We introduce SelfCite, a novel self-supervised approach that aligns LLMs to generate high-quality, fine-grained, sentence-level citations for the statements in their generated responses. Instead of only relying on costly and labor-intensive annotations, SelfCite leverages a reward signal provided by the LLM itself through context ablation: If a citation is necessary, removing the cited text from the context should prevent the same response; if sufficient, retaining the cited text alone should preserve the same response. This reward can guide the inference-time best-of-N sampling strategy to improve citation quality significantly, as well as be used in preference optimization to directly fine-tune the models for generating better citations. The effectiveness of SelfCite is demonstrated by increasing citation F1 up to 5.3 points on the LongBench-Cite benchmark across five long-form question answering tasks. The source code is available at https://github.com/facebookresearch/SelfCite
@article{arxiv.2502.09604,
title = {SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models},
author = {Yung-Sung Chuang and Benjamin Cohen-Wang and Shannon Zejiang Shen and Zhaofeng Wu and Hu Xu and Xi Victoria Lin and James Glass and Shang-Wen Li and Wen-tau Yih},
journal= {arXiv preprint arXiv:2502.09604},
year = {2025}
}
Comments
ICML 2025 main conference paper. The source code is available at https://github.com/facebookresearch/SelfCite