{"url":"/dataset/race","name":"RACE","full_name":"ReAding Comprehension dataset from Examinations","description_markdown":"The **ReAding Comprehension dataset from Examinations** (**RACE**) dataset is a machine reading comprehension dataset consisting of 27,933 passages and 97,867 questions from English exams, targeting Chinese students aged 12-18. RACE consists of two subsets, RACE-M and RACE-H, from middle school and high school exams, respectively. RACE-M has 28,293 questions and RACE-H has 69,574. Each question is associated with 4 candidate answers, one of which is correct. The data generation process of RACE differs from most machine reading comprehension datasets - instead of generating questions and answers by heuristics or crowd-sourcing, questions in RACE are specifically designed for testing human reading skills, and are created by domain experts.\r\n\r\nSource: [Dynamic Fusion Networks for Machine Reading Comprehension](https://arxiv.org/abs/1711.04964)\r\nImage Source: [Lai et al](https://arxiv.org/pdf/1704.04683v5.pdf)","description_withheld":null,"homepage":"https://www.cs.cmu.edu/~glai1/data/race/","introduced_date":"2017-01-01","introduced_date_note":null,"introduced_by":{"paper":"/paper/race-large-scale-reading-comprehension","title":"RACE: Large-scale ReAding Comprehension Dataset From Examinations","first_author":"Guokun Lai","url":null},"license":{"name":"Custom (research-only, non-commercial)","url":"https://www.cs.cmu.edu/~glai1/data/race/#:~:text=notes"},"modalities":[{"name":"Texts","url":"/datasets/modality/texts"}],"tasks":[{"name":"Question Answering","url":"/task/question-answering","datasets_with_task":"/datasets/task/question-answering"},{"name":"Reading Comprehension","url":"/task/reading-comprehension","datasets_with_task":"/datasets/task/reading-comprehension"},{"name":"Distractor Generation","url":"/task/distractor-generation","datasets_with_task":"/datasets/task/distractor-generation"}],"languages":[{"name":"English","url":"/datasets/language/english"}],"variants":["RACE"],"data_loaders":[{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/bfattori/race","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/race","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/EleutherAI/race","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/ehovy/race","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/MBZUAI-Paris/EgyptianRACE","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/tensorflow/datasets","url":"https://www.tensorflow.org/datasets/catalog/race","frameworks":["tf","jax"]},{"repo":"https://github.com/RUCAIBox/LLMBox","url":"https://github.com/RUCAIBox/LLMBox/blob/main/docs/utilization/supported-datasets.md","frameworks":["pytorch"]}],"num_papers_in_archive":412,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[{"leaderboard":"/sota/reading-comprehension-on-race","task":"Reading Comprehension","dataset_variant":"RACE","rows":24,"metrics":["Accuracy","Accuracy (Middle)","Accuracy (High)"],"first_row_in_archive_order":{"model":"ALBERT (Ensemble)","paper":"/paper/improving-machine-reading-comprehension-with-2","metrics":{"Accuracy":"91.4"},"code_links":[]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"},{"leaderboard":"/sota/question-answering-on-race","task":"Question Answering","dataset_variant":"RACE","rows":7,"metrics":["RACE-m","RACE-h","RACE"],"first_row_in_archive_order":{"model":"XLNet","paper":"/paper/xlnet-generalized-autoregressive-pretraining","metrics":{"RACE":"81.75","RACE-m":"85.45"},"code_links":[{"title":"huggingface/transformers","url":"https://github.com/huggingface/transformers"},{"title":"PaddlePaddle/PaddleNLP","url":"https://github.com/PaddlePaddle/PaddleNLP/tree/develop/examples/language_model/xlnet"},{"title":"zihangdai/xlnet","url":"https://github.com/zihangdai/xlnet"},{"title":"kaushaltrivedi/fast-bert","url":"https://github.com/kaushaltrivedi/fast-bert"},{"title":"utterworks/fast-bert","url":"https://github.com/utterworks/fast-bert"},{"title":"graykode/xlnet-Pytorch","url":"https://github.com/graykode/xlnet-Pytorch"},{"title":"facebookresearch/anli","url":"https://github.com/facebookresearch/anli"},{"title":"lvyufeng/bert4ms","url":"https://github.com/lvyufeng/bert4ms/blob/master/bert4ms/models/xlnet.py"},{"title":"huggingface/xlnet","url":"https://github.com/huggingface/xlnet"},{"title":"cuhksz-nlp/SAPar","url":"https://github.com/cuhksz-nlp/SAPar"},{"title":"joshuaWang-bit/Textclassification-pytorch","url":"https://github.com/joshuaWang-bit/Textclassification-pytorch"},{"title":"fanchenyou/transformer-study","url":"https://github.com/fanchenyou/transformer-study"},{"title":"https-seyhan/BugAI","url":"https://github.com/https-seyhan/BugAI"},{"title":"NathanDuran/Sentence-Encoding-for-DA-Classification","url":"https://github.com/NathanDuran/Sentence-Encoding-for-DA-Classification"},{"title":"2miatran/Natural-Language-Processing","url":"https://github.com/2miatran/Natural-Language-Processing"},{"title":"chesterdu/contrastive_summary","url":"https://github.com/chesterdu/contrastive_summary"},{"title":"pauldevos/python-notes","url":"https://github.com/pauldevos/python-notes"},{"title":"pwc-1/Paper-9","url":"https://github.com/pwc-1/Paper-9/tree/main/5/xlnet"},{"title":"MindCode-4/code-5","url":"https://github.com/MindCode-4/code-5/tree/main/xlnet"},{"title":"pwc-1/Paper-9","url":"https://github.com/pwc-1/Paper-9/tree/main/1/xlnet"},{"title":"samwisegamjeee/pytorch-transformers","url":"https://github.com/samwisegamjeee/pytorch-transformers"},{"title":"SambhawDrag/XLNet.jl","url":"https://github.com/SambhawDrag/XLNet.jl"},{"title":"tomgoter/nlp_finalproject","url":"https://github.com/tomgoter/nlp_finalproject"},{"title":"MS-P3/code7","url":"https://github.com/MS-P3/code7/tree/main/xlnet"},{"title":"zaradana/Fast_BERT","url":"https://github.com/zaradana/Fast_BERT"},{"title":"jonahwinninghoff/Text-Summarization","url":"https://github.com/jonahwinninghoff/Text-Summarization"},{"title":"listenviolet/XLNet","url":"https://github.com/listenviolet/XLNet"}]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"},{"leaderboard":"/sota/distractor-generation-on-race","task":"Distractor Generation","dataset_variant":"RACE","rows":1,"metrics":["BLEU-1","BLEU-2","BLEU-3","BLEU-4","ROUGE-L"],"first_row_in_archive_order":{"model":"BDG p.m.","paper":"/paper/a-bert-based-distractor-generation-scheme","metrics":{"BLEU-1":"39.81","BLEU-2":"24.81","BLEU-3":"17.66","BLEU-4":"13.56","ROUGE-L":"34.01"},"code_links":[{"title":"voidful/BDG","url":"https://github.com/voidful/BDG"}]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"}],"papers_with_a_benchmark_row":[{"paper":"/paper/orca-2-teaching-small-language-models-how-to","title":"Orca 2: Teaching Small Language Models How to Reason","date":"2023-11-18","rows_on_this_dataset":2,"code_links":0,"syntology":null},{"paper":"/paper/bloomberggpt-a-large-language-model-for","title":"BloombergGPT: A Large Language Model for Finance","date":"2023-03-30","rows_on_this_dataset":4,"code_links":2,"syntology":null},{"paper":"/paper/llama-open-and-efficient-foundation-language-1","title":"LLaMA: Open and Efficient Foundation Language Models","date":"2023-02-27","rows_on_this_dataset":4,"code_links":57,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":58,"samples_ran":26,"samples_unverified":32,"pointer_only_for_licence":4,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/palm-scaling-language-modeling-with-pathways-1","title":"PaLM: Scaling Language Modeling with Pathways","date":"2022-04-05","rows_on_this_dataset":3,"code_links":7,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":37,"samples_ran":30,"samples_unverified":7,"pointer_only_for_licence":0,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/hierarchical-learning-for-generation-with","title":"Hierarchical Learning for Generation with Long Source Sequences","date":"2021-04-15","rows_on_this_dataset":1,"code_links":0,"syntology":null},{"paper":"/paper/improving-machine-reading-comprehension-with-2","title":"Improving Machine Reading Comprehension with Single-choice Decision and Transfer Learning","date":"2020-11-06","rows_on_this_dataset":1,"code_links":0,"syntology":null},{"paper":"/paper/a-bert-based-distractor-generation-scheme","title":"A BERT-based Distractor Generation Scheme with Multi-tasking and Negative Answer Training Strategies","date":"2020-10-12","rows_on_this_dataset":1,"code_links":1,"syntology":null},{"paper":"/paper/funnel-transformer-filtering-out-sequential","title":"Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing","date":"2020-06-05","rows_on_this_dataset":1,"code_links":3,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":2,"samples_ran":2,"samples_unverified":0,"pointer_only_for_licence":0,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/deberta-decoding-enhanced-bert-with","title":"DeBERTa: Decoding-enhanced BERT with Disentangled Attention","date":"2020-06-05","rows_on_this_dataset":1,"code_links":14,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":13,"samples_ran":4,"samples_unverified":9,"pointer_only_for_licence":3,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/language-models-are-few-shot-learners","title":"Language Models are Few-Shot Learners","date":"2020-05-28","rows_on_this_dataset":4,"code_links":67,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":65,"samples_ran":15,"samples_unverified":50,"pointer_only_for_licence":4,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/dual-multi-head-co-attention-for-multi-choice","title":"DUMA: Reading Comprehension with Transposition Thinking","date":"2020-01-26","rows_on_this_dataset":1,"code_links":3,"syntology":null},{"paper":"/paper/megatron-lm-training-multi-billion-parameter","title":"Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism","date":"2019-09-17","rows_on_this_dataset":2,"code_links":10,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":47,"samples_ran":12,"samples_unverified":35,"pointer_only_for_licence":15,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/roberta-a-robustly-optimized-bert-pretraining","title":"RoBERTa: A Robustly Optimized BERT Pretraining Approach","date":"2019-07-26","rows_on_this_dataset":1,"code_links":67,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":48,"samples_ran":22,"samples_unverified":26,"pointer_only_for_licence":23,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/xlnet-generalized-autoregressive-pretraining","title":"XLNet: Generalized Autoregressive Pretraining for Language Understanding","date":"2019-06-19","rows_on_this_dataset":2,"code_links":27,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":24,"samples_ran":10,"samples_unverified":14,"pointer_only_for_licence":3,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/option-comparison-network-for-multiple-choice","title":"Option Comparison Network for Multiple-choice Reading Comprehension","date":"2019-03-07","rows_on_this_dataset":1,"code_links":0,"syntology":null},{"paper":"/paper/dual-co-matching-network-for-multi-choice","title":"Dual Co-Matching Network for Multi-choice Reading Comprehension","date":"2019-01-27","rows_on_this_dataset":1,"code_links":0,"syntology":null},{"paper":"/paper/improving-language-understanding-by","title":"Improving Language Understanding by Generative Pre-Training","date":"2018-06-11","rows_on_this_dataset":1,"code_links":13,"syntology":null},{"paper":"/paper/multi-range-reasoning-for-machine","title":"Multi-range Reasoning for Machine Comprehension","date":"2018-03-24","rows_on_this_dataset":1,"code_links":0,"syntology":null}],"syntology_totals":{"read_at":"2026-09-24T18:15:14+00:00","papers_with_samples":8,"samples_harvested":294,"samples_ran":121,"samples_unverified":173,"pointer_only_for_licence":52,"papers_with_no_sample_that_ran":0,"note":"the per-paper counts above, summed; not a rate"},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks."}