@inproceedings{zhang-etal-2026-context-fidelity,
    title = "Context-Fidelity Boosting: Enhancing Faithful Generation through Watermark-Inspired Decoding",
    author = "Zhang, Weixu  and
      Ye, Fanghua  and
      Gao, Qiang  and
      Li, Jian  and
      Wu, Haolun  and
      Tian, Yuxing  and
      Duan, Sijing  and
      Du, Nan  and
      Li, Xiaolong",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.findings-acl.2121/",
    pages = "42748--42759",
    ISBN = "979-8-89176-395-1",
    abstract = "Large language models (LLMs) often produce content that contradicts or overlooks information provided in the input context, a phenomenon known as faithfulness hallucination. In this paper, we propose Context-Fidelity Boosting (CFB), a lightweight and general decoding-time framework that effectively reduces such hallucinations by boosting the generation probability of context-relevant tokens. Motivated by logit-shaping principles in watermarking techniques, CFB leverages token-level logit adjustments based on their presence or salience in the input context. Specifically, we develop three boosting strategies, static, context-aware, and token-aware that progressively incorporate distributional divergence, attention scores, and semantic similarity. Notably, CFB requires no retraining or architectural changes, making it compatible with a wide range of LLMs. Experiments on summarization and question answering tasks across multiple open-source LLMs show that CFB consistently improves faithfulness metrics, with minimal generation overhead. Our implementation is fully open-sourced."
}