sentence_good
stringlengths
40
81
sentence_bad
stringlengths
40
81
field
stringclasses
1 value
linguistics_term
stringclasses
1 value
UID
stringclasses
1 value
simple_LM_method
bool
1 class
one_prefix_method
bool
1 class
two_prefix_method
bool
1 class
lexically_identical
bool
1 class
pair_id
int32
0
999
Who should Derek hug after shocking Richard?
Who should Derek hug Richard after shocking?
syntax
island_effects
adjunct_island
true
false
false
true
0
What had Theresa walked through while talking about that high school?
What had Theresa walked through that high school while talking about?
syntax
island_effects
adjunct_island
true
false
false
true
1
Who will Katherine discover without hiring Erin?
Who will Katherine discover Erin without hiring?
syntax
island_effects
adjunct_island
true
false
false
true
2
Who has Colleen aggravated before kissing Judy?
Who has Colleen aggravated Judy before kissing?
syntax
island_effects
adjunct_island
true
false
false
true
3
What could a lot of cats break while finding all convertibles?
What could a lot of cats break all convertibles while finding?
syntax
island_effects
adjunct_island
true
false
false
true
4
Who have most people discovered while embarrassing Erin?
Who have most people discovered Erin while embarrassing?
syntax
island_effects
adjunct_island
true
false
false
true
5
Who was William firing before talking about Maria?
Who was William firing Maria before talking about?
syntax
island_effects
adjunct_island
true
false
false
true
6
What is Denise descending while hiding a lot of hills?
What is Denise descending a lot of hills while hiding?
syntax
island_effects
adjunct_island
true
false
false
true
7
Who does John leave while alarming Beverly?
Who does John leave Beverly while alarming?
syntax
island_effects
adjunct_island
true
false
false
true
8
What was Melanie going to after taking those rivers?
What was Melanie going to those rivers after taking?
syntax
island_effects
adjunct_island
true
false
false
true
9
Who could Bethany research before discussing Winston Churchill?
Who could Bethany research Winston Churchill before discussing?
syntax
island_effects
adjunct_island
true
false
false
true
10
What could Jessica sell before noticing these spotlights?
What could Jessica sell these spotlights before noticing?
syntax
island_effects
adjunct_island
true
false
false
true
11
What had Helen biked to without bothering Spain?
What had Helen biked to Spain without bothering?
syntax
island_effects
adjunct_island
true
false
false
true
12
Who is Mary irritating after approaching Kenneth?
Who is Mary irritating Kenneth after approaching?
syntax
island_effects
adjunct_island
true
false
false
true
13
Who might Rose flee from before returning to this customer?
Who might Rose flee from this customer before returning to?
syntax
island_effects
adjunct_island
true
false
false
true
14
What will Janice research after boasting about politics?
What will Janice research politics after boasting about?
syntax
island_effects
adjunct_island
true
false
false
true
15
Who had Karla aggravated without finding Donald?
Who had Karla aggravated Donald without finding?
syntax
island_effects
adjunct_island
true
false
false
true
16
Who should a government reference before shocking Jesus?
Who should a government reference Jesus before shocking?
syntax
island_effects
adjunct_island
true
false
false
true
17
What had Aaron sounded like while cleaning the museum?
What had Aaron sounded like the museum while cleaning?
syntax
island_effects
adjunct_island
true
false
false
true
18
What is Brenda arriving at while exiting many schools?
What is Brenda arriving at many schools while exiting?
syntax
island_effects
adjunct_island
true
false
false
true
19
Who is a window hurting without disturbing Sabrina?
Who is a window hurting Sabrina without disturbing?
syntax
island_effects
adjunct_island
true
false
false
true
20
Who had Jesus bothered after leaving Lori?
Who had Jesus bothered Lori after leaving?
syntax
island_effects
adjunct_island
true
false
false
true
21
Who are a lot of hamsters disgusting while finding Ronald?
Who are a lot of hamsters disgusting Ronald while finding?
syntax
island_effects
adjunct_island
true
false
false
true
22
Who could Amelia leave while appreciating Paul?
Who could Amelia leave Paul while appreciating?
syntax
island_effects
adjunct_island
true
false
false
true
23
What have many ladies dropped by without talking about all malls?
What have many ladies dropped by all malls without talking about?
syntax
island_effects
adjunct_island
true
false
false
true
24
What is Larry dropping by after lifting a grocery store?
What is Larry dropping by a grocery store after lifting?
syntax
island_effects
adjunct_island
true
false
false
true
25
Who is that student hugging before stunning Catherine?
Who is that student hugging Catherine before stunning?
syntax
island_effects
adjunct_island
true
false
false
true
26
Who were those schools appreciating without scaring Alice?
Who were those schools appreciating Alice without scaring?
syntax
island_effects
adjunct_island
true
false
false
true
27
What did Dan sell after revealing some plane?
What did Dan sell some plane after revealing?
syntax
island_effects
adjunct_island
true
false
false
true
28
What had Marcus climbed down before buying many mountains?
What had Marcus climbed down many mountains before buying?
syntax
island_effects
adjunct_island
true
false
false
true
29
What has Jacqueline scanned before revealing that report?
What has Jacqueline scanned that report before revealing?
syntax
island_effects
adjunct_island
true
false
false
true
30
Who is Tanya returning to while astounding Donald?
Who is Tanya returning to Donald while astounding?
syntax
island_effects
adjunct_island
true
false
false
true
31
Who had Kirsten referenced without boring Liam?
Who had Kirsten referenced Liam without boring?
syntax
island_effects
adjunct_island
true
false
false
true
32
What are some drivers touring after going to every hill?
What are some drivers touring every hill after going to?
syntax
island_effects
adjunct_island
true
false
false
true
33
Who would Mark visit while kissing Marla?
Who would Mark visit Marla while kissing?
syntax
island_effects
adjunct_island
true
false
false
true
34
What is Scott cleaning after cleaning a pie?
What is Scott cleaning a pie after cleaning?
syntax
island_effects
adjunct_island
true
false
false
true
35
Who has some doctor listened to after firing that actress?
Who has some doctor listened to that actress after firing?
syntax
island_effects
adjunct_island
true
false
false
true
36
Who has that lady approached after alarming Ellen?
Who has that lady approached Ellen after alarming?
syntax
island_effects
adjunct_island
true
false
false
true
37
What can Randolf bike to before discussing the lakes?
What can Randolf bike to the lakes before discussing?
syntax
island_effects
adjunct_island
true
false
false
true
38
What had some senator toured before hiding those cafes?
What had some senator toured those cafes before hiding?
syntax
island_effects
adjunct_island
true
false
false
true
39
Who was Eric impressing without admiring Christina?
Who was Eric impressing Christina without admiring?
syntax
island_effects
adjunct_island
true
false
false
true
40
What do most drivers climb up without noticing most hills?
What do most drivers climb up most hills without noticing?
syntax
island_effects
adjunct_island
true
false
false
true
41
Who had Gregory worked with while noticing Julie?
Who had Gregory worked with Julie while noticing?
syntax
island_effects
adjunct_island
true
false
false
true
42
Who was Rose astounding while hiding many waiters?
Who was Rose astounding many waiters while hiding?
syntax
island_effects
adjunct_island
true
false
false
true
43
What is Karen researching before walking through that mountain?
What is Karen researching that mountain before walking through?
syntax
island_effects
adjunct_island
true
false
false
true
44
Who could Dana boast about after disturbing Brenda?
Who could Dana boast about Brenda after disturbing?
syntax
island_effects
adjunct_island
true
false
false
true
45
What was this bird aggravating before annoying every hospital?
What was this bird aggravating every hospital before annoying?
syntax
island_effects
adjunct_island
true
false
false
true
46
Who had Colleen known after concealing Stephen?
Who had Colleen known Stephen after concealing?
syntax
island_effects
adjunct_island
true
false
false
true
47
What are these customers arriving at without alarming Spain?
What are these customers arriving at Spain without alarming?
syntax
island_effects
adjunct_island
true
false
false
true
48
What has Angela admired while cleaning shoes?
What has Angela admired shoes while cleaning?
syntax
island_effects
adjunct_island
true
false
false
true
49
What is Becky selling before noticing some hospital?
What is Becky selling some hospital before noticing?
syntax
island_effects
adjunct_island
true
false
false
true
50
Who had many cashiers concealed after investigating Nina?
Who had many cashiers concealed Nina after investigating?
syntax
island_effects
adjunct_island
true
false
false
true
51
Who had Melanie admired before listening to Chad?
Who had Melanie admired Chad before listening to?
syntax
island_effects
adjunct_island
true
false
false
true
52
Who does Bethany visit before scaring Amanda?
Who does Bethany visit Amanda before scaring?
syntax
island_effects
adjunct_island
true
false
false
true
53
What is Tiffany exiting before climbing down the mountain?
What is Tiffany exiting the mountain before climbing down?
syntax
island_effects
adjunct_island
true
false
false
true
54
Who had Susan insulted without boring Heather?
Who had Susan insulted Heather without boring?
syntax
island_effects
adjunct_island
true
false
false
true
55
Who has Curtis noticed while investigating Steve?
Who has Curtis noticed Steve while investigating?
syntax
island_effects
adjunct_island
true
false
false
true
56
Who has Kayla disliked after disgusting Derek?
Who has Kayla disliked Derek after disgusting?
syntax
island_effects
adjunct_island
true
false
false
true
57
What is Martin skated around without buying these hospitals?
What is Martin skated around these hospitals without buying?
syntax
island_effects
adjunct_island
true
false
false
true
58
What was Beth dropping by after bringing all glaciers?
What was Beth dropping by all glaciers after bringing?
syntax
island_effects
adjunct_island
true
false
false
true
59
Who had Rhonda distracted while finding Chad?
Who had Rhonda distracted Chad while finding?
syntax
island_effects
adjunct_island
true
false
false
true
60
What has Bethany driven to before biking to a lot of hills?
What has Bethany driven to a lot of hills before biking to?
syntax
island_effects
adjunct_island
true
false
false
true
61
Who have the slopes hurt before disgusting a lot of adults?
Who have the slopes hurt a lot of adults before disgusting?
syntax
island_effects
adjunct_island
true
false
false
true
62
What will Brad wear before noticing these scarves?
What will Brad wear these scarves before noticing?
syntax
island_effects
adjunct_island
true
false
false
true
63
What had Steven boycotted before running around the art gallery?
What had Steven boycotted the art gallery before running around?
syntax
island_effects
adjunct_island
true
false
false
true
64
Who do the Borgias like before aggravating Mark?
Who do the Borgias like Mark before aggravating?
syntax
island_effects
adjunct_island
true
false
false
true
65
What had the patients resembled without hiding most sketches?
What had the patients resembled most sketches without hiding?
syntax
island_effects
adjunct_island
true
false
false
true
66
Who would Jesus hug without discovering Bradley?
Who would Jesus hug Bradley without discovering?
syntax
island_effects
adjunct_island
true
false
false
true
67
Who has Mark sounded like before finding that doctor?
Who has Mark sounded like that doctor before finding?
syntax
island_effects
adjunct_island
true
false
false
true
68
Who has Rebecca fled from without listening to Kirsten?
Who has Rebecca fled from Kirsten without listening to?
syntax
island_effects
adjunct_island
true
false
false
true
69
Who was Ellen escaping from before curing Julia?
Who was Ellen escaping from Julia before curing?
syntax
island_effects
adjunct_island
true
false
false
true
70
What does Travis fix after breaking every carriage?
What does Travis fix every carriage after breaking?
syntax
island_effects
adjunct_island
true
false
false
true
71
What might Tanya walk through after hiding every cafe?
What might Tanya walk through every cafe after hiding?
syntax
island_effects
adjunct_island
true
false
false
true
72
What has Ella brought after finding all icicles?
What has Ella brought all icicles after finding?
syntax
island_effects
adjunct_island
true
false
false
true
73
What was Valerie selling after climbing down some slope?
What was Valerie selling some slope after climbing down?
syntax
island_effects
adjunct_island
true
false
false
true
74
What could Douglas bike to without admiring every mountain?
What could Douglas bike to every mountain without admiring?
syntax
island_effects
adjunct_island
true
false
false
true
75
What would those teenagers research before boycotting some malls?
What would those teenagers research some malls before boycotting?
syntax
island_effects
adjunct_island
true
false
false
true
76
What can Dan clean before cleaning every screen?
What can Dan clean every screen before cleaning?
syntax
island_effects
adjunct_island
true
false
false
true
77
Who would Deborah see while hugging Christina?
Who would Deborah see Christina while hugging?
syntax
island_effects
adjunct_island
true
false
false
true
78
Who can Gregory talk about before hugging Emily?
Who can Gregory talk about Emily before hugging?
syntax
island_effects
adjunct_island
true
false
false
true
79
What should Elaine argue about before admiring books?
What should Elaine argue about books before admiring?
syntax
island_effects
adjunct_island
true
false
false
true
80
What had Jason dropped by before selling those schools?
What had Jason dropped by those schools before selling?
syntax
island_effects
adjunct_island
true
false
false
true
81
What will the guys notice before shocking a lot of museums?
What will the guys notice a lot of museums before shocking?
syntax
island_effects
adjunct_island
true
false
false
true
82
Who could Monet worry while thinking about Dana?
Who could Monet worry Dana while thinking about?
syntax
island_effects
adjunct_island
true
false
false
true
83
Who could Thomas observe without distracting Peter?
Who could Thomas observe Peter without distracting?
syntax
island_effects
adjunct_island
true
false
false
true
84
Who has Suzanne admired after scaring Connie?
Who has Suzanne admired Connie after scaring?
syntax
island_effects
adjunct_island
true
false
false
true
85
Who can some closets disgust without shocking that dancer?
Who can some closets disgust that dancer without shocking?
syntax
island_effects
adjunct_island
true
false
false
true
86
Who has Bruce annoyed while investigating Kevin?
Who has Bruce annoyed Kevin while investigating?
syntax
island_effects
adjunct_island
true
false
false
true
87
What was Bethany lifting without returning to these cafes?
What was Bethany lifting these cafes without returning to?
syntax
island_effects
adjunct_island
true
false
false
true
88
What should Wendy boycott after questioning every hospital?
What should Wendy boycott every hospital after questioning?
syntax
island_effects
adjunct_island
true
false
false
true
89
What are these pedestrians selling after finding that bird?
What are these pedestrians selling that bird after finding?
syntax
island_effects
adjunct_island
true
false
false
true
90
What is Laurie buying before cleaning the car?
What is Laurie buying the car before cleaning?
syntax
island_effects
adjunct_island
true
false
false
true
91
What does Matt embarrass before bothering every legislature?
What does Matt embarrass every legislature before bothering?
syntax
island_effects
adjunct_island
true
false
false
true
92
What have some drivers worn without finding a shawl?
What have some drivers worn a shawl without finding?
syntax
island_effects
adjunct_island
true
false
false
true
93
Who can Kimberley help before healing Martin?
Who can Kimberley help Martin before healing?
syntax
island_effects
adjunct_island
true
false
false
true
94
Who have most students distracted before impressing Marla?
Who have most students distracted Marla before impressing?
syntax
island_effects
adjunct_island
true
false
false
true
95
What was Dawn boycotting after describing public parks?
What was Dawn boycotting public parks after describing?
syntax
island_effects
adjunct_island
true
false
false
true
96
Who was Cynthia observing while thinking about Rose?
Who was Cynthia observing Rose while thinking about?
syntax
island_effects
adjunct_island
true
false
false
true
97
What do many fish break before breaking the truck?
What do many fish break the truck before breaking?
syntax
island_effects
adjunct_island
true
false
false
true
98
Who was Ann thinking about without distracting some cashiers?
Who was Ann thinking about some cashiers without distracting?
syntax
island_effects
adjunct_island
true
false
false
true
99

Dataset Card for "blimp"

Dataset Summary

BLiMP is a challenge set for evaluating what language models (LMs) know about major grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each containing 1000 minimal pairs isolating specific contrasts in syntax, morphology, or semantics. The data is automatically generated according to expert-crafted grammars.

Supported Tasks and Leaderboards

More Information Needed

Languages

More Information Needed

Dataset Structure

Data Instances

adjunct_island

  • Size of downloaded dataset files: 0.36 MB
  • Size of the generated dataset: 0.17 MB
  • Total amount of disk used: 0.52 MB

An example of 'train' looks as follows.

{
    "UID": "tough_vs_raising_1",
    "field": "syntax_semantics",
    "lexically_identical": false,
    "linguistics_term": "control_raising",
    "one_prefix_method": false,
    "pair_id": 2,
    "sentence_bad": "Benjamin's tutor was certain to boast about.",
    "sentence_good": "Benjamin's tutor was easy to boast about.",
    "simple_LM_method": true,
    "two_prefix_method": false
}

anaphor_gender_agreement

  • Size of downloaded dataset files: 0.44 MB
  • Size of the generated dataset: 0.14 MB
  • Total amount of disk used: 0.57 MB

An example of 'train' looks as follows.

{
    "UID": "tough_vs_raising_1",
    "field": "syntax_semantics",
    "lexically_identical": false,
    "linguistics_term": "control_raising",
    "one_prefix_method": false,
    "pair_id": 2,
    "sentence_bad": "Benjamin's tutor was certain to boast about.",
    "sentence_good": "Benjamin's tutor was easy to boast about.",
    "simple_LM_method": true,
    "two_prefix_method": false
}

anaphor_number_agreement

  • Size of downloaded dataset files: 0.45 MB
  • Size of the generated dataset: 0.14 MB
  • Total amount of disk used: 0.59 MB

An example of 'train' looks as follows.

{
    "UID": "tough_vs_raising_1",
    "field": "syntax_semantics",
    "lexically_identical": false,
    "linguistics_term": "control_raising",
    "one_prefix_method": false,
    "pair_id": 2,
    "sentence_bad": "Benjamin's tutor was certain to boast about.",
    "sentence_good": "Benjamin's tutor was easy to boast about.",
    "simple_LM_method": true,
    "two_prefix_method": false
}

animate_subject_passive

  • Size of downloaded dataset files: 0.46 MB
  • Size of the generated dataset: 0.15 MB
  • Total amount of disk used: 0.61 MB

An example of 'train' looks as follows.

{
    "UID": "tough_vs_raising_1",
    "field": "syntax_semantics",
    "lexically_identical": false,
    "linguistics_term": "control_raising",
    "one_prefix_method": false,
    "pair_id": 2,
    "sentence_bad": "Benjamin's tutor was certain to boast about.",
    "sentence_good": "Benjamin's tutor was easy to boast about.",
    "simple_LM_method": true,
    "two_prefix_method": false
}

animate_subject_trans

  • Size of downloaded dataset files: 0.43 MB
  • Size of the generated dataset: 0.13 MB
  • Total amount of disk used: 0.57 MB

An example of 'train' looks as follows.

{
    "UID": "tough_vs_raising_1",
    "field": "syntax_semantics",
    "lexically_identical": false,
    "linguistics_term": "control_raising",
    "one_prefix_method": false,
    "pair_id": 2,
    "sentence_bad": "Benjamin's tutor was certain to boast about.",
    "sentence_good": "Benjamin's tutor was easy to boast about.",
    "simple_LM_method": true,
    "two_prefix_method": false
}

Data Fields

The data fields are the same among all splits.

adjunct_island

  • sentence_good: a string feature.
  • sentence_bad: a string feature.
  • field: a string feature.
  • linguistics_term: a string feature.
  • UID: a string feature.
  • simple_LM_method: a bool feature.
  • one_prefix_method: a bool feature.
  • two_prefix_method: a bool feature.
  • lexically_identical: a bool feature.
  • pair_id: a int32 feature.

anaphor_gender_agreement

  • sentence_good: a string feature.
  • sentence_bad: a string feature.
  • field: a string feature.
  • linguistics_term: a string feature.
  • UID: a string feature.
  • simple_LM_method: a bool feature.
  • one_prefix_method: a bool feature.
  • two_prefix_method: a bool feature.
  • lexically_identical: a bool feature.
  • pair_id: a int32 feature.

anaphor_number_agreement

  • sentence_good: a string feature.
  • sentence_bad: a string feature.
  • field: a string feature.
  • linguistics_term: a string feature.
  • UID: a string feature.
  • simple_LM_method: a bool feature.
  • one_prefix_method: a bool feature.
  • two_prefix_method: a bool feature.
  • lexically_identical: a bool feature.
  • pair_id: a int32 feature.

animate_subject_passive

  • sentence_good: a string feature.
  • sentence_bad: a string feature.
  • field: a string feature.
  • linguistics_term: a string feature.
  • UID: a string feature.
  • simple_LM_method: a bool feature.
  • one_prefix_method: a bool feature.
  • two_prefix_method: a bool feature.
  • lexically_identical: a bool feature.
  • pair_id: a int32 feature.

animate_subject_trans

  • sentence_good: a string feature.
  • sentence_bad: a string feature.
  • field: a string feature.
  • linguistics_term: a string feature.
  • UID: a string feature.
  • simple_LM_method: a bool feature.
  • one_prefix_method: a bool feature.
  • two_prefix_method: a bool feature.
  • lexically_identical: a bool feature.
  • pair_id: a int32 feature.

Data Splits

name train
adjunct_island 1000
anaphor_gender_agreement 1000
anaphor_number_agreement 1000
animate_subject_passive 1000
animate_subject_trans 1000

Dataset Creation

Curation Rationale

More Information Needed

Source Data

Initial Data Collection and Normalization

More Information Needed

Who are the source language producers?

More Information Needed

Annotations

Annotation process

More Information Needed

Who are the annotators?

More Information Needed

Personal and Sensitive Information

More Information Needed

Considerations for Using the Data

Social Impact of Dataset

More Information Needed

Discussion of Biases

More Information Needed

Other Known Limitations

More Information Needed

Additional Information

Dataset Curators

More Information Needed

Licensing Information

BLiMP is distributed under a CC-BY license. Source: https://github.com/alexwarstadt/blimp#license

Citation Information

@article{warstadt2020blimp,
    author = {Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei and Wang, Sheng-Fu and Bowman, Samuel R.},
    title = {BLiMP: The Benchmark of Linguistic Minimal Pairs for English},
    journal = {Transactions of the Association for Computational Linguistics},
    volume = {8},
    number = {},
    pages = {377-392},
    year = {2020},
    doi = {10.1162/tacl\_a\_00321},
    URL = {https://doi.org/10.1162/tacl_a_00321},
    eprint = {https://doi.org/10.1162/tacl_a_00321},
    abstract = { We introduce The Benchmark of Linguistic Minimal Pairs (BLiMP),1 a challenge set for evaluating the linguistic knowledge of language models (LMs) on major grammatical phenomena in English. BLiMP consists of 67 individual datasets, each containing 1,000 minimal pairs—that is, pairs of minimally different sentences that contrast in grammatical acceptability and isolate specific phenomenon in syntax, morphology, or semantics. We generate the data according to linguist-crafted grammar templates, and human aggregate agreement with the labels is 96.4\%. We evaluate n-gram, LSTM, and Transformer (GPT-2 and Transformer-XL) LMs by observing whether they assign a higher probability to the acceptable sentence in each minimal pair. We find that state-of-the-art models identify morphological contrasts related to agreement reliably, but they struggle with some subtle semantic and syntactic phenomena, such as negative polarity items and extraction islands. }
}

Errata

Some results were misreported in the published TACL version. Please refer to the corrected version on arXiv: https://arxiv.org/abs/1912.00582

Contributions

Thanks to @lhoestq, @patrickvonplaten, @thomwolf for adding this dataset.

Downloads last month
415

Models trained or fine-tuned on nyu-mll/blimp