https://github.com/anshitag/memit_csk

Source repository for Editing Common Sense in Transformers (EMNLP 2023)

https://github.com/anshitag/memit_csk

Keywords

Repository

Source repository for Editing Common Sense in Transformers (EMNLP 2023)

https://github.com/anshitag/memit_csk/blob/main/

# Editing Common Sense in Transformers
EMNLP 2023 Paper: [Editing Common Sense in Transformers](https://arxiv.org/abs/2305.14956)

We find that commonsense judgments by GPT-2 are associated with localized parameters in early MLP layers of the models by conducting causal mediation analyses. We propose to correct commonsense judgments in transformer models by $MEMIT_{CSK}$, an adaptation of [Mass-Editing Memory in a Transformer](https://github.com/kmeng01/memit) that can edit subject, verb, or object token positions and features a robust editing layer selection strategy.

## Table of Contents

- [Installation](#installation)
- [Datasets](#datasets)
- [Evaluation Metrics](#evaluation-metrics)
- [Base Finetuning](#base-finetuning)
- [Causal Tracing](#causal-tracing)
- [Repair Finetuning](#repair-finetuning)
- [MEMIT_CSK Experiment](#memit_csk-experiment)
- [How to Cite](#how-to-cite)

## Installation

Similar to [MEMIT](https://github.com/kmeng01/memit) installation instructions. 
We recommend `conda` for managing Python, CUDA, and PyTorch; `pip` is for everything else. To get started, simply install `conda` and run:
```bash
CONDA_HOME=$CONDA_HOME ./scripts/setup_conda.sh
```

`$CONDA_HOME` should be the path to your `conda` installation, e.g., `~/miniconda3`.

## Datasets
The 20 Question and PEP 3K datasets are under the [data](data) folder. Each dataset consists of Train, Edit Validation, Edit, and Probe Sets.
- **Training Set** and **Edit Validation Set**: Formed by randomly dividing the validation set from Porada et al. (2021) into an 80%-20% split.
- **Edit Set**: The test set from Porada et al. (2021).
- **Probe Set**: For the subset of Edit Set that was incorrectly predicted by both GPT-2 Large and XL Base Model, we augment each instance with semantically related instances generated by GPT-3 text-davinci-003. The types of relations we consider include unaffected neighborhood, affected neighborhood, affected paraphrase, and affected reasoning, which are detailed in Sec. 4.2.1 of our paper.

## Evaluation Metrics
We report three metrics on the **Edit Validation Set** and **Edit Set**:
- F1 Score (), a measure of overall performance
- Efficacy (), the percentage of previously incorrect predictions that are corrected by an update method
- Relapse (), the percentage of instances that were previously predicted correctly but are now predicted incorrectly

We report accuracy on different types of augmented instances in the **Probe Set**.

The metrics are at [`eval_utils_csk.py`](memit_csk_experiment/py/eval_utils_csk.py).

## Base Finetuning

[`script_base_finetuning.sh`](base_finetune_experiments/script_base_finetuning.sh) can be used for running base finetuning on the GPT2-XL model for the 20q dataset. Similar command can be used for running experiments for GPT2-Large model and PEP 3K dataset.

## Causal Tracing

[`script_causal_trace_zero_shot.sh`](causal_tracing_experiment/script_causal_trace_zero_shot.sh) can be used for performing causal tracing experiment for zero shot model.

[`script_causal_trace.sh`](causal_tracing_experiment/script_causal_trace.sh) can be used for performing causal tracing experiment for base finetuned model, by passing its checkpoint location as a parameter and output inference file.

[`script_causal_trace_severed.sh`](causal_tracing_experiment/script_causal_trace_severed.sh) can be used for performing *severed* causal tracing experiment for base finetuned model, by passing it's checkpoint location as a parameter and output inference file.

## Repair Finetuning

[`script_repair_finetuning.sh`](repair_finetune_experiments/script_repair_finetuning.sh) can be used for running repair finetuning on the base finetuned GPT2-XL model for the 20q dataset. Similar command can be used for running experiments for GPT2-Large model and PEP 3K dataset. 

It includes commands to evaluate the affected and unaffected metrics for the repair finetuned model.

## MEMIT_CSK Experiment

[`script_memit_csk.sh`](script_memit_csk.sh) can be used for running $MEMIT_{CSK}$ on the base finetuned GPT2-XL model for the 20q dataset. Similar commands can be used for running experiments for GPT2-Large model and PEP 3K dataset. 

It includes commands to (1) find best hyperparamters for the Edit Validation Set and perform *configuration generalization* evaluation on the Edit Set, and (2) run *semantic generalization* evaluation on the Probe Set. Please refer to Section 3.3-3.4 in our paper for descriptions of the two types of generalization.

## How to Cite

```bibtex
@inproceedings{gupta2023editing,
    title = "Editing Common Sense in Transformers",
    author={Gupta, Anshita and Mondal, Debanjan and Sheshadri, Akshay Krishna and Zhao, Wenlong and Li, Xiang Lorraine and Wiegreffe, Sarah and Tandon, Niket},
    booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2305.14956",
}
```

Owner

GitHub Events