dataset_infos.json

{"abstract_algebra": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "abstract_algebra", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 19316, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 2012, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 818, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160623403, "size_in_bytes": 326808363}, "anatomy": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "anatomy", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 33109, "num_examples": 135, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 3128, "num_examples": 14, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 955, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160638449, "size_in_bytes": 326823409}, "astronomy": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "astronomy", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 46759, "num_examples": 152, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 5015, "num_examples": 16, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2064, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160655095, "size_in_bytes": 326840055}, "business_ethics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "business_ethics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 33240, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 3026, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2178, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160639701, "size_in_bytes": 326824661}, "clinical_knowledge": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "clinical_knowledge", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 62742, "num_examples": 265, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 6652, "num_examples": 29, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1198, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160671849, "size_in_bytes": 326856809}, "college_biology": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "college_biology", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 48785, "num_examples": 144, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 4807, "num_examples": 16, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1520, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160656369, "size_in_bytes": 326841329}, "college_chemistry": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "college_chemistry", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 24696, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 2316, "num_examples": 8, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1319, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160629588, "size_in_bytes": 326814548}, "college_computer_science": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "college_computer_science", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 42629, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 4651, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2753, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160651290, "size_in_bytes": 326836250}, "college_mathematics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "college_mathematics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 24699, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 2656, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1481, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160630093, "size_in_bytes": 326815053}, "college_medicine": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "college_medicine", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 82385, "num_examples": 173, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 7897, "num_examples": 22, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1658, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160693197, "size_in_bytes": 326878157}, "college_physics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "college_physics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 30169, "num_examples": 102, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 3478, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1400, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160636304, "size_in_bytes": 326821264}, "computer_security": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "computer_security", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 27112, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 4537, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1089, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160633995, "size_in_bytes": 326818955}, "conceptual_physics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "conceptual_physics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 40697, "num_examples": 235, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 4462, "num_examples": 26, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 922, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160647338, "size_in_bytes": 326832298}, "econometrics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "econometrics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 46535, "num_examples": 114, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 4955, "num_examples": 12, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1632, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160654379, "size_in_bytes": 326839339}, "electrical_engineering": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "electrical_engineering", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 25130, "num_examples": 145, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 2891, "num_examples": 16, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 960, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160630238, "size_in_bytes": 326815198}, "elementary_mathematics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "elementary_mathematics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 70096, "num_examples": 378, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 8976, "num_examples": 41, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1428, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160681757, "size_in_bytes": 326866717}, "formal_logic": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "formal_logic", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 49773, "num_examples": 126, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 6240, "num_examples": 14, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1745, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160659015, "size_in_bytes": 326843975}, "global_facts": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "global_facts", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 18391, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 1853, "num_examples": 10, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1217, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160622718, "size_in_bytes": 326807678}, "high_school_biology": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_biology", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 109720, "num_examples": 310, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 11010, "num_examples": 32, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1661, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160723648, "size_in_bytes": 326908608}, "high_school_chemistry": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_chemistry", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 58452, "num_examples": 203, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 7080, "num_examples": 22, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1208, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160667997, "size_in_bytes": 326852957}, "high_school_computer_science": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_computer_science", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 44464, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 3331, "num_examples": 9, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2906, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160651958, "size_in_bytes": 326836918}, "high_school_european_history": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_european_history", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 270288, "num_examples": 165, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 29620, "num_examples": 18, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 11552, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160912717, "size_in_bytes": 327097677}, "high_school_geography": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_geography", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 42022, "num_examples": 198, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 4320, "num_examples": 22, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1391, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160648990, "size_in_bytes": 326833950}, "high_school_government_and_politics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_government_and_politics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 66062, "num_examples": 193, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 7051, "num_examples": 21, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1767, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160676137, "size_in_bytes": 326861097}, "high_school_macroeconomics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_macroeconomics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 117675, "num_examples": 390, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 13008, "num_examples": 43, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1316, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160733256, "size_in_bytes": 326918216}, "high_school_mathematics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_mathematics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 54842, "num_examples": 270, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 5753, "num_examples": 29, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1285, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160663137, "size_in_bytes": 326848097}, "high_school_microeconomics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_microeconomics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 75691, "num_examples": 238, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 7541, "num_examples": 26, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1286, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160685775, "size_in_bytes": 326870735}, "high_school_physics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_physics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 59526, "num_examples": 151, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 6759, "num_examples": 17, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1477, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160669019, "size_in_bytes": 326853979}, "high_school_psychology": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_psychology", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 159395, "num_examples": 545, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 17257, "num_examples": 60, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1893, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160779802, "size_in_bytes": 326964762}, "high_school_statistics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_statistics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 110690, "num_examples": 216, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 9985, "num_examples": 23, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2516, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160724448, "size_in_bytes": 326909408}, "high_school_us_history": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_us_history", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 296722, "num_examples": 204, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 31694, "num_examples": 22, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 8852, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160938525, "size_in_bytes": 327123485}, "high_school_world_history": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "high_school_world_history", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 378605, "num_examples": 237, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 45489, "num_examples": 26, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 4870, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 161030221, "size_in_bytes": 327215181}, "human_aging": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "human_aging", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 46086, "num_examples": 223, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 4695, "num_examples": 23, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 996, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160653034, "size_in_bytes": 326837994}, "human_sexuality": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "human_sexuality", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 32098, "num_examples": 131, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 2409, "num_examples": 12, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1065, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160636829, "size_in_bytes": 326821789}, "international_law": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "international_law", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 53519, "num_examples": 121, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 6461, "num_examples": 13, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2406, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160663643, "size_in_bytes": 326848603}, "jurisprudence": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "jurisprudence", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 33974, "num_examples": 108, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 3717, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1291, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160640239, "size_in_bytes": 326825199}, "logical_fallacies": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "logical_fallacies", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 50105, "num_examples": 163, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 5091, "num_examples": 18, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1561, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160658014, "size_in_bytes": 326842974}, "machine_learning": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "machine_learning", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 33868, "num_examples": 112, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 3220, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2311, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160640656, "size_in_bytes": 326825616}, "management": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "management", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 19990, "num_examples": 103, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 1808, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 886, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160623941, "size_in_bytes": 326808901}, "marketing": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "marketing", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 63013, "num_examples": 234, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 7382, "num_examples": 25, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1469, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160673121, "size_in_bytes": 326858081}, "medical_genetics": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "medical_genetics", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 20852, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 2993, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1077, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160626179, "size_in_bytes": 326811139}, "miscellaneous": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "miscellaneous", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 147692, "num_examples": 783, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 14318, "num_examples": 86, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 687, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160763954, "size_in_bytes": 326948914}, "moral_disputes": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "moral_disputes", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 107806, "num_examples": 346, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 12408, "num_examples": 38, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1743, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160723214, "size_in_bytes": 326908174}, "moral_scenarios": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "moral_scenarios", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 374014, "num_examples": 895, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 42326, "num_examples": 100, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2046, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 161019643, "size_in_bytes": 327204603}, "nutrition": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "nutrition", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 92398, "num_examples": 306, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 8424, "num_examples": 33, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2073, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160704152, "size_in_bytes": 326889112}, "philosophy": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "philosophy", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 80061, "num_examples": 311, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 9172, "num_examples": 34, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 976, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160691466, "size_in_bytes": 326876426}, "prehistory": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "prehistory", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 89582, "num_examples": 324, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 10273, "num_examples": 35, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1866, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160702978, "size_in_bytes": 326887938}, "professional_accounting": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "professional_accounting", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 124538, "num_examples": 282, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 14360, "num_examples": 31, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2136, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160742291, "size_in_bytes": 326927251}, "professional_law": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "professional_law", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 1891750, "num_examples": 1534, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 203507, "num_examples": 170, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 6598, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 162703112, "size_in_bytes": 328888072}, "professional_medicine": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "professional_medicine", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 217549, "num_examples": 272, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 23835, "num_examples": 31, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 3795, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160846436, "size_in_bytes": 327031396}, "professional_psychology": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "professional_psychology", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 225887, "num_examples": 612, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 29089, "num_examples": 69, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 2255, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160858488, "size_in_bytes": 327043448}, "public_relations": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "public_relations", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 28748, "num_examples": 110, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 4554, "num_examples": 12, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1484, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160636043, "size_in_bytes": 326821003}, "security_studies": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "security_studies", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 204832, "num_examples": 245, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 22625, "num_examples": 27, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 5323, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160834037, "size_in_bytes": 327018997}, "sociology": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "sociology", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 66231, "num_examples": 201, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 7172, "num_examples": 22, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1601, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160676261, "size_in_bytes": 326861221}, "us_foreign_policy": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "us_foreign_policy", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 28431, "num_examples": 100, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 3252, "num_examples": 11, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1599, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160634539, "size_in_bytes": 326819499}, "virology": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "virology", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 38747, "num_examples": 166, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 5451, "num_examples": 18, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 1084, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160646539, "size_in_bytes": 326831499}, "world_religions": {"description": "This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.\n", "citation": "@article{hendryckstest2021,\n      title={Measuring Massive Multitask Language Understanding},\n      author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},\n      journal={Proceedings of the International Conference on Learning Representations (ICLR)},\n      year={2021}\n    }\n", "homepage": "https://github.com/hendrycks/test", "license": "", "features": {"question": {"dtype": "string", "id": null, "_type": "Value"}, "choices": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"num_classes": 4, "names": ["A", "B", "C", "D"], "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "mmlu", "config_name": "world_religions", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"auxiliary_train": {"name": "auxiliary_train", "num_bytes": 160601257, "num_examples": 99842, "dataset_name": "mmlu"}, "test": {"name": "test", "num_bytes": 25262, "num_examples": 171, "dataset_name": "mmlu"}, "validation": {"name": "validation", "num_bytes": 2753, "num_examples": 19, "dataset_name": "mmlu"}, "dev": {"name": "dev", "num_bytes": 658, "num_examples": 5, "dataset_name": "mmlu"}}, "download_checksums": {"data.tar": {"num_bytes": 166184960, "checksum": "bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b"}}, "download_size": 166184960, "post_processing_size": null, "dataset_size": 160629930, "size_in_bytes": 326814890}}