Dual-Debiasing-Noisy-ICL

Implementation of Dual Debiasing for Noisy In-Context Learning for Text Generation, accepted by Findings of the Association for Computational Linguistics 2025.

Prepareration

pip install -r requirements.txt

Dataset Prepare

Unzip the ICL-Data.zip into Dual-Debiasing-Noisy-ICL/ICL-Data folder.

Run the script

Check Dual-Debiasing-Noisy-ICL/run_dual_debias_pipeline.sh for complete script demo of our dual debiasing method.

For naive ICL baseline and random-delete baseline, please check Dual-Debiasing-Noisy-ICL/run_baseline_pipeline.sh.

Citation

@inproceedings{liang-etal-2025-dual,
    title = "Dual Debiasing for Noisy In-Context Learning for Text Generation",
    author = "Liang, Siqi  and
      Ahn, Sumyeong  and
      Dhillon, Paramveer  and
      Zhou, Jiayu",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.findings-acl.665/",
    doi = "10.18653/v1/2025.findings-acl.665",
    pages = "12855--12868",
    ISBN = "979-8-89176-256-5",
}