Skip to content

Latest commit

 

History

History
26 lines (20 loc) · 938 Bytes

README.md

File metadata and controls

26 lines (20 loc) · 938 Bytes

Injecting Numerical Reasoning Skills into Language Models

This repository contains the accompanying code for the paper:

"Injecting Numerical Reasoning Skills into Language Models." Mor Geva*, Ankit Gupta* and Jonathan Berant. In ACL, 2020. [PDF]

Structure

The repository contains:

  • Implementation/pre-training/finetuning of GenBERT on MLM/synthetic-data/DROP/SQuAD (in pre_training dir)
  • Code and vocabularies for textual data generation (in textual_data_generation dir)
  • Code for numerical data generation (in pre_training/numeric_data_generation dir)

Instructions for downloading our data + models are in the README of pre_training dir.


Citation

@inproceedings{ggb2020injecting,
  title={Injecting Numerical Reasoning Skills into Language Models},
  author={Geva, Mor and Gupta, Ankit and Berant, Jonathan},
  booktitle={ ACL },
  year={ 2020 }
}