naacl25-prompt-compositions_finetune-baseline

Adaptive Prompting: Ad-hoc Prompt Composition for Social Bias Detection

Finetune baseline models for the paper Adaptive Prompting: Ad-hoc Prompt Composition for Social Bias Detection.

For details, please see the published paper and the GitHub repository.

@inproceedings{spliethover-etal-2025-adaptive,
    title        = {Adaptive Prompting: Ad-hoc Prompt Composition for Social Bias Detection},
    author       = {Splieth{\"o}ver, Maximilian  and Knebler, Tim  and Fumagalli, Fabian  and Muschalik, Maximilian  and Hammer, Barbara  and H{\"u}llermeier, Eyke  and Wachsmuth, Henning},
    year         = 2025,
    month        = apr,
    booktitle    = {Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)},
    publisher    = {Association for Computational Linguistics},
    address      = {Albuquerque, New Mexico},
    pages        = {2421--2449},
    isbn         = {979-8-89176-189-6},
    url          = {https://aclanthology.org/2025.naacl-long.122/},
    editor       = {Chiruzzo, Luis and Ritter, Alan and Wang, Lu}
}

Note on finetune baseline models

Unfortunately, we did not keep the original finetuning baseline models, for which scores are reported in the paper. We did, however, keep the prediction results of these models.

We did retrain the models on the same splits, same seeds, same python version, and same library versions. The new models and also the new (and old) prediction results are uploaded in this repository.