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

2025-03-19

AI Summary

This paper addresses the lack of fine-grained provenance tracing for large language model (LLM)-generated text in high-stakes domains such as healthcare and law. We formally define and systematically model sentence-level textual provenance: localizing the original source sentence and classifying the semantic relationship between target and source—namely, quotation, compression, inference, or paraphrase. To support this task, we introduce TROVE, a high-quality, cross-lingual, multi-document, long-context benchmark featuring a four-category relation annotation schema and a three-stage hybrid annotation pipeline (GPT-assisted pre-screening, human expert refinement, and retrieval-augmented validation). Evaluating 11 LLMs via both direct prompting and RAG-based methods, we demonstrate that retrieval is critical for accurate provenance tracing. While closed-source models generally outperform open-source ones, the latter achieve substantial gains when augmented with RAG.

Technology Category

• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

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 such as healthcare, law, and news, 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 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.

Problem

• Trace sentences to specific source sentences in multi-document inputs.
• Classify fine-grained relationships between target and source sentences.
• Evaluate LLMs on provenance tasks in diverse scenarios and languages.

Innovation

• TROVE traces sentences to specific sources
• Annotates fine-grained relationships between sentences
• Uses three-stage annotation for high-quality data