Can Large Language Models Capture Dissenting Human Voices? (EMNLP 2023)

Authors: Noah Lee* Na Min An* James Thorne

• We test generative LLMs jointly on the performance and human disagreement on NLI.
• We suggest two probability distribution estimation techniques for LLMs to represent disagreement and perform empirical evaluations to with respect to the human disagreement distribution.
• LLMs do not excel as expected on NLI tasks and fail to align with human disagreement levels.

Environment Setting

conda create -n humllm
conda activate humllm
pip install -r requirements.txt

Datasets

The datasets used for the research are as the following:

• ChaosNLI
• QNLI
• ANLI
• PK2019
• NLI_Disagreement

Usage

• All the script examples can be found in ./scripts/

Preprocess & Sample

Sample random 100 samples & hardest 100 samples

bash ./scripts/sample.sh

Generate

Generation of a LLM output can be done by bash ./scripts/generate.sh or either:

python generate.py --data_dir <input data directory> \
                    --data_type <input data type> \
                    --model <model name> \
                    --file_name <output file name> \
                    --out_dir <output directory> \
                    --max_length <maximum token lengths> \
                    --gen_type <generation type> \
                    --num_iter <iteration number> \ 
                    --num_samples <sample number> # num_iter x num_samples = total sample size \
                    --prompt_variations <use prompt variations> \
                    --few_shot <few shot number>

Evaluate

Evaluation of generated distribution is available by bash ./scripts/evaluate.sh or either:

python evaluate.py --data_dir <input data directory> \
                    --data_type <input data type> \
                    --gen_type <generation type>

Citation

Please consider citing our work if you find this work helpful for your research.

@misc{lee2023large,
      title={Can Large Language Models Capture Dissenting Human Voices?}, 
      author={Noah Lee and Na Min An and James Thorne},
      year={2023},
      eprint={2305.13788},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Contact

• Noah Lee: noah.lee@kaist.ac.kr
• Na Min An: naminan@kaist.ac.kr