Evaluating Verifiability in Generative Search Engines

By Nelson F. Liu, Tianyi Zhang, Percy Liang. Conference on Empirical Methods in Natural Language Processing, 2023

407 citations recorded by the source corpus.

Generative search engines directly generate responses to user queries, along with in-line citations. A prerequisite trait of a trustworthy generative search engine is verifiability, i.e., systems should cite comprehensively (high citation recall; all statements are fully supported by citations) and accurately (high citation precision; every cite supports its associated statement). We conduct human evaluation to audit four popular generative search engines -- Bing Chat, NeevaAI, perplexity.ai, and YouChat -- across a diverse set of queries from a variety of sources (e.g., historical Google user queries, dynamically-collected open-ended questions on Reddit, etc.). We find that responses from existing generative search engines are fluent and appear informative, but frequently contain unsupported statements and inaccurate citations: on average, a mere 51.5% of generated sentences are fully supported by citations and only 74.5% of citations support their associated sentence. We believe that these results are concerningly low for systems that may serve as a primary tool for information-seeking users, especially given their facade of trustworthiness. We hope that our results further motivate the development of trustworthy generative search engines and help researchers and users better understand the shortcomings of existing commercial systems.

Subfields: Artificial Intelligence, Information Systems, Sociology and Political Science

View Paper: https://www.semanticscholar.org/paper/e03dc5bd72bea7d1db03de380d4c43b7e610fbee

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 Information Systems

In Sociology and Political Science

Citations in the corpus

Benchmarking Large Language Models in Retrieval-Augmented Generation (2023, background): “However, Retrieval-Augmented Generation brings not only positive effects to LLMs (Liu, Zhang, and Liang 2023; Maynez et al. 2020).”
Opportunities and Challenges for ChatGPT and Large Language Models in Biomedicine and Health (2023, background): “However, such systems are still not free from errors [54] and there is a need for more systematic evaluations [52].”
Evaluating Very Long-Term Conversational Memory of LLM Agents (2024, background): “…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 the correct context…”
How Language Model Hallucinations Can Snowball (2023, background): “Despite their diverse capabilities and extreme fluency, a major open challenge is that LMs still hallucinate by making up facts or citing sources that do not exist (Maynez et al., 2020; Liu et al., 2023, i.a.), often while sounding extremely plausible.”
A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions (2023, unknown): “Another concern lies in the arena of generative retrievals, which occasionally suffer from citation inaccuracies (Rashkin et al.; Liu et al., 2023f; Yue et al., 2023; Gao et al., 2023a; Chen et al., 2023a).”
ExpertQA: Expert-Curated Questions and Attributed Answers (2023, unknown): “Attribution Analysis Prior work has conducted analysis of attributions produced by systems in response to queries from existing NLP datasets (Rashkin et al., 2021; Bohnet et al., 2022; Dziri et al., 2022; Liu et al., 2023; Muller et al., 2023).”
Enabling Large Language Models to Generate Text with Citations (2023, methodology): “A concurrent work (Liu et al., 2023) conduct human evaluation on commercial generative search engines like Bing Chat to examine their citation qualities.”
Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (2023, background): “The effectiveness of GPT-4 evaluations in fine-grained aspects has shown to be effective in prior work (Liu et al., 2023a).”
FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation (2023, methodology): “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).”
BITEM at the NTCIR-19 R2C2 Task: Predicting Confidence from Agentic RAG Pipeline Signals (2026, unknown): “Liu et al. [8] check after the fact, auditing four generative search engines and finding 51.5% of their statements fully supported by the citations given for them.”
SAGE-RAI: Design Patterns for Transparent RAG Systems (2026, unknown): “The scope-breadth tension validates Liu et al.’s [22] finding that citation quality affects trust: constraining knowledge bases improves attribution but reduces breadth.”
Retrieval And Structuring Augmented Generation with Large Language Models (2025, unknown): “In specialized fields such as scientific research, healthcare, or technical domains, LLMs often lack the precise and detailed knowledge required for reliable performance [50, 95, 145].”