PreCog: Exploring the Relation between Memorization and Performance in Pre-trained Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2023
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866929584142811136 |
|---|---|
| author | Ranaldi, Leonardo Ruzzetti, Elena Sofia Zanzotto, Fabio Massimo |
| author_facet | Ranaldi, Leonardo Ruzzetti, Elena Sofia Zanzotto, Fabio Massimo |
| contents | Pre-trained Language Models such as BERT are impressive machines with the ability to memorize, possibly generalized learning examples. We present here a small, focused contribution to the analysis of the interplay between memorization and performance of BERT in downstream tasks. We propose PreCog, a measure for evaluating memorization from pre-training, and we analyze its correlation with the BERT's performance. Our experiments show that highly memorized examples are better classified, suggesting memorization is an essential key to success for BERT. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2305_04673 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | PreCog: Exploring the Relation between Memorization and Performance in Pre-trained Language Models Ranaldi, Leonardo Ruzzetti, Elena Sofia Zanzotto, Fabio Massimo Computation and Language Artificial Intelligence Pre-trained Language Models such as BERT are impressive machines with the ability to memorize, possibly generalized learning examples. We present here a small, focused contribution to the analysis of the interplay between memorization and performance of BERT in downstream tasks. We propose PreCog, a measure for evaluating memorization from pre-training, and we analyze its correlation with the BERT's performance. Our experiments show that highly memorized examples are better classified, suggesting memorization is an essential key to success for BERT. |
| title | PreCog: Exploring the Relation between Memorization and Performance in Pre-trained Language Models |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2305.04673 |