LLM-MCQ-Bias

Official repository for our ICLR 2024 Spotlight paper "Large Language Models Are Not Robust Multiple Choice Selectors"

If you find this repository useful or our work is related to your research, please kindly cite it:

@inproceedings{
  llm-mcq-bias,
  title={Large Language Models Are Not Robust Multiple Choice Selectors},
  author={Chujie Zheng and Hao Zhou and Fandong Meng and Jie Zhou and Minlie Huang},
  booktitle={The Twelfth International Conference on Learning Representations},
  year={2024},
  url={https://openreview.net/forum?id=shr9PXz7T0}
}

Code and Data

See code for the experimental code for reproducing our experimental results.

We also release the experimental results in another repo: https://github.com/chujiezheng/LLM-MCQ-Bias_data