Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2408.00612 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909298867568640 |
|---|---|
| author | Baksi, Arkadeep Singh, Rahul Joshi, Tarun |
| author_facet | Baksi, Arkadeep Singh, Rahul Joshi, Tarun |
| contents | The advent of transformer-based architectures and large language models (LLMs) have significantly advanced the performance of natural language processing (NLP) models. Since these LLMs are trained on huge corpuses of data from the web and other sources, there has been a major concern about harmful prejudices that may potentially be transferred from the data. In many applications, these pre-trained LLMs are fine-tuned on task specific datasets, which can further contribute to biases. This paper studies the extent of biases absorbed by LLMs during pre-training as well as task-specific behaviour after fine-tuning. We found that controlled interventions on pre-trained LLMs, prior to fine-tuning, have minimal effect on lowering biases in classifiers. However, the biases present in domain-specific datasets play a much bigger role, and hence mitigating them at this stage has a bigger impact. While pre-training does matter, but after the model has been pre-trained, even slight changes to co-occurrence rates in the fine-tuning dataset has a significant effect on the bias of the model. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_00612 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Downstream bias mitigation is all you need Baksi, Arkadeep Singh, Rahul Joshi, Tarun Computation and Language The advent of transformer-based architectures and large language models (LLMs) have significantly advanced the performance of natural language processing (NLP) models. Since these LLMs are trained on huge corpuses of data from the web and other sources, there has been a major concern about harmful prejudices that may potentially be transferred from the data. In many applications, these pre-trained LLMs are fine-tuned on task specific datasets, which can further contribute to biases. This paper studies the extent of biases absorbed by LLMs during pre-training as well as task-specific behaviour after fine-tuning. We found that controlled interventions on pre-trained LLMs, prior to fine-tuning, have minimal effect on lowering biases in classifiers. However, the biases present in domain-specific datasets play a much bigger role, and hence mitigating them at this stage has a bigger impact. While pre-training does matter, but after the model has been pre-trained, even slight changes to co-occurrence rates in the fine-tuning dataset has a significant effect on the bias of the model. |
| title | Downstream bias mitigation is all you need |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2408.00612 |