Individual results

View docs

View in-depth performance of a single language model on a single test suite.

Region-by-region surprisal
Sample item for Negative Polarity Licensing (ever; with object relative clause)
Item
Condition
Licensor np compl rc_dp rc_subj rc_verb has npi continuation
Item Condition Licensor np compl rc_dp rc_subj rc_verb has npi continuation
1 neg_pos No author that the senators liked has ever been popular
1 neg_neg No author that no senators liked has ever been popular
1 pos_pos The author that the senators liked has ever been popular
1 pos_neg The author that no senators liked has ever been popular
Prediction performance for GPT-2 XL on Negative Polarity Licensing (ever; with object relative clause)
Accuracy
Formula
Description
AccuracyPredictionDescription
100.00% (591,neg_pos/8,npi) < (589,pos_pos/8,npi) No description provided.
100.00% (590,neg_neg/8,npi) < (592,pos_neg/8,npi) No description provided.
100.00% (591,neg_pos/8,npi) < (592,pos_neg/8,npi) No description provided.