# AtomCite: Verification and Correction of Supplied Page-Level Citations in Multi-Page Documents

arXiv: https://arxiv.org/abs/2609.05802 · HTML: https://mlsanity.com/paper/2609.05802/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 28 references that are in the map.
- First posted on arXiv 5 Sep 2026.

## Builds on (most cited first, 19 of 19)

- [DocVQA: A Dataset for VQA on Document Images](https://mlsanity.com/paper/2007.00398/) (2020)
- [FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation](https://mlsanity.com/paper/2305.14251/) (2023)
- [Enabling Large Language Models to Generate Text with Citations](https://mlsanity.com/paper/2305.14627/) (2023)
- [SummaC: Re-Visiting NLI-based Models for Inconsistency Detection in Summarization](https://mlsanity.com/paper/2111.09525/) (2021)
- [Evaluating Verifiability in Generative Search Engines](https://mlsanity.com/paper/2304.09848/) (2023)
- [AlignScore: Evaluating Factual Consistency with a Unified Alignment Function](https://mlsanity.com/paper/2305.16739/) (2023)
- [Measuring Attribution in Natural Language Generation Models](https://mlsanity.com/paper/2112.12870/) (2021)
- [Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools](https://mlsanity.com/paper/2405.20362/) (2024)
- [Document Understanding Dataset and Evaluation (DUDE)](https://mlsanity.com/paper/2305.08455/) (2023)
- [LettuceDetect: A Hallucination Detection Framework for RAG Applications](https://mlsanity.com/paper/2502.17125/) (2025)
- [MinerU2.5-Pro: Pushing the Limits of Data-Centric Document Parsing at Scale](https://mlsanity.com/paper/2604.04771/) (2026)
- [AttributionBench: How Hard is Automatic Attribution Evaluation?](https://mlsanity.com/paper/2402.15089/) (2024)
- [CiteEval: Principle-Driven Citation Evaluation for Source Attribution](https://mlsanity.com/paper/2506.01829/) (2025)
- [CiteFix: Enhancing RAG Accuracy Through Post-Processing Citation Correction](https://mlsanity.com/paper/2504.15629/) (2025)
- [Generation-Time vs. Post-hoc Citation: A Holistic Evaluation of LLM Attribution](https://mlsanity.com/paper/2509.21557/) (2025)
- [HalluMix: A Task-Agnostic, Multi-Domain Benchmark for Real-World Hallucination Detection](https://mlsanity.com/paper/2505.00506/) (2025)
- [DocScope: Benchmarking Verifiable Reasoning for Trustworthy Long-Document Understanding](https://mlsanity.com/paper/2605.08888/) (2026)
- [The Gray Zone of Faithfulness: Taming Ambiguity in Unfaithfulness Detection](https://mlsanity.com/paper/2510.21118/) (2025)
- [SG-LegalCite: A Principle-Augmented Benchmark for Legal Citation Retrieval in Singapore Law](https://mlsanity.com/paper/2605.21057/) (2026)

## Abstract (arXiv)

Large language models answering questions over multi-page documents are expected to cite the supporting pages, yet supplied citations are sometimes inaccurate, and current evaluations score citations at generation time or against text passages: no existing benchmark evaluates whether a system can verify and correct a page-level citation already attached to an answer. We propose AtomCite, an agentic framework that parses an answer into claims, checks each claim against the image of its cited page, and applies a deterministic repair policy. To evaluate it, we introduce DocCite, to our knowledge the first benchmark for systems that verify and correct page-level citations in document images. Built on MP-DocVQA and DUDE, it combines 928 validated injected instances with 2,468 candidate natural errors harvested from frontier- and efficiency-tier models, of which a two-annotator audit confirms 1,909 as genuine errors. Primary labels are assigned deterministically, not by LLM judges, with the human audit as a separate validation layer. Across three model families (Gemini, Claude, and GPT), AtomCite reaches around 93% binary verification accuracy on the injected benchmark, significantly outperforming every OCR-only condition, including a compute-matched control, and exceeding every prior text-based baseline given the same OCR text. Its repair policy lifts citation precision on the injected mix from a constructed 34% to 87-90% while retaining over 90% of correct claims. AtomCite also transfers: with frozen prompts and zero training, it raises the hallucination-detection scores of two open 7-8B models on five public benchmarks above the same models prompted as direct judges. Finally, the audit shows that noise in automatic labels biases measured verifier accuracy and can reverse system rankings, so evaluations relying only on synthetic or automatic labels risk mismeasuring verification capability.