy_data.zip
new file mode 100644
index 00000000000..0893e0f622c
Binary files /dev/null and b/datasets/russian_super_glue/dummy/terra/0.0.1/dummy_data.zip differ
diff --git a/datasets/russian_super_glue/russian_super_glue.py b/datasets/russian_super_glue/russian_super_glue.py
new file mode 100644
index 00000000000..8daf3cc5557
--- /dev/null
+++ b/datasets/russian_super_glue/russian_super_glue.py
@@ -0,0 +1,531 @@
+# coding=utf-8
+# Copyright 2020 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# Lint as: python3
+"""The Russian SuperGLUE Benchmark"""
+
+import json
+import os
+from typing import List, Union
+
+import datasets
+
+
+_RUSSIAN_SUPER_GLUE_CITATION = """\
+@article{shavrina2020russiansuperglue,
+                  title={RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark},
+                  author={Shavrina, Tatiana and Fenogenova, Alena and Emelyanov, Anton and Shevelev, Denis and Artemova,
+                  Ekaterina and Malykh, Valentin and Mikhailov, Vladislav and Tikhonova, Maria and Chertok, Andrey and
+                  Evlampiev, Andrey},
+                  journal={arXiv preprint arXiv:2010.15925},
+                  year={2020}
+                  }
+"""
+
+_MUSERC_CITATION = """\
+@inproceedings{fenogenova-etal-2020-read,
+    title = "Read and Reason with {M}u{S}e{RC} and {R}u{C}o{S}: Datasets for Machine Reading Comprehension for {R}ussian",
+    author = "Fenogenova, Alena  and
+      Mikhailov, Vladislav  and
+      Shevelev, Denis",
+    booktitle = "Proceedings of the 28th International Conference on Computational Linguistics",
+    month = dec,
+    year = "2020",
+    address = "Barcelona, Spain (Online)",
+    publisher = "International Committee on Computational Linguistics",
+    url = "https://aclanthology.org/2020.coling-main.570",
+    doi = "10.18653/v1/2020.coling-main.570",
+    pages = "6481--6497",
+    abstract = "The paper introduces two Russian machine reading comprehension (MRC) datasets, called MuSeRC and RuCoS,
+    which require reasoning over multiple sentences and commonsense knowledge to infer the answer. The former follows
+    the design of MultiRC, while the latter is a counterpart of the ReCoRD dataset. The datasets are included
+    in RussianSuperGLUE, the Russian general language understanding benchmark. We provide a comparative analysis
+    and demonstrate that the proposed tasks are relatively more complex as compared to the original ones for English.
+    Besides, performance results of human solvers and BERT-based models show that MuSeRC and RuCoS represent a challenge
+    for recent advanced neural models. We thus hope to facilitate research in the field of MRC for Russian and prompt
+    the study of multi-hop reasoning in a cross-lingual scenario.",
+}
+"""
+
+_RUSSE_CITATION = """\
+@inproceedings{RUSSE2018,
+  author    = {Panchenko, Alexander and Lopukhina, Anastasia and Ustalov, Dmitry and Lopukhin, Konstantin and Arefyev,
+               Nikolay and Leontyev, Alexey and Loukachevitch, Natalia},
+  title     = {{RUSSE'2018: A Shared Task on Word Sense Induction for the Russian Language}},
+  booktitle = {Computational Linguistics and Intellectual Technologies:
+               Papers from the Annual International Conference ``Dialogue''},
+  year      = {2018},
+  pages     = {547--564},
+  url       = {http://www.dialog-21.ru/media/4539/panchenkoaplusetal.pdf},
+  address   = {Moscow, Russia},
+  publisher = {RSUH},
+  issn      = {2221-7932},
+  language  = {english},
+}
+"""
+
+_DANETQA_CITATION = """\
+@InProceedings{10.1007/978-3-030-72610-2_4,
+author="Glushkova, Taisia
+and Machnev, Alexey
+and Fenogenova, Alena
+and Shavrina, Tatiana
+and Artemova, Ekaterina
+and Ignatov, Dmitry I.",
+editor="van der Aalst, Wil M. P.
+and Batagelj, Vladimir
+and Ignatov, Dmitry I.
+and Khachay, Michael
+and Koltsova, Olessia
+and Kutuzov, Andrey
+and Kuznetsov, Sergei O.
+and Lomazova, Irina A.
+and Loukachevitch, Natalia
+and Napoli, Amedeo
+and Panchenko, Alexander
+and Pardalos, Panos M.
+and Pelillo, Marcello
+and Savchenko, Andrey V.
+and Tutubalina, Elena",
+title="DaNetQA: A Yes/No Question Answering Dataset for the Russian Language",
+booktitle="Analysis of Images, Social Networks and Texts",
+year="2021",
+publisher="Springer International Publishing",
+address="Cham",
+pages="57--68",
+abstract="DaNetQA, a new question-answering corpus, follows BoolQ [2] design: it comprises natural yes/no questions.
+Each question is paired with a paragraph from Wikipedia and an answer, derived from the paragraph. The task is to take
+both the question and a paragraph as input and come up with a yes/no answer, i.e. to produce a binary output. In this
+paper, we present a reproducible approach to DaNetQA creation and investigate transfer learning methods for task and
+language transferring. For task transferring we leverage three similar sentence modelling tasks: 1) a corpus of
+paraphrases, Paraphraser, 2) an NLI task, for which we use the Russian part of XNLI, 3) another question answering task,
+SberQUAD. For language transferring we use English to Russian translation together
+with multilingual language fine-tuning.",
+isbn="978-3-030-72610-2"
+}
+"""
+
+_RUCOS_CITATION = _MUSERC_CITATION
+
+_RUSSIAN_SUPER_GLUE_DESCRIPTION = """\
+Recent advances in the field of universal language models and transformers require the development of a methodology for
+their broad diagnostics and testing for general intellectual skills - detection of natural language inference,
+commonsense reasoning, ability to perform simple logical operations regardless of text subject or lexicon. For the first
+time, a benchmark of nine tasks, collected and organized analogically to the SuperGLUE methodology, was developed from
+scratch for the Russian language. We provide baselines, human level evaluation, an open-source framework for evaluating
+models and an overall leaderboard of transformer models for the Russian language.
+"""
+
+_HOMEPAGE = "https://russiansuperglue.com/"
+
+_LICENSE = "MIT License"
+
+_LIDIRUS_DESCRIPTION = """"\
+LiDiRus (Linguistic Diagnostic for Russian) is a diagnostic dataset that covers a large volume of linguistic phenomena,
+while allowing you to evaluate information systems on a simple test of textual entailment recognition.
+See more details diagnostics.
+"""
+
+_RCB_DESCRIPTION = """\
+The Russian Commitment Bank is a corpus of naturally occurring discourses whose final sentence contains
+a clause-embedding predicate under an entailment canceling operator (question, modal, negation, antecedent
+of conditional).
+"""
+
+_PARUS_DESCRIPTION = """\
+Choice of Plausible Alternatives for Russian language
+Choice of Plausible Alternatives for Russian language (PARus) evaluation provides researchers with a tool for assessing
+progress in open-domain commonsense causal reasoning. Each question in PARus is composed of a premise and two
+alternatives, where the task is to select the alternative that more plausibly has a causal relation with the premise.
+The correct alternative is randomized so that the expected performance of randomly guessing is 50%.
+"""
+
+_MUSERC_DESCRIPTION = """\
+We present a reading comprehension challenge in which questions can only be answered by taking into account information
+from multiple sentences. The dataset is the first to study multi-sentence inference at scale, with an open-ended set of
+question types that requires reasoning skills.
+"""
+
+_TERRA_DESCRIPTION = """\
+Textual Entailment Recognition has been proposed recently as a generic task that captures major semantic inference
+needs across many NLP applications, such as Question Answering, Information Retrieval, Information Extraction,
+and Text Summarization. This task requires to recognize, given two text fragments, whether the meaning of one text is
+entailed (can be inferred) from the other text.
+"""
+
+_RUSSE_DESCRIPTION = """\
+WiC: The Word-in-Context Dataset A reliable benchmark for the evaluation of context-sensitive word embeddings.
+Depending on its context, an ambiguous word can refer to multiple, potentially unrelated, meanings. Mainstream static
+word embeddings, such as Word2vec and GloVe, are unable to reflect this dynamic semantic nature. Contextualised word
+embeddings are an attempt at addressing this limitation by computing dynamic representations for words which can adapt
+based on context.
+Russian SuperGLUE task borrows original data from the Russe project, Word Sense Induction and Disambiguation
+shared task (2018)
+"""
+
+_RWSD_DESCRIPTION = """\
+A Winograd schema is a pair of sentences that differ in only one or two words and that contain an ambiguity that is
+resolved in opposite ways in the two sentences and requires the use of world knowledge and reasoning for its resolution.
+The schema takes its name from a well-known example by Terry Winograd.
+The set would then be presented as a challenge for AI programs, along the lines of the Turing test. The strengths of
+the challenge are that it is clear-cut, in that the answer to each schema is a binary choice; vivid, in that it is
+obvious to non-experts that a program that fails to get the right answers clearly has serious gaps in its understanding;
+and difficult, in that it is far beyond the current state of the art.
+"""
+
+_DANETQA_DESCRIPTION = """\
+DaNetQA is a question answering dataset for yes/no questions. These questions are naturally occurring -- they are
+generated in unprompted and unconstrained settings.
+
+Each example is a triplet of (question, passage, answer), with the title of the page as optional additional context.
+The text-pair classification setup is similar to existing natural language inference tasks.
+
+By sampling questions from a distribution of information-seeking queries (rather than prompting annotators for
+text pairs), we observe significantly more challenging examples compared to existing NLI datasets.
+"""
+
+_RUCOS_DESCRIPTION = """\
+Russian reading comprehension with Commonsense reasoning (RuCoS) is a large-scale reading comprehension dataset which
+requires commonsense reasoning. RuCoS consists of queries automatically generated from CNN/Daily Mail news articles;
+the answer to each query is a text span from a summarizing passage of the corresponding news. The goal of RuCoS is to
+evaluate a machine`s ability of commonsense reasoning in reading comprehension.
+"""
+
+
+class RussianSuperGlueConfig(datasets.BuilderConfig):
+    """BuilderConfig for the Russian SuperGLUE."""
+
+    VERSION = datasets.Version("0.0.1")
+
+    def __init__(
+        self,
+        features: List[str],
+        data_url: str,
+        citation: str,
+        url: str,
+        label_classes: List[str] = ("False", "True"),
+        **kwargs,
+    ):
+        """BuilderConfig for the Russian SuperGLUE.
+
+        Args:
+          features: `list[string]`, list of the features that will appear in the
+            feature dict. Should not include "label".
+          data_url: `string`, url to download the zip file from.
+          citation: `string`, citation for the data set.
+          url: `string`, url for information about the data set.
+          label_classes: `list[string]`, the list of classes for the label if the
+            label is present as a string. Non-string labels will be cast to either
+            'False' or 'True'.
+          **kwargs: keyword arguments forwarded to super.
+        """
+        # 0.0.1: Initial version.
+        super(RussianSuperGlueConfig, self).__init__(version=self.VERSION, **kwargs)
+        self.features = features
+        self.label_classes = label_classes
+        self.data_url = data_url
+        self.citation = citation
+        self.url = url
+
+
+class RussianSuperGlue(datasets.GeneratorBasedBuilder):
+
+    BUILDER_CONFIGS = [
+        RussianSuperGlueConfig(
+            name="lidirus",
+            description=_LIDIRUS_DESCRIPTION,
+            features=[
+                "sentence1",
+                "sentence2",
+                "knowledge",
+                "lexical-semantics",
+                "logic",
+                "predicate-argument-structure",
+            ],
+            label_classes=["entailment", "not_entailment"],
+            data_url="https://russiansuperglue.com/tasks/download/LiDiRus",
+            citation="",
+            url="https://russiansuperglue.com/tasks/task_info/LiDiRus",
+        ),
+        RussianSuperGlueConfig(
+            name="rcb",
+            description=_RCB_DESCRIPTION,
+            features=["premise", "hypothesis", "verb", "negation"],
+            label_classes=["entailment", "contradiction", "neutral"],
+            data_url="https://russiansuperglue.com/tasks/download/RCB",
+            citation="",
+            url="https://russiansuperglue.com/tasks/task_info/RCB",
+        ),
+        RussianSuperGlueConfig(
+            name="parus",
+            description=_PARUS_DESCRIPTION,
+            label_classes=["choice1", "choice2"],
+            features=["premise", "choice1", "choice2", "question"],
+            data_url="https://russiansuperglue.com/tasks/download/PARus",
+            citation="",
+            url="https://russiansuperglue.com/tasks/task_info/PARus",
+        ),
+        RussianSuperGlueConfig(
+            name="muserc",
+            description=_MUSERC_DESCRIPTION,
+            features=["paragraph", "question", "answer"],
+            data_url="https://russiansuperglue.com/tasks/download/MuSeRC",
+            citation=_MUSERC_CITATION,
+            label_classes=["False", "True"],
+            url="https://russiansuperglue.com/tasks/task_info/MuSeRC",
+        ),
+        RussianSuperGlueConfig(
+            name="terra",
+            description=_TERRA_DESCRIPTION,
+            features=["premise", "hypothesis"],
+            label_classes=["entailment", "not_entailment"],
+            data_url="https://russiansuperglue.com/tasks/download/TERRa",
+            citation="",
+            url="https://russiansuperglue.com/tasks/task_info/TERRa",
+        ),
+        RussianSuperGlueConfig(
+            name="russe",
+            description=_RUSSE_DESCRIPTION,
+            features=[
+                "word",
+                "sentence1",
+                "sentence2",
+                "start1",
+                "start2",
+                "end1",
+                "end2",
+                "gold_sense1",
+                "gold_sense2",
+            ],
+            data_url="https://russiansuperglue.com/tasks/download/RUSSE",
+            citation=_RUSSE_CITATION,
+            label_classes=["False", "True"],
+            url="https://russiansuperglue.com/tasks/task_info/RUSSE",
+        ),
+        RussianSuperGlueConfig(
+            name="rwsd",
+            description=_RWSD_DESCRIPTION,
+            features=["text", "span1_index", "span2_index", "span1_text", "span2_text"],
+            data_url="https://russiansuperglue.com/tasks/download/RWSD",
+            citation="",
+            label_classes=["False", "True"],
+            url="https://russiansuperglue.com/tasks/task_info/RWSD",
+        ),
+        RussianSuperGlueConfig(
+            name="danetqa",
+            description=_DANETQA_DESCRIPTION,
+            features=["question", "passage"],
+            data_url="https://russiansuperglue.com/tasks/download/DaNetQA",
+            citation=_DANETQA_CITATION,
+            label_classes=["False", "True"],
+            url="https://russiansuperglue.com/tasks/task_info/DaNetQA",
+        ),
+        RussianSuperGlueConfig(
+            name="rucos",
+            description=_RUCOS_DESCRIPTION,
+            features=["passage", "query", "entities", "answers"],
+            data_url="https://russiansuperglue.com/tasks/download/RuCoS",
+            citation=_RUCOS_CITATION,
+            url="https://russiansuperglue.com/tasks/task_info/RuCoS",
+        ),
+    ]
+
+    def _info(self):
+
+        if self.config.name == "russe":
+            features = {feature: datasets.Value("string") for feature in ("word", "sentence1", "sentence2")}
+            features["start1"] = datasets.Value("int32")
+            features["start2"] = datasets.Value("int32")
+            features["end1"] = datasets.Value("int32")
+            features["end2"] = datasets.Value("int32")
+            features["gold_sense1"] = datasets.Value("int32")
+            features["gold_sense2"] = datasets.Value("int32")
+
+        else:
+            features = {feature: datasets.Value("string") for feature in self.config.features}
+
+        if self.config.name == "rwsd":
+            features["span1_index"] = datasets.Value("int32")
+            features["span2_index"] = datasets.Value("int32")
+
+        if self.config.name == "muserc":
+            features["idx"] = dict(
+                {
+                    "paragraph": datasets.Value("int32"),
+                    "question": datasets.Value("int32"),
+                    "answer": datasets.Value("int32"),
+                }
+            )
+        elif self.config.name == "rucos":
+            features["idx"] = dict(
+                {
+                    "passage": datasets.Value("int32"),
+                    "query": datasets.Value("int32"),
+                }
+            )
+        else:
+            features["idx"] = datasets.Value("int32")
+
+        if self.config.name == "rucos":
+            # Entities are the set of possible choices for the placeholder.
+            features["entities"] = datasets.features.Sequence(datasets.Value("string"))
+            # Answers are the subset of entities that are correct.
+            features["answers"] = datasets.features.Sequence(datasets.Value("string"))
+        else:
+            features["label"] = datasets.features.ClassLabel(names=self.config.label_classes)
+
+        return datasets.DatasetInfo(
+            description=_RUSSIAN_SUPER_GLUE_DESCRIPTION + self.config.description,
+            features=datasets.Features(features),
+            homepage=self.config.url,
+            citation=self.config.citation + "\n" + _RUSSIAN_SUPER_GLUE_CITATION,
+        )
+
+    def _split_generators(self, dl_manager: datasets.DownloadManager):
+        dl_dir = dl_manager.download_and_extract(self.config.data_url) or ""
+        task_name = _get_task_name_from_data_url(self.config.data_url)
+        dl_dir = os.path.join(dl_dir, task_name)
+        if self.config.name == "lidirus":
+            return [
+                datasets.SplitGenerator(
+                    name=datasets.Split.TEST,
+                    gen_kwargs={
+                        "data_file": os.path.join(dl_dir, f"{task_name}.jsonl"),
+                        "split": datasets.Split.TEST,
+                    },
+                ),
+            ]
+        else:
+            return [
+                datasets.SplitGenerator(
+                    name=datasets.Split.TRAIN,
+                    gen_kwargs={
+                        "data_file": os.path.join(dl_dir, "train.jsonl"),
+                        "split": datasets.Split.TRAIN,
+                    },
+                ),
+                datasets.SplitGenerator(
+                    name=datasets.Split.VALIDATION,
+                    gen_kwargs={
+                        "data_file": os.path.join(dl_dir, "val.jsonl"),
+                        "split": datasets.Split.VALIDATION,
+                    },
+                ),
+                datasets.SplitGenerator(
+                    name=datasets.Split.TEST,
+                    gen_kwargs={
+                        "data_file": os.path.join(dl_dir, "test.jsonl"),
+                        "split": datasets.Split.TEST,
+                    },
+                ),
+            ]
+
+    def _generate_examples(self, data_file: str, split: datasets.Split):
+        with open(data_file, encoding="utf-8") as file:
+            for line in file:
+                row = json.loads(line)
+
+                if self.config.name == "muserc":
+
+                    paragraph = row["passage"]
+                    for question in paragraph["questions"]:
+                        for answer in question["answers"]:
+                            label = answer.get("label")
+                            key = "%s_%s_%s" % (row["idx"], question["idx"], answer["idx"])
+                            yield key, {
+                                "paragraph": paragraph["text"],
+                                "question": question["question"],
+                                "answer": answer["text"],
+                                "label": -1 if label is None else _cast_label(bool(label)),
+                                "idx": {"paragraph": row["idx"], "question": question["idx"], "answer": answer["idx"]},
+                            }
+
+                elif self.config.name == "rucos":
+                    passage = row["passage"]
+                    for qa in row["qas"]:
+                        yield qa["idx"], {
+                            "passage": passage["text"],
+                            "query": qa["query"],
+                            "entities": _get_rucos_entities(passage),
+                            "answers": _get_rucos_answers(qa),
+                            "idx": {"passage": row["idx"], "query": qa["idx"]},
+                        }
+                else:
+                    if self.config.name in ("lidirus", "rcb"):
+                        # features may be missing
+                        example = {feature: row.get(feature, "") for feature in self.config.features}
+                    elif self.config.name == "russe" and split == datasets.Split.TEST:
+                        # gold senses are not available in `test` split
+                        example = {
+                            feature: row[feature]
+                            for feature in self.config.features
+                            if feature not in ("gold_sense1", "gold_sense2")
+                        }
+                        example["gold_sense1"] = -1
+                        example["gold_sense2"] = -1
+                    else:
+                        if self.config.name == "rwsd":
+                            row.update(row["target"])
+
+                        example = {feature: row[feature] for feature in self.config.features}
+
+                    example["idx"] = row["idx"]
+
+                    if "label" in row:
+                        if self.config.name == "parus":
+                            example["label"] = "choice2" if row["label"] else "choice1"
+                        else:
+                            example["label"] = _cast_label(row["label"])
+                    else:
+                        assert split == datasets.Split.TEST, row
+                        example["label"] = -1
+
+                    yield example["idx"], example
+
+
+def _get_task_name_from_data_url(data_url: str) -> str:
+    return data_url.split("/")[-1]
+
+
+def _cast_label(label: Union[str, bool, int]) -> str:
+    """Converts the label into the appropriate string version."""
+    if isinstance(label, str):
+        return label
+    elif isinstance(label, bool):
+        return "True" if label else "False"
+    elif isinstance(label, int):
+        assert label in (0, 1)
+        return str(label)
+    else:
+        raise ValueError("Invalid label format.")
+
+
+def _get_rucos_entities(passage: dict) -> List[str]:
+    """Returns the unique set of entities."""
+    text = passage["text"]
+    entities = set()
+    for entity in passage["entities"]:
+        entities.add(text[entity["start"] : entity["end"] + 1])
+    return sorted(entities)
+
+
+def _get_rucos_answers(qa: dict) -> List[str]:
+    """Returns the unique set of answers."""
+    if "answers" not in qa:
+        return []
+    answers = set()
+    for answer in qa["answers"]:
+        answers.add(answer["text"])
+    return sorted(answers)