Evaluating Attribution in Dialogue Systems: The BEGIN Benchmark

By Nouha Dziri, Hannah Rashkin, Tal Linzen, David Reitter. 2021

102 citations recorded by the source corpus.

Abstract Knowledge-grounded dialogue systems powered by large language models often generate responses that, while fluent, are not attributable to a relevant source of information. Progress towards models that do not exhibit this issue requires evaluation metrics that can quantify its prevalence. To this end, we introduce the Benchmark for Evaluation of Grounded INteraction (Begin), comprising 12k dialogue turns generated by neural dialogue systems trained on three knowledge-grounded dialogue corpora. We collect human annotations assessing the extent to which the models’ responses can be attributed to the given background information. We then use Begin to analyze eight evaluation metrics. We find that these metrics rely on spurious correlations, do not reliably distinguish attributable abstractive responses from unattributable ones, and perform substantially worse when the knowledge source is longer. Our findings underscore the need for more sophisticated and robust evaluation metrics for knowledge-grounded dialogue. We make Begin publicly available at https://github.com/google/BEGIN-dataset.

Subfields: Artificial Intelligence, Natural Language Processing, Evaluation & Benchmarks

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 Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

Measuring Attribution in Natural Language Generation Models (2021, methodology): “These have been measured via dialogue-specific natural language inference methods, often via human studies and data creation (Welleck et al. 2019; Mehri and Eskenazi 2020; Gupta et al. 2021; Honovich et al. 2021; Dziri et al. 2021; Santhanam et al. 2021).”
Survey of Hallucination in Natural Language Generation (2022, background): “[31] propose the BEGIN benchmark, which consists of samples taken from Dinan et al.”
ExpertQA: Expert-Curated Questions and Attributed Answers (2023, methodology): “…have been conducted by repurposing Wikipedia citations (Kamoi et al., 2023; Petroni et al., 2022), or existing datasets like Natural Questions (Bohnet et al., 2022; Kwiatkowski et al., 2019), and relying on large-scale human annotation (Chen et al., 2022; Dziri et al., 2022; Kamalloo et al., 2023).”
RAGTruth: A Hallucination Corpus for Developing Trustworthy Retrieval-Augmented Language Models (2023, background): “In different contexts, the definition and criteria for hallucination vary, and the annotation of hallucina-310 tion is not always straightforward.”
AlignScore: Evaluating Factual Consistency with A Unified Alignment Function (2023, background): “2021), MNBM (Maynez et al., 2020), QAGSCNNDM (Wang et al., 2020), QAGS-XSum (Wang et al., 2020), BEGIN (Dziri et al., 2022), Q2dataset (Honovich et al., 2021), DialFact (Gupta et al., 2022), PAWS (Zhang et al., 2019), FEVER (Nie et al., 2019b; Thorne et al., 2018) and VitaminC (Schuster et al., 2021”
DialFact: A Benchmark for Fact-Checking in Dialogue (2021, background): “recent works have proposed evaluation metrics and benchmarks for factual consistency in knowledge grounded response generation (Honovich et al., 2021; Dziri et al., 2021b).”
Large Language Models: A Survey (2024, unknown): “• NLI-Based Metrics : Use Natural Language Inference datasets to evaluate the truthfulness of a generated hypothesis based on a given premise (see [153]).”
HaluEval: A Large-Scale Hallucination Evaluation Benchmark for Large Language Models (2023, background): “For instance, The BEGIN benchmark (Dziri et al., 2022b) classifies the utterances generated by dialogue systems into three categories, i.e., fully attributable, not fully attributable, and generic; and the Attributable to Identified Sources (AIS) benchmark (Rashkin et al., 2021) assesses whether…”