Enabling Large Language Models to Generate Text with Citations

By Tianyu Gao, Howard Yen, Jiatong Yu, Danqi Chen. Conference on Empirical Methods in Natural Language Processing, 2023

689 citations recorded by the source corpus.

Large language models (LLMs) have emerged as a widely-used tool for information seeking, but their generated outputs are prone to hallucination. In this work, our aim is to allow LLMs to generate text with citations, improving their factual correctness and verifiability. Existing work mainly relies on commercial search engines and human evaluation, making it challenging to reproduce and compare different modeling approaches. We propose ALCE, the first benchmark for Automatic LLMs' Citation Evaluation. ALCE collects a diverse set of questions and retrieval corpora and requires building end-to-end systems to retrieve supporting evidence and generate answers with citations. We develop automatic metrics along three dimensions -- fluency, correctness, and citation quality -- and demonstrate their strong correlation with human judgements. Our experiments with state-of-the-art LLMs and novel prompting strategies show that current systems have considerable room for improvement -- For example, on the ELI5 dataset, even the best models lack complete citation support 50% of the time. Our analyses further highlight promising future directions, including developing better retrievers, advancing long-context LLMs, and improving the ability to synthesize information from multiple sources.

Subfields: Artificial Intelligence, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/e7c97e953849f1a8e5d85ceb4cfcc0a5d54d2365

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

Citations in the corpus

Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models (2023, positive): “The first step entails accurately obtaining knowledge re-16 Method Timing of Using Knowledge Source Application Task WebGPT (Nakano et al., 2021) Generation-Time Search API QA Adaptive-Retrieval (Mallen et al., 2023) Generation-Time Wikipedia QA ReACT (Yao et al., 2022) Generation-Time Wikipedia QA ”
ARES: An Automated Evaluation Framework for Retrieval-Augmented Generation Systems (2023, neutral): “New guidelines like LongEval (Krishna et al., 2023) and datasets like Hagrid and ALCE (Kamalloo et al., 2023; Gao et al., 2023) provide resources for analyzing evaluating knowledge-intensive LLM pipelines.”
Fine-Grained Human Feedback Gives Better Rewards for Language Model Training (2023, background): “Furthermore, previous research [8, 18, 11, 45, 25, 43] into automated evaluation of generated text shows that it can be challenging for human annotators to reliably compare the overall quality of two or more model outputs when the outputs contain a mixture of diverse undesired behaviors.”
RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval (2024, neutral): “The summarization/snippet model by Gao et al. (2023) uses summarizations and snippets of passages, which improves correctness on most datasets but can sometimes be a lossy means of compression.”
Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (2023, methodology): “Moreover, the output is not guaranteed to be consistent with retrieved relevant passages (Gao et al., 2023) since the models are not explicitly trained to leverage and follow facts from provided passages.”
Instruction Tuning for Large Language Models: A Survey (2023, neutral): “LLMs such as GPT-3 [13], PaLM [22], and LLaMA [113] have demonstrated impressive capabilities across a wide range of natural language tasks [1, 19, 38, 40, 41, 59, 60, 75, 76, 91, 104–106, 115, 117, 118, 126, 127, 135, 136, 150].”
Improving Text Embeddings with Large Language Models (2023, background): “Source attribution of generated text is another important application of text embeddings [13] that can improve the interpretability and trustworthiness of LLMs. Previous studies have demonstrated that weighted average of pre-trained word embeddings [34, 1] is a strong baseline for measuring…”
Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts (2023, background): “Such parametric memory could help solve downstream tasks (Wang et al., 2021; Yu et al., 2023; Sun et al., 2023).”
FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation (2023, background): “A series of concurrent work verifies the precision of the citations (attributions) provided by the model (Gao et al., 2022; Liu et al., 2023a; Yue et al., 2023; Gao et al., 2023).”
Evaluating Very Long-Term Conversational Memory of LLM Agents (2024, background): “…co-references and missing content ( i.e., anaphora) (Anantha et al., 2021), which further complicate the retrieval process (Mallen et al., 2023; Gao et al., 2023b; Liu et al., 2023b); (2) Challenges arise in reasoning over retrieved documents, especially when the model struggles to identify…”
Tool learning with large language models: a survey (2024, neutral): “Additionally, the integration of advanced natural language processing tools can augment the linguistic understanding of LLMs, thereby optimizing dialogue management and intent recognition [18, 89, 90].”
MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (2024, background): “RAG improves LLM’s response (Borgeaud et al., 2022) and also mitigates the occurrence of hallucinations, thereby enhancing the models’ credibility (Gao et al., 2023).”
GigaEmbeddings: Efficient Russian Language Embedding Model (2025, unknown): “Their role extends to retrieval-augmented generation (RAG) (Lewis et al., 2020), where they dynamically ground large language models (LLMs) in external knowledge, and to source attribution frameworks (Gao et al., 2023), enhancing the transparency of LLM outputs.”
Deep Literature Survey Automation with an Iterative Workflow (2025, unknown): “Citation quality is typically measured with NLI-based protocols (Gao et al., 2023), and Yan et al. (2025) additionally evaluate coverage by comparing system outputs with human-written surveys.”
RAS: Retrieval-And-Structuring for Knowledge-Intensive LLM Generation (2025, unknown): “For ASQA and ELI5, we maintain methodological consistency with prior work (Lyu et al., 2024b; Asai et al., 2023; Gao et al., 2023), utilizing their predetermined five-document context.”