diff --git "a/dataset_infos.json" "b/dataset_infos.json" deleted file mode 100644--- "a/dataset_infos.json" +++ /dev/null @@ -1,4558 +0,0 @@ -{ - "adjunct_island": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "adjunct_island", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 165894, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 62231, - "dataset_size": 165894, - "size_in_bytes": 228125 - }, - "anaphor_gender_agreement": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "anaphor_gender_agreement", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 130918, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 39201, - "dataset_size": 130918, - "size_in_bytes": 170119 - }, - "anaphor_number_agreement": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "anaphor_number_agreement", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 139879, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 41547, - "dataset_size": 139879, - "size_in_bytes": 181426 - }, - "animate_subject_passive": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "animate_subject_passive", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 144423, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 47282, - "dataset_size": 144423, - "size_in_bytes": 191705 - }, - "animate_subject_trans": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "animate_subject_trans", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 127798, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 49651, - "dataset_size": 127798, - "size_in_bytes": 177449 - }, - "causative": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "causative", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 122772, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 48963, - "dataset_size": 122772, - "size_in_bytes": 171735 - }, - "complex_NP_island": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "complex_NP_island", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 198972, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 78211, - "dataset_size": 198972, - "size_in_bytes": 277183 - }, - "coordinate_structure_constraint_complex_left_branch": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "coordinate_structure_constraint_complex_left_branch", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 210912, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 67908, - "dataset_size": 210912, - "size_in_bytes": 278820 - }, - "coordinate_structure_constraint_object_extraction": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "coordinate_structure_constraint_object_extraction", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 171655, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 51584, - "dataset_size": 171655, - "size_in_bytes": 223239 - }, - "determiner_noun_agreement_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "determiner_noun_agreement_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 156120, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 49893, - "dataset_size": 156120, - "size_in_bytes": 206013 - }, - "determiner_noun_agreement_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "determiner_noun_agreement_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 156204, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 49527, - "dataset_size": 156204, - "size_in_bytes": 205731 - }, - "determiner_noun_agreement_irregular_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "determiner_noun_agreement_irregular_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 164473, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 47274, - "dataset_size": 164473, - "size_in_bytes": 211747 - }, - "determiner_noun_agreement_irregular_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "determiner_noun_agreement_irregular_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 161074, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 47422, - "dataset_size": 161074, - "size_in_bytes": 208496 - }, - "determiner_noun_agreement_with_adj_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "determiner_noun_agreement_with_adj_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 179666, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 56346, - "dataset_size": 179666, - "size_in_bytes": 236012 - }, - "determiner_noun_agreement_with_adj_irregular_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "determiner_noun_agreement_with_adj_irregular_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 184529, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 54405, - "dataset_size": 184529, - "size_in_bytes": 238934 - }, - "determiner_noun_agreement_with_adj_irregular_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "determiner_noun_agreement_with_adj_irregular_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 184396, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 54064, - "dataset_size": 184396, - "size_in_bytes": 238460 - }, - "determiner_noun_agreement_with_adjective_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "determiner_noun_agreement_with_adjective_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 185126, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 55682, - "dataset_size": 185126, - "size_in_bytes": 240808 - }, - "distractor_agreement_relational_noun": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "distractor_agreement_relational_noun", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 191473, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 59641, - "dataset_size": 191473, - "size_in_bytes": 251114 - }, - "distractor_agreement_relative_clause": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "distractor_agreement_relative_clause", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 216756, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 77897, - "dataset_size": 216756, - "size_in_bytes": 294653 - }, - "drop_argument": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "drop_argument", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 109806, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 39961, - "dataset_size": 109806, - "size_in_bytes": 149767 - }, - "ellipsis_n_bar_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "ellipsis_n_bar_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 217590, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 92776, - "dataset_size": 217590, - "size_in_bytes": 310366 - }, - "ellipsis_n_bar_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "ellipsis_n_bar_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 233161, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 98882, - "dataset_size": 233161, - "size_in_bytes": 332043 - }, - "existential_there_object_raising": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "existential_there_object_raising", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 223741, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 76641, - "dataset_size": 223741, - "size_in_bytes": 300382 - }, - "existential_there_quantifiers_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "existential_there_quantifiers_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 162931, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 51576, - "dataset_size": 162931, - "size_in_bytes": 214507 - }, - "existential_there_quantifiers_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "existential_there_quantifiers_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 164826, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 52092, - "dataset_size": 164826, - "size_in_bytes": 216918 - }, - "existential_there_subject_raising": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "existential_there_subject_raising", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 200063, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 59519, - "dataset_size": 200063, - "size_in_bytes": 259582 - }, - "expletive_it_object_raising": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "expletive_it_object_raising", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 238615, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 88607, - "dataset_size": 238615, - "size_in_bytes": 327222 - }, - "inchoative": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "inchoative", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 104319, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 39842, - "dataset_size": 104319, - "size_in_bytes": 144161 - }, - "intransitive": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "intransitive", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 111097, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 42387, - "dataset_size": 111097, - "size_in_bytes": 153484 - }, - "irregular_past_participle_adjectives": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "irregular_past_participle_adjectives", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 144661, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 36654, - "dataset_size": 144661, - "size_in_bytes": 181315 - }, - "irregular_past_participle_verbs": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "irregular_past_participle_verbs", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 125692, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 37297, - "dataset_size": 125692, - "size_in_bytes": 162989 - }, - "irregular_plural_subject_verb_agreement_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "irregular_plural_subject_verb_agreement_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 165584, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 50725, - "dataset_size": 165584, - "size_in_bytes": 216309 - }, - "irregular_plural_subject_verb_agreement_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "irregular_plural_subject_verb_agreement_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 153843, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 42707, - "dataset_size": 153843, - "size_in_bytes": 196550 - }, - "left_branch_island_echo_question": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "left_branch_island_echo_question", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 147840, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 50481, - "dataset_size": 147840, - "size_in_bytes": 198321 - }, - "left_branch_island_simple_question": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "left_branch_island_simple_question", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 150060, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 50293, - "dataset_size": 150060, - "size_in_bytes": 200353 - }, - "matrix_question_npi_licensor_present": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "matrix_question_npi_licensor_present", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 153262, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 51899, - "dataset_size": 153262, - "size_in_bytes": 205161 - }, - "npi_present_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "npi_present_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 138465, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 51981, - "dataset_size": 138465, - "size_in_bytes": 190446 - }, - "npi_present_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "npi_present_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 127636, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 51661, - "dataset_size": 127636, - "size_in_bytes": 179297 - }, - "only_npi_licensor_present": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "only_npi_licensor_present", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 148516, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 51361, - "dataset_size": 148516, - "size_in_bytes": 199877 - }, - "only_npi_scope": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "only_npi_scope", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 208902, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 84970, - "dataset_size": 208902, - "size_in_bytes": 293872 - }, - "passive_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "passive_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 145882, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 53931, - "dataset_size": 145882, - "size_in_bytes": 199813 - }, - "passive_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "passive_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 113960, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 40499, - "dataset_size": 113960, - "size_in_bytes": 154459 - }, - "principle_A_c_command": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "principle_A_c_command", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 188490, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 67867, - "dataset_size": 188490, - "size_in_bytes": 256357 - }, - "principle_A_case_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "principle_A_case_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 170398, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 61092, - "dataset_size": 170398, - "size_in_bytes": 231490 - }, - "principle_A_case_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "principle_A_case_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 170412, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 56430, - "dataset_size": 170412, - "size_in_bytes": 226842 - }, - "principle_A_domain_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "principle_A_domain_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 171170, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 59120, - "dataset_size": 171170, - "size_in_bytes": 230290 - }, - "principle_A_domain_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "principle_A_domain_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 165333, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 58464, - "dataset_size": 165333, - "size_in_bytes": 223797 - }, - "principle_A_domain_3": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "principle_A_domain_3", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 158998, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 52859, - "dataset_size": 158998, - "size_in_bytes": 211857 - }, - "principle_A_reconstruction": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "principle_A_reconstruction", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 152104, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 44480, - "dataset_size": 152104, - "size_in_bytes": 196584 - }, - "regular_plural_subject_verb_agreement_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "regular_plural_subject_verb_agreement_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 158819, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 49466, - "dataset_size": 158819, - "size_in_bytes": 208285 - }, - "regular_plural_subject_verb_agreement_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "regular_plural_subject_verb_agreement_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 153609, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 43365, - "dataset_size": 153609, - "size_in_bytes": 196974 - }, - "sentential_negation_npi_licensor_present": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "sentential_negation_npi_licensor_present", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 171864, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 54830, - "dataset_size": 171864, - "size_in_bytes": 226694 - }, - "sentential_negation_npi_scope": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "sentential_negation_npi_scope", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 232098, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 90157, - "dataset_size": 232098, - "size_in_bytes": 322255 - }, - "sentential_subject_island": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "sentential_subject_island", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 172432, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 56666, - "dataset_size": 172432, - "size_in_bytes": 229098 - }, - "superlative_quantifiers_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "superlative_quantifiers_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 159290, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 48453, - "dataset_size": 159290, - "size_in_bytes": 207743 - }, - "superlative_quantifiers_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "superlative_quantifiers_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 159340, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 50480, - "dataset_size": 159340, - "size_in_bytes": 209820 - }, - "tough_vs_raising_1": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "tough_vs_raising_1", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 148636, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 44779, - "dataset_size": 148636, - "size_in_bytes": 193415 - }, - "tough_vs_raising_2": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "tough_vs_raising_2", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 169684, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 61465, - "dataset_size": 169684, - "size_in_bytes": 231149 - }, - "transitive": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "transitive", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 133104, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 55090, - "dataset_size": 133104, - "size_in_bytes": 188194 - }, - "wh_island": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "wh_island", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 142340, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 52808, - "dataset_size": 142340, - "size_in_bytes": 195148 - }, - "wh_questions_object_gap": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "wh_questions_object_gap", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 193045, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 70049, - "dataset_size": 193045, - "size_in_bytes": 263094 - }, - "wh_questions_subject_gap": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "wh_questions_subject_gap", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 195593, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 71632, - "dataset_size": 195593, - "size_in_bytes": 267225 - }, - "wh_questions_subject_gap_long_distance": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "wh_questions_subject_gap_long_distance", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 268270, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 98913, - "dataset_size": 268270, - "size_in_bytes": 367183 - }, - "wh_vs_that_no_gap": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "wh_vs_that_no_gap", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 188872, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 71710, - "dataset_size": 188872, - "size_in_bytes": 260582 - }, - "wh_vs_that_no_gap_long_distance": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "wh_vs_that_no_gap_long_distance", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 247039, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 95504, - "dataset_size": 247039, - "size_in_bytes": 342543 - }, - "wh_vs_that_with_gap": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "wh_vs_that_with_gap", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 173386, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 60291, - "dataset_size": 173386, - "size_in_bytes": 233677 - }, - "wh_vs_that_with_gap_long_distance": { - "description": "\nBLiMP is a challenge set for evaluating what language models (LMs) know about\nmajor grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each\ncontaining 1000 minimal pairs isolating specific contrasts in syntax,\nmorphology, or semantics. The data is automatically generated according to\nexpert-crafted grammars.\n", - "citation": "\n@article{warstadt2019blimp,\n title={BLiMP: A Benchmark of Linguistic Minimal Pairs for English},\n author={Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei, and Wang, Sheng-Fu and Bowman, Samuel R},\n journal={arXiv preprint arXiv:1912.00582},\n year={2019}\n}\n", - "homepage": "https://github.com/alexwarstadt/blimp/tree/master/", - "license": "", - "features": { - "sentence_good": { - "dtype": "string", - "_type": "Value" - }, - "sentence_bad": { - "dtype": "string", - "_type": "Value" - }, - "field": { - "dtype": "string", - "_type": "Value" - }, - "linguistics_term": { - "dtype": "string", - "_type": "Value" - }, - "UID": { - "dtype": "string", - "_type": "Value" - }, - "simple_LM_method": { - "dtype": "bool", - "_type": "Value" - }, - "one_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "two_prefix_method": { - "dtype": "bool", - "_type": "Value" - }, - "lexically_identical": { - "dtype": "bool", - "_type": "Value" - }, - "pair_id": { - "dtype": "int32", - "_type": "Value" - } - }, - "builder_name": "blimp", - "dataset_name": "blimp", - "config_name": "wh_vs_that_with_gap_long_distance", - "version": { - "version_str": "0.1.0", - "major": 0, - "minor": 1, - "patch": 0 - }, - "splits": { - "train": { - "name": "train", - "num_bytes": 231595, - "num_examples": 1000, - "dataset_name": null - } - }, - "download_size": 84147, - "dataset_size": 231595, - "size_in_bytes": 315742 - } -} \ No newline at end of file