# TROVE: A Challenge for Fine-Grained Text Provenance via Source Sentence Tracing and Relationship Classification

arXiv: https://arxiv.org/abs/2503.15289 · HTML: https://mlsanity.com/paper/2503.15289/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 36 references that are in the map.
- First posted on arXiv 19 Mar 2025.

## Builds on (most cited first, 20 of 20)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding](https://mlsanity.com/paper/2308.14508/) (2023)
- [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://mlsanity.com/paper/2405.04434/) (2024)
- [Enabling Large Language Models to Generate Text with Citations](https://mlsanity.com/paper/2305.14627/) (2023)
- [Sparse, Dense, and Attentional Representations for Text Retrieval](https://mlsanity.com/paper/2005.00181/) (2020)
- [Complex Claim Verification with Evidence Retrieved in the Wild](https://mlsanity.com/paper/2305.11859/) (2023)
- [Reasoning Over Semantic-Level Graph for Fact Checking](https://mlsanity.com/paper/1909.03745/) (2019)
- [Attribute First, then Generate: Locally-attributable Grounded Text Generation](https://mlsanity.com/paper/2403.17104/) (2024)
- [GERE: Generative Evidence Retrieval for Fact Verification](https://mlsanity.com/paper/2204.05511/) (2022)
- [Knowledge-Grounded Dialogue Generation with a Unified Knowledge Representation](https://mlsanity.com/paper/2112.07924/) (2021)
- [MultiCite: Modeling realistic citations requires moving beyond the single-sentence single-label setting](https://mlsanity.com/paper/2107.00414/) (2021)
- [Verifiable Generation with Subsentence-Level Fine-Grained Citations](https://mlsanity.com/paper/2406.06125/) (2024)
- [Learning to Generate Answers with Citations via Factual Consistency Models](https://mlsanity.com/paper/2406.13124/) (2024)
- [CaLM: Contrasting Large and Small Language Models to Verify Grounded Generation](https://mlsanity.com/paper/2406.05365/) (2024)
- [Towards Abstractive Grounded Summarization of Podcast Transcripts](https://mlsanity.com/paper/2203.11425/) (2022)
- [Analysis of Plan-based Retrieval for Grounded Text Generation](https://mlsanity.com/paper/2408.10490/) (2024)
- [Decker: Double Check with Heterogeneous Knowledge for Commonsense Fact Verification](https://mlsanity.com/paper/2305.05921/) (2023)
- [Grounded Keys-to-Text Generation: Towards Factual Open-Ended Generation](https://mlsanity.com/paper/2212.01956/) (2022)
- [Contextualizing Generated Citation Texts](https://mlsanity.com/paper/2402.18054/) (2024)

## Built on by (most cited first, 4 of 4)

- [TRACER: Verifiable Generative Provenance for Multimodal Tool-Using Agents](https://mlsanity.com/paper/2605.09934/) (2026)
- [GenProve: Learning to Generate Text with Fine-Grained Provenance](https://mlsanity.com/paper/2601.04932/) (2026)
- [Citation Failure: Definition, Analysis and Efficient Mitigation](https://mlsanity.com/paper/2510.20303/) (2025)
- [Fine-grained Verbal Attack Detection via a Hierarchical Divide-and-Conquer Framework](https://mlsanity.com/paper/2601.06907/) (2026)

## Abstract (arXiv)

LLMs have achieved remarkable fluency and coherence in text generation, yet their widespread adoption has raised concerns about content reliability and accountability. In high-stakes domains, it is crucial to understand where and how the content is created. To address this, we introduce the Text pROVEnance (TROVE) challenge, designed to trace each sentence of a target text back to specific source sentences within potentially lengthy or multi-document inputs. Beyond identifying sources, TROVE annotates the fine-grained relationships (quotation, compression, inference, and others), providing a deep understanding of how each target sentence is formed. To benchmark TROVE, we construct our dataset by leveraging three public datasets covering 11 diverse scenarios (e.g., QA and summarization) in English and Chinese, spanning source texts of varying lengths (0-5k, 5-10k, 10k+), emphasizing the multi-document and long-document settings essential for provenance. To ensure high-quality data, we employ a three-stage annotation process: sentence retrieval, GPT-4o provenance, and human provenance. We evaluate 11 LLMs under direct prompting and retrieval-augmented paradigms, revealing that retrieval is essential for robust performance, larger models perform better in complex relationship classification, and closed-source models often lead, yet open-source models show significant promise, particularly with retrieval augmentation. We make our dataset available here: https://github.com/ZNLP/ZNLP-Dataset.