Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2409.18857 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916740908187648 |
|---|---|
| author | Choi, Hyeong Kyu Xu, Weijie Xue, Chi Eckman, Stephanie Reddy, Chandan K. |
| author_facet | Choi, Hyeong Kyu Xu, Weijie Xue, Chi Eckman, Stephanie Reddy, Chandan K. |
| contents | Large language models (LLMs) often exhibit systematic preferences for certain answer choices when responding to multiple-choice questions-a behavior known as selection bias. This bias reduces the accuracy and reliability of LLM outputs, limiting their usefulness in decision-critical applications. While prior work has focused on adjusting model inputs or outputs to mitigate this issue, our work takes a fundamentally different approach by identifying and removing the internal sources of bias. We introduce two methods: Bias Node Pruning (BNP), which prunes parameters that contribute to selection bias, and Auxiliary Option Injection (AOI), which introduces an additional answer choice to reduce bias in both white-box and black-box settings. To address the shortcomings of existing evaluation metrics, we propose Choice Kullback-Leibler Divergence (CKLD), a new metric that captures distributional imbalances in model predictions. Experiments on three LLMs across multiple datasets demonstrate that our methods consistently improve answer accuracy while reducing selection bias, providing a robust solution for both open- and closed-source models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_18857 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Mitigating Selection Bias with Node Pruning and Auxiliary Options Choi, Hyeong Kyu Xu, Weijie Xue, Chi Eckman, Stephanie Reddy, Chandan K. Artificial Intelligence Large language models (LLMs) often exhibit systematic preferences for certain answer choices when responding to multiple-choice questions-a behavior known as selection bias. This bias reduces the accuracy and reliability of LLM outputs, limiting their usefulness in decision-critical applications. While prior work has focused on adjusting model inputs or outputs to mitigate this issue, our work takes a fundamentally different approach by identifying and removing the internal sources of bias. We introduce two methods: Bias Node Pruning (BNP), which prunes parameters that contribute to selection bias, and Auxiliary Option Injection (AOI), which introduces an additional answer choice to reduce bias in both white-box and black-box settings. To address the shortcomings of existing evaluation metrics, we propose Choice Kullback-Leibler Divergence (CKLD), a new metric that captures distributional imbalances in model predictions. Experiments on three LLMs across multiple datasets demonstrate that our methods consistently improve answer accuracy while reducing selection bias, providing a robust solution for both open- and closed-source models. |
| title | Mitigating Selection Bias with Node Pruning and Auxiliary Options |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2409.18857 |