MiRa
Collection
8 items
•
Updated
Dataset Viewer
input
stringlengths 41
66
| reference
stringlengths 3
12
|
|---|---|
The librarian documented which journalist the celebrities had
|
interviewed
|
The librarian documented which celebrities the journalist had
|
interviewed
|
the old widower remembered which ghost the villager had
|
haunted
|
the old widower remembered which villager the ghost had
|
haunted
|
the firefighter reported which paramedic the victim had
|
saved
|
the firefighter reported which victim the paramedic had
|
saved
|
the historian documented which assassin the prince had
|
killed
|
the historian documented which prince the assassin had
|
killed
|
the judo master recognized which assailant the defender had
|
attacked
|
the judo master recognized which defender the assailant had
|
attacked
|
the naturalist observed which deer the predators had
|
avoided
|
the naturalist observed which predators the deer had
|
avoided
|
the restaurant owner forgot which waitress the customer had
|
served
|
the restaurant owner forgot which customer the waitress had
|
served
|
the nanny knew which billionaire the housekeeper had
|
hired
|
the nanny knew which housekeeper the billionaire had
|
hired
|
the researcher learned which biologist the whale had
|
studied
|
the researcher learned which whale the biologist had
|
studied
|
Tracy announced which administrator the applicant had
|
chosen
|
Tracy announced which applicant the administrator had
|
chosen
|
the birthday boy saw which clown the friend had
|
scared
|
the birthday boy saw which friend the clown had
|
scared
|
the columnist publicized which celebrity the photographer had
|
punched
|
the columnist publicized which photographer the celebrity had
|
punched
|
the police officer noted which burglar the jeweler had
|
robbed
|
the police officer noted which jeweler the burglar had
|
robbed
|
the royal cook asked which chef the lord had
|
served
|
the royal cook asked which lord the chef had
|
served
|
the wildlife curator mentioned which zookeeper the ape had
|
fed
|
the wildlife curator mentioned which ape the zookeeper had
|
fed
|
the aid worker realized which philanthropist the refugees had
|
helped
|
the aid worker realized which refugees the philanthropist had
|
helped
|
the camper reported which bear the girl had
|
eaten
|
the camper reported which girl the bear had
|
eaten
|
the dance instructor noticed which performers the onlookers had
|
impressed
|
the dance instructor noticed which onlookers the performers had
|
impressed
|
the travel agent recorded which guests the innkeeper had
|
liked
|
the travel agent recorded which innkeeper the guests had
|
liked
|
the bird watcher saw which hawk the photographer had
|
attacked
|
the bird watcher saw which photographer the hawk had
|
attacked
|
the costume designer confirmed which barber the actor had
|
shaved
|
the costume designer confirmed which actor the barber had
|
shaved
|
the investigator discovered which abductor the couple had
|
taken
|
the investigator discovered which couple the abductor had
|
taken
|
the jail keeper forgot which prisoner the guard had
|
attacked
|
the jail keeper forgot which guard the prisoner had
|
attacked
|
the sheriff recalled which gangsters the locals had
|
robbed
|
the sheriff recalled which locals the gangsters had
|
robbed
|
the undead king recognized which zombie the woman had
|
bitten
|
the undead king recognized which woman the zombie had
|
bitten
|
the broadcaster explained which judge the contestant had
|
disqualified
|
the broadcaster explained which contestant the judge had
|
disqualified
|
the family counselor understood which stepparent the daughter had
|
abused
|
the family counselor understood which daughter the stepparent had
|
abused
|
the newscaster confirmed which policeman the fugitive had
|
arrested
|
the newscaster confirmed which fugitive the policeman had
|
arrested
|
the opera composer revealed which singer the pirate had
|
portrayed
|
the opera composer revealed which pirate the singer had
|
portrayed
|
the prison warden guessed which inmate the jailer had
|
stabbed
|
the prison warden guessed which jailer the inmate had
|
stabbed
|
the producer observed which acrobats the viewers had
|
impressed
|
the producer observed which viewers the acrobats had
|
impressed
|
the queen recognized which comedian the nobleman had
|
entertained
|
the queen recognized which nobleman the comedian had
|
entertained
|
The ape researcher identified which poacher the chimpanzee had
|
shot
|
The ape researcher identified which chimpanzee the poacher had
|
shot
|
the committee member guessed which president the advisors had
|
appointed
|
the committee member guessed which advisors the president had
|
appointed
|
the farmer indicated which townsperson the donkey had
|
bought
|
the farmer indicated which donkey the townsperson had
|
bought
|
the military man knew which general the soldiers had
|
promoted
|
the military man knew which soldiers the general had
|
promoted
|
the parole officer confirmed which psychologist the delinquent had
|
treated
|
the parole officer confirmed which delinquent the psychologist had
|
treated
|
the policeman knew which detective the suspect had
|
arrested
|
the policeman knew which suspect the detective had
|
arrested
|
Jack observed which conductor the passenger had
|
removed
|
Jack observed which passenger the conductor had
|
removed
|
Teresa knew which parrot the neighbor had
|
bitten
|
Teresa knew which neighbor the parrot had
|
bitten
|
the art critic identified which artist the model had
|
drawn
|
the art critic identified which model the artist had
|
drawn
|
The cotton farmer wondered which slave the master had
|
escaped
|
The cotton farmer wondered which master the slave had
|
escaped
|
the farmhand recorded which breeder the goats had
|
sold
|
the farmhand recorded which goats the breeder had
|
sold
|
the four-year-old knew which dragon the hero had
|
eaten
|
the four-year-old knew which hero the dragon had
|
eaten
|
Dataset Card for LM Diagnostics (cprag) Clone
This repository contains a diagnostic dataset (cprag) for What BERT is not: Lessons from a new suite of psycholinguistic diagnostics for language models, by Allyson Ettinger.
Licensing Information
The dataset is released under the MIT License.
Citation Information
@article{10.1162/tacl_a_00298,
author = {Ettinger, Allyson},
title = {What BERT Is Not: Lessons from a New Suite of Psycholinguistic Diagnostics for Language Models},
journal = {Transactions of the Association for Computational Linguistics},
volume = {8},
pages = {34-48},
year = {2020},
month = {01},
abstract = {Pre-training by language modeling has become a popular and successful approach to NLP tasks, but we have yet to understand exactly what linguistic capacities these pre-training processes confer upon models. In this paper we introduce a suite of diagnostics drawn from human language experiments, which allow us to ask targeted questions about information used by language models for generating predictions in context. As a case study, we apply these diagnostics to the popular BERT model, finding that it can generally distinguish good from bad completions involving shared category or role reversal, albeit with less sensitivity than humans, and it robustly retrieves noun hypernyms, but it struggles with challenging inference and role-based event prediction— and, in particular, it shows clear insensitivity to the contextual impacts of negation.},
issn = {2307-387X},
doi = {10.1162/tacl_a_00298},
url = {https://doi.org/10.1162/tacl_a_00298},
eprint = {https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00298/1923116/tacl_a_00298.pdf},
}
- Downloads last month
- 9
Size of downloaded dataset files:
4.43 kB
Size of the auto-converted Parquet files:
4.43 kB
Number of rows:
88