SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chuang, Yung-Sung, Cohen-Wang, Benjamin, Shen, Shannon Zejiang, Wu, Zhaofeng, Xu, Hu, Lin, Xi Victoria, Glass, James, Li, Shang-Wen, Yih, Wen-tau
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913893271470080
author Chuang, Yung-Sung
Cohen-Wang, Benjamin
Shen, Shannon Zejiang
Wu, Zhaofeng
Xu, Hu
Lin, Xi Victoria
Glass, James
Li, Shang-Wen
Yih, Wen-tau
author_facet Chuang, Yung-Sung
Cohen-Wang, Benjamin
Shen, Shannon Zejiang
Wu, Zhaofeng
Xu, Hu
Lin, Xi Victoria
Glass, James
Li, Shang-Wen
Yih, Wen-tau
contents 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
format Preprint
id arxiv_https___arxiv_org_abs_2502_09604
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models
Chuang, Yung-Sung
Cohen-Wang, Benjamin
Shen, Shannon Zejiang
Wu, Zhaofeng
Xu, Hu
Lin, Xi Victoria
Glass, James
Li, Shang-Wen
Yih, Wen-tau
Computation and Language
Artificial Intelligence
Machine Learning
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
title SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2502.09604