This is the large data set as featured in the OCR H240 exam series.

Questions about this dataset will be featured in the statistics paper

The LDS is a .xlsx file containing 5 tables, four data, one information. The data is drawn from the UK censuses from the years 2001 and 2011. It is designed for you to make comparisons and analyses of the changes in demographic and behavioural features of the populace. There is the age structure of each local authority and the method of travel within each local authority.

MLFMF MLFMF (Machine Learning for Mathematical Formalization) is a collection of data sets for benchmarking recommendation systems used to support formalization of mathematics with proof assistants. These systems help humans identify which previous entries (theorems, constructions, datatypes, and postulates) are relevant in proving a new theorem or carrying out a new construction. The MLFMF data sets provide solid benchmarking support for further investigation of the numerous machine learning approaches to formalized mathematics. With more than 250,000 entries in total, this is currently the largest collection of formalized mathematical knowledge in machine learnable format. In addition to benchmarking the recommendation systems, the data sets can also be used for benchmarking node classification and link prediction algorithms. The four data sets Each data set is derived from a library of formalized mathematics written in proof assistants Agda or Lean. The collection includes

the largest Lean 4 library Mathlib, the three largest Agda libraries:

the standard library the library of univalent mathematics Agda-unimath, and the TypeTopology library. Each data set represents the corresponding library in two ways: as a heterogeneous network, and as a list of syntax trees of all the entries in the library. The network contains the (modular) structure of the library and the references between entries, while the syntax trees give complete and easily parsed information about each entry. The Lean library data set was obtained by converting .olean files into s-expressions (see the lean2sexp tool). The Agda data sets were obtained with an s-expression extension of the official Agda repository (use either master-sexp or release-2.6.3-sexp branch). For more details, see our arXiv copy of the paper. Directory structure First, the mlfmf.zip archive needs to be unzipped. It contains a separate directory for every library (for example, the standard library of Agda can be found in the stdlib directory) and some auxiliary files. Every library directory contains

the network file from which the heterogeneous network can be loaded, a zip of the entries directory that contains (many) files with abstract syntax trees. Each of those files describes a single entry of the library. In addition to the auxiliary files which are used for loading the data (and described below), the zipped sources of lean2sexp and Agda s-expression extension are present. Loading the data In addition to the data files, there is also a simple python script main.py for loading the data. To run it, you will have to install the packages listed in the file requirements.txt: tqdm and networkx. The easiest way to do so is calling pip install -r requirements.txt. When running main.py for the first time, the script will unzip the entry files into the directory named entries. After that, the script loads the syntax trees of the entries (see the Entry class) and the network (as networkx.MultiDiGraph object). Note. The entry files have extension .dag (directed acyclic graph), since Lean uses node sharing, which breaks the tree structure (a shared node has more than one parent node). More information For more information about the data collection process, detailed data (and data format) description, and baseline experiments that were already performed with these data, see our arXiv copy of the paper. For the code that was used to perform the experiments and data format description, visit our github repository https://github.com/ul-fmf/mlfmf-data. Funding Since not all the funders are available in the Zenodo's database, we list them here:

This material is based upon work supported by the Air Force Office of Scientific Research under award number FA9550-21-1-0024. The authors also acknowledge the financial support of the Slovenian Research Agency via the research core funding No. P2-0103 and No. P1-0294.

Airoboros LLMs Math Dataset

Mastering Complex Mathematical Operations in Machine Learning

By Huggingface Hub [source]

About this dataset

The Airoboros-3.1 dataset is the perfect tool to help machine learning models excel in the difficult realm of complicated mathematical operations. This data collection features thousands of conversations between machines and humans, formatted in ShareGPT to maximize optimization in an OS ecosystem. The dataset’s focus on advanced subjects like factorials, trigonometry, and larger numerical values will help drive machine learning models to the next level - facilitating critical acquisition of sophisticated mathematical skills that are essential for ML success. As AI technology advances at such a rapid pace, training neural networks to correspondingly move forward can be a daunting and complicated challenge - but with Airoboros-3.1’s powerful datasets designed around difficult mathematical operations it just became one step closer to achievable!

More Datasets

For more datasets, click here.

Featured Notebooks

• 🚨 Your notebook can be here! 🚨!

How to use the dataset

To get started, download the dataset from Kaggle and use the train.csv file. This file contains over two thousand examples of conversations between ML models and humans which have been formatted using ShareGPT - fast and efficient OS ecosystem fine-tuning tools designed to help with understanding mathematical operations more easily. The file includes two columns: category and conversations, both of which are marked as strings in the data itself.

Once you have downloaded the train file you can begin setting up your own ML training environment by using any of your preferred frameworks or methods. Your model should focus on predicting what kind of mathematical operations will likely be involved in future conversations by referring back to previous dialogues within this dataset for reference (category column). You can also create your own test sets from this data, adding new conversation topics either by modifying existing rows or creating new ones entirely with conversation topics related to mathematics. Finally, compare your model’s results against other established models or algorithms that are already published online!

Happy training!

Research Ideas

• It can be used to build custom neural networks or machine learning algorithms that are specifically designed for complex mathematical operations.
• This data set can be used to teach and debug more general-purpose machine learning models to recognize large numbers, and intricate calculations within natural language processing (NLP).
• The Airoboros-3.1 dataset can also be utilized as a supervised learning task: models could learn from the conversations provided in the dataset how to respond correctly when presented with complex mathematical operations

Acknowledgements

If you use this dataset in your research, please credit the original authors. Data Source

License

License: CC0 1.0 Universal (CC0 1.0) - Public Domain Dedication No Copyright - You can copy, modify, distribute and perform the work, even for commercial purposes, all without asking permission. See Other Information.

Columns

File: train.csv | Column name | Description | |:------------------|:-----------------------------------------------------------------------------| | category | The type of mathematical operation being discussed. (String) | | conversations | The conversations between the machine learning model and the human. (String) |

Acknowledgements

If you use this dataset in your research, please credit the original authors. If you use this dataset in your research, please credit Huggingface Hub.

Math data for paper "Expanding RL with Verifiable Rewards Across Diverse Domains". we use a large-scale dataset of 773k Chinese Question Answering (QA) pairs, collected under authorized licenses from educational websites. This dataset covers three educational levels: elementary, middle, and high school. Unlike well-structured yet small-scale benchmarks such as MATH (Hendrycks et al., 2021b) and GSM8K (Cobbe et al., 2021b), our reference answers are inherently free-form, often interwoven with… See the full description on the dataset page: https://huggingface.co/datasets/virtuoussy/Math-RLVR.

Maths-Grade-School I am releasing large Grade School level Mathematics datatset. This extensive dataset, comprising nearly one million instructions in JSON format, encapsulates a diverse array of topics fundamental to building a strong mathematical foundation. This dataset is in instruction format so that model developers, researchers etc. can easily use this dataset. Following Fields & sub Fields are covered: Calculus Probability Algebra Liner Algebra Trigonometry Differential Equations… See the full description on the dataset page: https://huggingface.co/datasets/pt-sk/Maths-Grade-School.

MLFMF MLFMF (Machine Learning for Mathematical Formalization) is a collection of data sets for benchmarking recommendation systems used to support formalization of mathematics with proof assistants. These systems help humans identify which previous entries (theorems, constructions, datatypes, and postulates) are relevant in proving a new theorem or carrying out a new construction. The MLFMF data sets provide solid benchmarking support for further investigation of the numerous machine learning approaches to formalized mathematics. With more than 250,000 entries in total, this is currently the largest collection of formalized mathematical knowledge in machine learnable format. In addition to benchmarking the recommendation systems, the data sets can also be used for benchmarking node classification and link prediction algorithms. The four data sets Each data set is derived from a library of formalized mathematics written in proof assistants Agda or Lean. The collection includes

the largest Lean 4 library Mathlib, the three largest Agda libraries:

the standard library the library of univalent mathematics Agda-unimath, and the TypeTopology library. Each data set represents the corresponding library in two ways: as a heterogeneous network, and as a list of syntax trees of all the entries in the library. The network contains the (modular) structure of the library and the references between entries, while the syntax trees give complete and easily parsed information about each entry. The Lean library data set was obtained by converting .olean files into s-expressions (see the lean2sexp tool). The Agda data sets were obtained with an s-expression extension of the official Agda repository (use either master-sexp or release-2.6.3-sexp branch). For more details, see our arXiv copy of the paper. Directory structure First, the mlfmf.zip archive needs to be unzipped. It contains a separate directory for every library (for example, the standard library of Agda can be found in the stdlib directory) and some auxiliary files. Every library directory contains

the network file from which the heterogeneous network can be loaded, a zip of the entries directory that contains (many) files with abstract syntax trees. Each of those files describes a single entry of the library. In addition to the auxiliary files which are used for loading the data (and described below), the zipped sources of lean2sexp and Agda s-expression extension are present. Loading the data In addition to the data files, there is also a simple python script main.py for loading the data. To run it, you will have to install the packages listed in the file requirements.txt: tqdm and networkx. The easiest way to do so is calling pip install -r requirements.txt. When running main.py for the first time, the script will unzip the entry files into the directory named entries. After that, the script loads the syntax trees of the entries (see the Entry class) and the network (as networkx.MultiDiGraph object). Note. The entry files have extension .dag (directed acyclic graph), since Lean uses node sharing, which breaks the tree structure (a shared node has more than one parent node). More information For more information about the data collection process, detailed data (and data format) description, and baseline experiments that were already performed with these data, see our arXiv copy of the paper. For the code that was used to perform the experiments and data format description, visit our github repository https://github.com/ul-fmf/mlfmf-data. Funding Since not all the funders are available in the Zenodo's database, we list them here:

This material is based upon work supported by the Air Force Office of Scientific Research under award number FA9550-21-1-0024. The authors also acknowledge the financial support of the Slovenian Research Agency via the research core funding No. P2-0103 and No. P1-0294.

Supplementary Information files for A gifted SNARC? Directional spatial-numerical associations in gifted children with high-level math skills do not differ from controlsThe SNARC (Spatial-Numerical Association of Response Codes) efect (i.e., a tendency to associate small/large magnitude numbers with the left/right hand side) is prevalent across the whole lifespan. Because the ability to relate numbers to space has been viewed as a cornerstone in the development of mathematical skills, the relationship between the SNARC efect and math skills has been frequently examined. The results remain largely inconsistent. Studies testing groups of people with very low or very high skill levels in math sometimes found relationships between SNARC and math skills. So far, however, studies testing such extreme math skills level groups were mostly investigating the SNARC efect in individuals revealing math difculties. Groups with above average math skills remain understudied, especially in regard to children. Here, we investigate the SNARC efect in gifted children, as compared to normally developing children (overall n=165). Frequentist and Bayesian analysis suggested that the groups did not difer from each other in the SNARC efect. These results are the frst to provide evidence for the SNARC efect in a relatively large sample of gifted (and mathematically highly skilled) children. In sum, our study provides another piece of evidence for no direct link between the SNARC efect and mathematical ability in childhood.

InCube - Large Math Dataset

  Overview

This dataset contains over 175 million mathematical questions and their answers, designed for training and evaluating machine learning models on mathematical reasoning tasks. It spans 17 different types of mathematical operations with varying levels of complexity.

  Dataset Details





  Size

Total examples: ~175 million File format: JSON Examples per operation: ~10 million (with some variation due to mathematical… See the full description on the dataset page: https://huggingface.co/datasets/evanto/Incube-large-math-dataset.

This dataset contains 10.44 million English-language test questions parsed and structured for large-scale educational AI and NLP applications. Each question record includes the title, answer, explanation (parse), subject, grade level, and question type. Covering a full range of academic stages from primary and middle school to high school and university, the dataset spans core subjects such as English, mathematics, biology, and accounting. The content follows the Anglo-American education system and supports tasks such as question answering, subject knowledge enhancement, educational chatbot training, and intelligent tutoring systems. All data are formatted for efficient machine learning use and comply with data protection regulations including GDPR, CCPA, and PIPL.

Mathematical reasoning, a fundamental aspect of human cognition, poses significant challenges for artificial intelligence (AI) systems. Despite recent advancements in natural language processing (NLP) and large language models (LLMs), AI's ability to replicate human-like reasoning, generalization, and efficiency remains an ongoing research challenge. In this dissertation, we address key limitations in MWP solving, focusing on the accuracy, generalization ability and efficiency of AI-based mathematical reasoners by applying human-like reasoning methods and principles.

This dissertation introduces several innovative approaches in mathematical reasoning. First, a numeracy-driven framework is proposed to enhance math word problem (MWP) solvers by integrating numerical reasoning into model training, surpassing human-level performance on benchmark datasets. Second, a novel multi-solution framework captures the diversity of valid solutions to math problems, improving the generalization capabilities of AI models. Third, a customized knowledge distillation technique, termed Customized Exercise for Math Learning (CEMAL), is developed to create tailored exercises for smaller models, significantly improving their efficiency and accuracy in solving MWPs. Additionally, a multi-view fine-tuning paradigm (MinT) is introduced to enable smaller models to handle diverse annotation styles from different datasets, improving their adaptability and generalization. To further advance mathematical reasoning, a benchmark, MathChat, is introduced to evaluate large language models (LLMs) in multi-turn reasoning and instruction-following tasks, demonstrating significant performance improvements. Finally, new inference-time verifiers, Math-Rev and Code-Rev, are developed to enhance reasoning verification, combining language-based and code-based solutions for improved accuracy in both math and code reasoning tasks.

In summary, this dissertation provides a comprehensive exploration of these challenges and contributes novel solutions that push the boundaries of AI-driven mathematical reasoning. Potential future research directions are also discussed to further extend the impact of this dissertation.

We extend the analysis of a riskless choice experiment reported recently by Hochman et al. (2014). Participants select from among sets of standard playing cards valued by a simple formula. In some sessions, participants are given a prepayment associated with some of the cards, which need not be the earnings-maximizing ones. Hochman et al. (2014) find that participants choose an earnings-maximizing card less frequently when another card is prepaid. We replicate this result under the original instructions, but not with instructions which explain the payment process more explicitly. Participants who state they do not consider themselves good at mathematics make earnings-maximizing choices much less frequently overall, but those who express self-confidence in mathematics drive the treatment effect. The results suggest that even when comparisons among choices require only simple quantitative reasoning steps, market designers and regulators may need to pay close attention to how the terms of offers are expressed, explained, and implemented.

This network project brings together economists, psychologists, computer and complexity scientists from three leading centres for behavioural social science at Nottingham, Warwick and UEA. This group will lead a research programme with two broad objectives: to develop and test cross-disciplinary models of human behaviour and behaviour change; to draw out their implications for the formulation and evaluation of public policy. Foundational research will focus on three inter-related themes: understanding individual behaviour and behaviour change; understanding social and interactive behaviour; rethinking the foundations of policy analysis. The project will explore implications of the basic science for policy via a series of applied projects connecting naturally with the three themes. These will include: the determinants of consumer credit behaviour; the formation of social values; strategies for evaluation of policies affecting health and safety. The research will integrate theoretical perspectives from multiple disciplines and utilise a wide range of complementary methodologies including: theoretical modeling of individuals, groups and complex systems; conceptual analysis; lab and field experiments; analysis of large data sets. The Network will promote high quality cross-disciplinary research and serve as a policy forum for understanding behaviour and behaviour change.

Dataset Description for Filtered Sigma Dolphin Dataset

Overview

This dataset is a cleaned and filtered version of the Sigma Dolphin dataset (https://www.kaggle.com/datasets/saurabhshahane/sigmadolphin), designed to aid in solving maths word problems using AI techniques. This was used as an effort towards taking part in the AI Mathematical Olympiad - Progress Prize 1 (https://www.kaggle.com/competitions/ai-mathematical-olympiad-prize/overview). The dataset was processed using TF-IDF vectorisation and K-means clustering, specifically targeting questions relevant to the AIME (American Invitational Mathematics Examination) and AMC 12 (American Mathematics Competitions).

Context

The Sigma Dolphin dataset is a project initiated by Microsoft Research Asia, aimed at building an intelligent system with natural language understanding and reasoning capacities to automatically solve maths word problems written in natural language. This project began in early 2013, and the dataset includes maths word problems from various sources, including community question-answering sites like Yahoo! Answers.

Source and Original Dataset Details

• Original Dataset: Sigma Dolphin (https://www.kaggle.com/datasets/saurabhshahane/sigmadolphin)
• Original Source: https://msropendata.com/datasets/f0e63bb3-717a-4a53-aa79-da339b0d7992
• Project Page: http://research.microsoft.com/en-us/projects/dolphin/
• References:
  • Shuming Shi, et al. "Automatically Solving Number Word Problems by Semantic Parsing and Reasoning." EMNLP 2015.
  • Danqing Huang, et al. "How Well Do Computers Solve Math Word Problems? Large-Scale Dataset Construction and Evaluation." ACL 2016.
  • JSON: http://json.org/

Content

The filtered dataset includes problems that are relevant for preparing for maths competitions such as AIME and AMC. The data is structured to facilitate the training and evaluation of AI models aimed at solving these types of problems.

Datasets:

There are several filtered versions of the dataset based on different similarity thresholds (0.3 and 0.5). These thresholds were used to determine the relevance of problems from the original Sigma Dolphin dataset to the AIME and AMC problems.

1. Number Word Problems Filtered at 0.3 Threshold:

   • File: number_word_test_filtered_0.3_Threshold.csv
   • Description: Contains problems filtered with a similarity threshold of 0.3, ensuring moderate relevance to AIME and AMC 12 problems.

2. Number Word Problems Filtered at 0.5 Threshold:

   • File: number_word_std.test_filtered_0.5_Threshold.csv
   • Description: Contains problems filtered with a higher similarity threshold of 0.5, ensuring higher relevance to AIME and AMC 12 problems.

3. Filtered Number Word Problems 2 at 0.3 Threshold:

   • File: filtered_number_word_problems2_Threshold.csv
   • Description: Another set of problems filtered at a 0.3 similarity threshold.

4. Filtered Number Word Problems 2 at 0.5 Threshold:

   • File: filtered_number_word_problems_Threshold.csv
   • Description: Another set of problems filtered at a 0.5 similarity threshold.

Why Different Similarity Thresholds?

Different similarity thresholds (0.3 and 0.5) are used to provide flexibility in selecting problems based on their relevance to AIME and AMC problems. A lower threshold (0.3) includes a broader range of problems, ensuring a diverse set of questions, while a higher threshold (0.5) focuses on problems with stronger relevance, offering a more targeted and precise dataset. This allows users to choose the level of specificity that best fits their needs.

For a detailed explanation of the preprocessing and filtering process, please refer to the Sigma Dolphin Filtered & Cleaned Notebook.

Acknowledgements

We extend our gratitude to all the original authors of the Sigma Dolphin dataset and the creators of the AIME and AMC problems. This project leverages the work of numerous researchers and datasets to build a comprehensive resource for AI-based problem solving in mathematics.

Usage

This dataset is intended for research and educational purposes. It can be used to train AI models for natural language processing and problem-solving tasks, specifically targeting maths word problems in competitive environments like AIME and AMC.

Licensing

This dataset is shared under the Computational Use of Data Agreement v1.0.

This description provides an extensive overview of the dataset, its sources, contents, and usage. If any specific details or additional sections are needed, please let me know!

Since the beginning of the 1960s, Statistics Sweden, in collaboration with various research institutions, has carried out follow-up surveys in the school system. These surveys have taken place within the framework of the IS project (Individual Statistics Project) at the University of Gothenburg and the UGU project (Evaluation through follow-up of students) at the University of Teacher Education in Stockholm, which since 1990 have been merged into a research project called 'Evaluation through Follow-up'. The follow-up surveys are part of the central evaluation of the school and are based on large nationally representative samples from different cohorts of students.

Evaluation through follow-up (UGU) is one of the country's largest research databases in the field of education. UGU is part of the central evaluation of the school and is based on large nationally representative samples from different cohorts of students. The longitudinal database contains information on nationally representative samples of school pupils from ten cohorts, born between 1948 and 2004. The sampling process was based on the student's birthday for the first two and on the school class for the other cohorts.

For each cohort, data of mainly two types are collected. School administrative data is collected annually by Statistics Sweden during the time that pupils are in the general school system (primary and secondary school), for most cohorts starting in compulsory school year 3. This information is provided by the school offices and, among other things, includes characteristics of school, class, special support, study choices and grades. Information obtained has varied somewhat, e.g. due to changes in curricula. A more detailed description of this data collection can be found in reports published by Statistics Sweden and linked to datasets for each cohort.

Survey data from the pupils is collected for the first time in compulsory school year 6 (for most cohorts). Questionnaire in survey in year 6 includes questions related to self-perception and interest in learning, attitudes to school, hobbies, school motivation and future plans. For some cohorts, questionnaire data are also collected in year 3 and year 9 in compulsory school and in upper secondary school.

Furthermore, results from various intelligence tests and standartized knowledge tests are included in the data collection year 6. The intelligence tests have been identical for all cohorts (except cohort born in 1987 from which questionnaire data were first collected in year 9). The intelligence test consists of a verbal, a spatial and an inductive test, each containing 40 tasks and specially designed for the UGU project. The verbal test is a vocabulary test of the opposite type. The spatial test is a so-called ‘sheet metal folding test’ and the inductive test are made up of series of numbers. The reliability of the test, intercorrelations and connection with school grades are reported by Svensson (1971).

For the first three cohorts (1948, 1953 and 1967), the standartized knowledge tests in year 6 consist of the standard tests in Swedish, mathematics and English that up to and including the beginning of the 1980s were offered to all pupils in compulsory school year 6. For the cohort 1972, specially prepared tests in reading and mathematics were used. The test in reading consists of 27 tasks and aimed to identify students with reading difficulties. The mathematics test, which was also offered for the fifth cohort, (1977) includes 19 assignments. After a changed version of the test, caused by the previously used test being judged to be somewhat too simple, has been used for the cohort born in 1982. Results on the mathematics test are not available for the 1987 cohort. The mathematics test was not offered to the students in the cohort in 1992, as the test did not seem to fully correspond with current curriculum intentions in mathematics. For further information, see the description of the dataset for each cohort.

For several of the samples, questionnaires were also collected from the students 'parents and teachers in year 6. The teacher questionnaire contains questions about the teacher, class size and composition, the teacher's assessments of the class' knowledge level, etc., school resources, working methods and parental involvement and questions about the existence of evaluations. The questionnaire for the guardians includes questions about the child's upbringing conditions, ambitions and wishes regarding the child's education, views on the school's objectives and the parents' own educational and professional situation.

The students are followed up even after they have left primary school. Among other things, data collection is done during the time they are in high school. Then school administrative data such as e.g. choice of upper secondary school line / program and grades after completing studies. For some of the cohorts, in addition to school administrative data, questionnaire data were also collected from the students.

he sample consisted of students born on the 5th, 15th and 25th of any month in 1953, a total of 10,723 students.

The data obtained in 1966 were: 1. School administrative data (school form, class type, year and grades). 2. Information about the parents' profession and education, number of siblings, the distance between home and school, etc.

This information was collected for 93% of all born on the current days. The reason for this is reduced resources for Statistics Sweden for follow-up work - reminders etc. Annual data for cohorts in 1953 were collected by Statistics Sweden up to and including academic year 1972/73.

1. Answers to certain questions that shed light on students' school motivation, leisure activities and study and career plans. Some of the questions changed significantly compared to the cohort in 1948 due to the fact that they did not function satisfactorily from a metrological point of view.
2. Results on three aptitude tests, one verbal, one spatial and one inductive.
3. Standard test results in reading, writing, mathematics and English, which were offered to the students who belonged to year 6.

Response rate for test and questionnaire data is 88% Standard test results were received for just over 85% of those who took the tests.

The sample included a total of 9955 students, for whom some form of information was obtained.

Part of the "Individual Statistics Project" together with cohort 1953.

This report, based on results from PISA 2012, shows that one way forward is to ensure that all students spend more “engaged” time learning core mathematics concepts and solving challenging mathematics tasks. The opportunity to learn mathematics content – the time students spend learning mathematics topics and practising maths tasks at school – can accurately predict mathematics literacy. Differences in students’ familiarity with mathematics concepts explain a substantial share of performance disparities in PISA between socio-economically advantaged and disadvantaged students. Widening access to mathematics content can raise average levels of achievement and, at the same time, reduce inequalities in education and in society at large.

Korean Test Questions Structured Analysis Processing Data, around 2.4 million questions, contains question types, questions, answers, explanations, etc..For subjects, include [Primary School] Korean, Mathematics, English, Social Studies, Science; [Middle School] Korean, English, Mathematics, Science, Social Studies; [High School] Korean, English, Mathematics, Physics, Chemistry, Biology, History, Geography; question Types indlude single-choice question, fill-in question, true or false question, short answer question, etc. This dataset can be used for large-scale subject knowledge enhancement tasks.

Data content Korean K12, university test question

Amount around 2.4 million questions

Data fields Contains question types, questions, answers, explanations, etc.

Subject and Grade Level K12, university；contains math,physics,chemistry,biology

Question Types single-choice question, fill-in question, true or false question, short answer question, etc.

Format Jsonl

Language Korean

This dataset tracks annual distribution of students across grade levels in Big Walnut Middle School

This paper presents the results of two randomized experiments conducted in schools in urban India. A remedial education program hired young women to teach students lagging behind in basic literacy and numeracy skills. It increased average test scores of all children in treatment schools by 0.28 standard deviation, mostly due to large gains experienced by children at the bottom of the test-score distribution. A computer-assisted learning program focusing on math increased math scores by 0.47 standard deviation. One year after the programs were over, initial gains remained significant for targeted children, but they faded to about 0.10 standard deviation.