Source-Aware Training Enables Knowledge Attribution in Language Models

2024/04/01 by Muhammad Khalifa, Khalifa, Muhammad, David Wadden +11 · 4 citations
Computer Science · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2404.01019

openalex publication_date 2024/04/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Large language models (LLMs) learn a vast amount of knowledge during pretraining, but they are often oblivious to the source(s) of such knowledge. We investigate the problem of intrinsic source citation, where LLMs are required to cite the pretraining source supporting a generated response. Intrinsic source citation can enhance LLM transparency, interpretability, and verifiability. To give LLMs such ability, we explore source-aware training -- a recipe that involves (i) training the LLM to associate unique source document identifiers with the knowledge in each document, followed by (ii) an instruction-tuning stage to teach the LLM to cite a supporting pretraining source when prompted. Source-aware training borrows from existing pretraining/fine-tuning frameworks and requires minimal changes to the model architecture or implementation. Through experiments on synthetic data, we demonstrate that our training recipe can enable faithful attribution to the pretraining data without a substantial impact on the model's perplexity compared to standard pretraining. Our findings also highlight the importance of pretraining data augmentation in achieving attribution. Code and data available here: \urlhttps://github.com/mukhal/intrinsic-source-citation

Cited by

• CAGE: Cognitive Attribution Graphs for Faithful Inline Citation Generation in Long-Form Question Answering
• LIME: Making LLM Data More Efficient with Linguistic Metadata Embeddings
• Beyond URLs: Metadata Diversity and Position for Efficient LLM Pretraining
• Rule-Based Explanations for Retrieval-Augmented LLM Systems
• The SMeL Test: A simple benchmark for media literacy in language models

Related

• Cite Pretrain: Retrieval-Free Knowledge Attribution for Large Language Models
• Source Attribution for Large Language Model-Generated Data
• Advancing Large Language Model Attribution through Self-Improving
• Scalable Influence and Fact Tracing for Large Language Model Pretraining
• Identifying the Source of Generation for Large Language Models