CiteAnchor: Source-Aware Grounding for Hallucinated Citation Detection and Resolution
Computer Science > Computation and Language arXiv:2610.30355 (cs) [Submitted on 21 Oct 2026] Title:CiteAnchor: Source-Aware Grounding for Hallucinated Citation Detection and Resolution Authors:Chenyi Tong, Ping Nie, Yu Zhang View a PDF of the paper titled CiteAnchor: Source-Aware Grounding for Hallucinated Citation Detection and Resolution, by Chenyi Tong and 2 other authors View PDF HTML (experimental)
计算机科学 > 计算与语言 arXiv:2610.30355 (cs) [提交于 2026 年 10 月 21 日] 标题：CiteAnchor：用于幻觉引用检测与解析的源感知基础（Source-Aware Grounding） 作者：Chenyi Tong, Ping Nie, Yu Zhang 查看标题为《CiteAnchor：用于幻觉引用检测与解析的源感知基础》的论文 PDF，作者：Chenyi Tong 及其他 2 位作者 查看 PDF HTML（实验性）
Abstract:Hallucinated citations make it harder to trace sources and verify the evidence behind scientific claims. Detecting them is difficult because bibliographic records can be incomplete or differ across source types and publication versions. A plausible match alone neither validates the citation nor establishes its underlying source. We introduce CiteResolve, a benchmark of 423 real-world hallucinated citations annotated for source recoverability. We further propose CiteAnchor, a source-aware grounding framework that adapts retrieval and field-level verification to each source type and separately determines citation validity and whether a unique source can be recovered.
摘要：幻觉引用使得追踪来源和验证科学主张背后的证据变得更加困难。检测这些引用之所以困难，是因为书目记录可能不完整，或者在不同来源类型和出版版本之间存在差异。仅凭一个看似合理的匹配既不能验证引用，也不能确定其潜在来源。我们引入了 CiteResolve，这是一个包含 423 个真实世界幻觉引用的基准，并标注了来源的可恢复性。我们进一步提出了 CiteAnchor，这是一个源感知基础框架，它针对每种来源类型调整检索和字段级验证，并分别确定引用的有效性以及是否可以恢复唯一来源。
On an existing detection benchmark, CiteAnchor achieves 0.841 F1 with GPT-5.6 Luna and 0.784 with an open-weight backbone, outperforming the baseline with the same backbones by 7.9 and 36.8 points. On CiteResolve, it achieves 74.7% and 70.7% end-to-end task success, outperforming the corresponding web search baselines by 16.3 and 65.3 percentage points. Our code and data are available at this https URL.
在现有的检测基准上，CiteAnchor 在使用 GPT-5.6 Luna 时达到了 0.841 的 F1 分数，在使用开源权重主干模型时达到了 0.784，比使用相同主干模型的基线分别高出 7.9 和 36.8 个百分点。在 CiteResolve 上，它实现了 74.7% 和 70.7% 的端到端任务成功率，比相应的网络搜索基线分别高出 16.3 和 65.3 个百分点。我们的代码和数据可在该 HTTPS 链接获取。